Stage verified Inflect v2 release candidate and evaluation evidence

.gitattributes

@@ -33,3 +33,10 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+samples/male/conversational.wav filter=lfs diff=lfs merge=lfs -text
+samples/male/names_places.wav filter=lfs diff=lfs merge=lfs -text
+samples/male/numbers.wav filter=lfs diff=lfs merge=lfs -text
+samples/male/punctuation.wav filter=lfs diff=lfs merge=lfs -text
+samples/male/technical.wav filter=lfs diff=lfs merge=lfs -text
+smoke/release_smoke.wav filter=lfs diff=lfs merge=lfs -text
+smoke/standalone_cpu.wav filter=lfs diff=lfs merge=lfs -text

LICENSE

@@ -0,0 +1,201 @@
+Apache License
+Version 2.0, January 2004
+http://www.apache.org/licenses/
+
+TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+1. Definitions.
+
+"License" shall mean the terms and conditions for use, reproduction,
+and distribution as defined by Sections 1 through 9 of this document.
+
+"Licensor" shall mean the copyright owner or entity authorized by
+the copyright owner that is granting the License.
+
+"Legal Entity" shall mean the union of the acting entity and all
+other entities that control, are controlled by, or are under common
+control with that entity. For the purposes of this definition,
+"control" means (i) the power, direct or indirect, to cause the
+direction or management of such entity, whether by contract or
+otherwise, or (ii) ownership of fifty percent (50%) or more of the
+outstanding shares, or (iii) beneficial ownership of such entity.
+
+"You" (or "Your") shall mean an individual or Legal Entity
+exercising permissions granted by this License.
+
+"Source" form shall mean the preferred form for making modifications,
+including but not limited to software source code, documentation
+source, and configuration files.
+
+"Object" form shall mean any form resulting from mechanical
+transformation or translation of a Source form, including but
+not limited to compiled object code, generated documentation,
+and conversions to other media types.
+
+"Work" shall mean the work of authorship, whether in Source or
+Object form, made available under the License, as indicated by a
+copyright notice that is included in or attached to the work
+(an example is provided in the Appendix below).
+
+"Derivative Works" shall mean any work, whether in Source or Object
+form, that is based on (or derived from) the Work and for which the
+editorial revisions, annotations, elaborations, or other modifications
+represent, as a whole, an original work of authorship. For the purposes
+of this License, Derivative Works shall not include works that remain
+separable from, or merely link (or bind by name) to the interfaces of,
+the Work and Derivative Works thereof.
+
+"Contribution" shall mean any work of authorship, including
+the original version of the Work and any modifications or additions
+to that Work or Derivative Works thereof, that is intentionally
+submitted to Licensor for inclusion in the Work by the copyright owner
+or by an individual or Legal Entity authorized to submit on behalf of
+the copyright owner. For the purposes of this definition, "submitted"
+means any form of electronic, verbal, or written communication sent
+to the Licensor or its representatives, including but not limited to
+communication on electronic mailing lists, source code control systems,
+and issue tracking systems that are managed by, or on behalf of, the
+Licensor for the purpose of discussing and improving the Work, but
+excluding communication that is conspicuously marked or otherwise
+designated in writing by the copyright owner as "Not a Contribution."
+
+"Contributor" shall mean Licensor and any individual or Legal Entity
+on behalf of whom a Contribution has been received by Licensor and
+subsequently incorporated within the Work.
+
+2. Grant of Copyright License. Subject to the terms and conditions of
+this License, each Contributor hereby grants to You a perpetual,
+worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+copyright license to reproduce, prepare Derivative Works of,
+publicly display, publicly perform, sublicense, and distribute the
+Work and such Derivative Works in Source or Object form.
+
+3. Grant of Patent License. Subject to the terms and conditions of
+this License, each Contributor hereby grants to You a perpetual,
+worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+(except as stated in this section) patent license to make, have made,
+use, offer to sell, sell, import, and otherwise transfer the Work,
+where such license applies only to those patent claims licensable
+by such Contributor that are necessarily infringed by their
+Contribution(s) alone or by combination of their Contribution(s)
+with the Work to which such Contribution(s) was submitted. If You
+institute patent litigation against any entity (including a
+cross-claim or counterclaim in a lawsuit) alleging that the Work
+or a Contribution incorporated within the Work constitutes direct
+or contributory patent infringement, then any patent licenses
+granted to You under this License for that Work shall terminate
+as of the date such litigation is filed.
+
+4. Redistribution. You may reproduce and distribute copies of the
+Work or Derivative Works thereof in any medium, with or without
+modifications, and in Source or Object form, provided that You
+meet the following conditions:
+
+(a) You must give any other recipients of the Work or
+Derivative Works a copy of this License; and
+
+(b) You must cause any modified files to carry prominent notices
+stating that You changed the files; and
+
+(c) You must retain, in the Source form of any Derivative Works
+that You distribute, all copyright, patent, trademark, and
+attribution notices from the Source form of the Work,
+excluding those notices that do not pertain to any part of
+the Derivative Works; and
+
+(d) If the Work includes a "NOTICE" text file as part of its
+distribution, then any Derivative Works that You distribute must
+include a readable copy of the attribution notices contained
+within such NOTICE file, excluding those notices that do not
+pertain to any part of the Derivative Works, in at least one
+of the following places: within a NOTICE text file distributed
+as part of the Derivative Works; within the Source form or
+documentation, if provided along with the Derivative Works; or,
+within a display generated by the Derivative Works, if and
+wherever such third-party notices normally appear. The contents
+of the NOTICE file are for informational purposes only and
+do not modify the License. You may add Your own attribution
+notices within Derivative Works that You distribute, alongside
+or as an addendum to the NOTICE text from the Work, provided
+that such additional attribution notices cannot be construed
+as modifying the License.
+
+You may add Your own copyright statement to Your modifications and
+may provide additional or different license terms and conditions
+for use, reproduction, or distribution of Your modifications, or
+for any such Derivative Works as a whole, provided Your use,
+reproduction, and distribution of the Work otherwise complies with
+the conditions stated in this License.
+
+5. Submission of Contributions. Unless You explicitly state otherwise,
+any Contribution intentionally submitted for inclusion in the Work
+by You to the Licensor shall be under the terms and conditions of
+this License, without any additional terms or conditions.
+Notwithstanding the above, nothing herein shall supersede or modify
+the terms of any separate license agreement you may have executed
+with Licensor regarding such Contributions.
+
+6. Trademarks. This License does not grant permission to use the trade
+names, trademarks, service marks, or product names of the Licensor,
+except as required for reasonable and customary use in describing the
+origin of the Work and reproducing the content of the NOTICE file.
+
+7. Disclaimer of Warranty. Unless required by applicable law or
+agreed to in writing, Licensor provides the Work (and each
+Contributor provides its Contributions) on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+implied, including, without limitation, any warranties or conditions
+of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+PARTICULAR PURPOSE. You are solely responsible for determining the
+appropriateness of using or redistributing the Work and assume any
+risks associated with Your exercise of permissions under this License.
+
+8. Limitation of Liability. In no event and under no legal theory,
+whether in tort (including negligence), contract, or otherwise,
+unless required by applicable law (such as deliberate and grossly
+negligent acts) or agreed to in writing, shall any Contributor be
+liable to You for damages, including any direct, indirect, special,
+incidental, or consequential damages of any character arising as a
+result of this License or out of the use or inability to use the
+Work (including but not limited to damages for loss of goodwill,
+work stoppage, computer failure or malfunction, or any and all
+other commercial damages or losses), even if such Contributor
+has been advised of the possibility of such damages.
+
+9. Accepting Warranty or Additional Liability. While redistributing
+the Work or Derivative Works thereof, You may choose to offer,
+and charge a fee for, acceptance of support, warranty, indemnity,
+or other liability obligations and/or rights consistent with this
+License. However, in accepting such obligations, You may act only
+on Your own behalf and on Your sole responsibility, not on behalf
+of any other Contributor, and only if You agree to indemnify,
+defend, and hold each Contributor harmless for any liability
+incurred by, or claims asserted against, such Contributor by reason
+of your accepting any such warranty or additional liability.
+
+END OF TERMS AND CONDITIONS
+
+APPENDIX: How to apply the Apache License to your work.
+
+To apply the Apache License to your work, attach the following
+boilerplate notice, with the fields enclosed by brackets "{}"
+replaced with your own identifying information. (Don't include
+the brackets.) The text should be enclosed in the appropriate
+comment syntax for the file format. We also recommend that a
+file or class name and description of purpose be included on the
+same "printed page" as the copyright notice for easier
+identification within third-party archives.
+
+Copyright {yyyy} {name of copyright owner}
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.

README.md

@@ -1,169 +1,154 @@
----
-language:
-  - en
-pipeline_tag: text-to-speech
-license: other
-tags:
-  - text-to-speech
-  - tts
-  - speech
-  - local
-  - cpu
-  - onnx
-  - small-model
-  - experimental
----
-
-# Inflect-Micro-v2
-
-> **Private release preview.** The selected male checkpoint is frozen for final
-> packaging. Female weights, inference files, samples, and release evaluations
-> will be added before the repository is made public.
-
-**A complete 24 kHz English text-to-speech model in 9.36 million inference
-parameters. The waveform decoder is included.**
-
-Inflect-Micro-v2 is the quality-focused member of the Inflect v2 family. It is
-being built for local assistants, games, embedded applications, and research on
-high-quality TTS under a strict 10M-parameter limit.
-
-[Inflect-Nano-v2](https://huggingface.co/owensong/Inflect-Nano-v2) -
-[Inflect-Nano-v1](https://huggingface.co/owensong/Inflect-Nano-v1)
-
-## Current Release Candidate
-
-| Property | Male candidate | Female candidate |
-| --- | ---: | ---: |
-| Selected checkpoint | 180K polish | Pending |
-| Total deployable parameters | **9,356,513** | Target: same architecture |
-| Canonical FP32 size | **37.47 MB** | Pending |
-| FP16 size | **18.75 MB** | Pending validation |
-| Sample rate | **24 kHz** | **24 kHz planned** |
-| Input | English text | English text |
-| Output | Waveform | Waveform |
-| External vocoder required | No | No |
-
-The parameter count describes the inference-only text-to-waveform model. It does
-not include optimizer state, discriminators, or training-only posterior paths,
-and it does not hide a separate vocoder.
-
-## Architecture
-
-Inflect-Micro-v2 uses a compact VITS-derived, single-stage text-to-waveform
-backbone configured specifically for the sub-10M regime:
-
-- a normalized English phoneme frontend with stress and preserved punctuation
-- a deterministic duration predictor for compact, stable deployment
-- monotonic alignment between phonemes and acoustic frames
-- a residual-coupling latent flow and posterior encoder during training
-- an integrated HiFi-GAN-style waveform decoder; no external vocoder is needed
-- multi-resolution STFT supervision during waveform training
-
-The selected checkpoint does not use the separate experimental HierJoint or
-HierNAR prosody-planner branches evaluated during development. Those models are
-not part of this release candidate.
-
-### Core Configuration
-
-| Setting | Value |
-| --- | ---: |
-| Latent channels | 192 |
-| Text hidden channels | 96 |
-| Feed-forward channels | 768 |
-| Text encoder layers | 3 |
-| Attention heads | 2 |
-| Posterior encoder layers | 3 |
-| Initial waveform-decoder channels | 320 |
-| Segment size | 16,384 samples |
-| Upsampling | 8 x 8 x 2 x 2 |
-
-## Evaluation Preview
-
-These measurements are **provisional development results**, not final release
-claims. The full matrix will be rerun from the exact downloadable male and
-female artifacts with published prompts, hashes, generation settings, raw
-outputs, and confidence intervals.
-
-| Metric | Current male 180K | Protocol |
-| --- | ---: | --- |
-| UTMOS22 | **4.370** | 200 matched prompts; automated MOS predictor |
-| Semantic WER | **3.55%** | Whisper-large-v3; 48 difficult OOD prompts |
-| FP32 artifact size | **37.47 MB** | Inference-only Safetensors |
-
-The current semantic-WER confidence interval overlaps the compact Piper and
-Kitten baselines. The supported interpretation is **competitive intelligibility
-on this prompt set**, not a definitive win.
-
-An earlier public listening study collected 196 pairwise decisions, but it did
-not record the precise Micro checkpoint in its metadata. It will not be used as
-evidence for this 180K release candidate. A new exact-artifact study is planned.
-
-## Planned Release Files
-
-```text
-weights/
-  male/model.fp32.safetensors
-  male/model.fp16.safetensors
-  female/model.fp32.safetensors
-  female/model.fp16.safetensors
-onnx/
-  male/model.onnx
-  female/model.onnx
-examples/
-evaluation/
-voices.json
-inference.py
-app.py
-```
-
-FP32 Safetensors will be the canonical quality artifact. FP16 and ONNX will only
-be released after waveform-level parity and listening checks. INT8/GGUF formats
-are not planned for launch because they require their own audio-quality gate.
-
-## Intended Use
-
-- local and offline speech generation
-- assistants, games, prototypes, and efficient deployment research
-- applications where model size and CPU execution matter
-
-## Known Limitations
-
-- English only
-- fixed voices; this is not zero-shot voice cloning
-- unfamiliar phrasing can become flatter or more artifact-prone
-- punctuation and prosody are learned rather than guaranteed
-- numbers, abbreviations, and homographs depend on the text frontend
-- long passages should be sentence-chunked by the runtime
-- synthetic teacher speech can transfer teacher-specific biases
-- not validated for critical medical, legal, emergency, or accessibility use
-
-## Training Data and License
-
-The model was trained primarily on synthetic English speech generated by a
-larger teacher TTS system. The final release will disclose corpus size, teacher
-configuration, filtering, train/test separation, and reference-voice
-provenance.
-
-Repository code is Apache 2.0 unless otherwise noted. The final model-weight and
-synthetic-data terms are still under review, so this preview uses
-`license: other`. Upstream code, teacher-model, dataset, and voice terms are not
-automatically relicensed by this repository.
-
-## Responsible Use
-
-Do not use the included voices to impersonate a real person, deceive listeners,
-or create fraudulent content. Disclose synthetic speech where context could
-otherwise mislead.
-
-## Release Gate
-
-Before this repository becomes public:
-
-- [ ] train and select the female checkpoint
-- [ ] upload inference-only FP32 artifacts and SHA-256 hashes
-- [ ] validate FP16 and ONNX parity
-- [ ] run exact-artifact WER, UTMOS22, runtime, and failure audits
-- [ ] complete an exact-checkpoint blind listening study
-- [ ] publish representative, average, OOD, long, and failure samples
-- [ ] verify installation and inference from a clean machine
-- [ ] finalize model and dataset licensing
+---
+license: apache-2.0
+language:
+  - en
+pipeline_tag: text-to-speech
+tags:
+  - text-to-speech
+  - speech-synthesis
+  - local-tts
+  - cpu
+  - small-model
+  - vits
+  - 24khz
+---
+
+<h1 align="center">Inflect-Micro-v2</h1>
+
+<p align="center"><strong>A complete 24 kHz text-to-waveform TTS system in 9,356,513 parameters.</strong><br>One English voice. CPU-ready. No external vocoder.</p>
+
+Inflect-Micro-v2 is designed for local applications where the entire speech stack must fit below **10M parameters**. The count includes text encoding, duration modeling, latent generation, and waveform decoding. It does not hide a separately downloaded vocoder.
+
+## Listen First
+
+All samples below use unseen evaluation text, not training-audio reconstructions.
+
+| Prompt | Male voice |
+| --- | --- |
+| **Conversational** - It wasn't until later that I realized what had actually happened. | <audio controls preload="none" src="samples/male/conversational.wav"></audio> |
+| **Punctuation** - First, close the window; second, turn off the lamp; finally, lock the door. | <audio controls preload="none" src="samples/male/punctuation.wav"></audio> |
+| **Numbers** - The package weighs twelve point six kilograms and arrived on July twenty-first. | <audio controls preload="none" src="samples/male/numbers.wav"></audio> |
+| **Names Places** - Gwendolyn photographed the eucalyptus trees outside Ljubljana. | <audio controls preload="none" src="samples/male/names_places.wav"></audio> |
+| **Technical** - The system runs on three core components that all have to stay in sync. | <audio controls preload="none" src="samples/male/technical.wav"></audio> |
+
+## Model Facts
+
+| | Inflect-Micro-v2 |
+| --- | ---: |
+| Deployable inference parameters | **9,356,513** |
+| PyTorch checkpoint size | **37.53 MB** |
+| Sample rate | **24 kHz** |
+| Included voices | **1 male voice** |
+| External vocoder | **None** |
+| Inference package | **PyTorch** |
+
+## Training Data
+
+The deployable Inflect weights were trained from scratch on a single-speaker synthetic English corpus produced with Qwen3-TTS. Qwen3-TTS is a training-time teacher only and is not downloaded or used for inference. The corpus itself is not included in this repository.
+
+## Evaluation
+
+![Matched unseen-text evaluation](assets/matched-evaluation.svg)
+
+| System | UTMOS22 higher is better | Whisper-large-v3 semantic WER lower is better |
+| --- | ---: | ---: |
+| Inflect-Micro-v2 | 4.406 | 1.23% |
+| Inflect-Nano-v2 | 4.392 | 1.53% |
+| Piper Ryan Low | 4.256 | 1.23% |
+| Kitten Nano Bruno | 4.246 | 0.96% |
+
+The selected Inflect-Micro-v2 checkpoint scored **4.406 UTMOS22** and **1.23% semantic WER** on the matched 200-prompt release set. Automated metrics are diagnostics, not substitutes for listening.
+
+<details>
+<summary><strong>Evaluation protocol and raw reports</strong></summary>
+
+- All four systems synthesize the same 200 unseen English prompts.
+- Semantic WER uses Whisper-large-v3 and the disclosed English text normalizer.
+- UTMOS22 uses `tarepan/SpeechMOS` v1.2.0 with 5,000 paired bootstrap samples.
+- NISQA dimensions, signal diagnostics, categorized ASR failures, speaker consistency, and multi-seed robustness are under `evaluation/`.
+- Competitor results apply only to the named checkpoints and voices, not every configuration of those projects.
+
+</details>
+
+## Quickstart
+
+```bash
+pip install torch numpy scipy soundfile phonemizer espeakng-loader num2words Unidecode huggingface_hub
+hf download owensong/Inflect-Micro-v2 --local-dir Inflect-Micro-v2
+python Inflect-Micro-v2/inference.py --model-dir Inflect-Micro-v2 --device cpu --text "Small models should still sound alive." --output sample.wav
+```
+
+Python:
+
+```python
+import sys
+from huggingface_hub import snapshot_download
+
+model_dir = snapshot_download("owensong/Inflect-Micro-v2")
+sys.path.insert(0, model_dir)
+from inference import InflectTTS
+
+tts = InflectTTS(model_dir, device="cpu")
+tts.save("The complete model runs locally.", "sample.wav", seed=7)
+```
+
+Long text is split at punctuation-aware boundaries by the packaged runtime. `speed` ranges from `0.5` to `2.0`; `variation` ranges from `0.0` to `1.0`.
+
+## Architecture
+
+Inflect v2 is a size-specific compact VITS-derived system:
+
+- English normalization and stress-marked eSpeak phonemes with punctuation retained
+- compact Transformer text encoder and deterministic duration predictor
+- monotonic alignment, posterior encoder, and residual coupling flow during training
+- integrated adversarial waveform decoder
+- multi-resolution STFT, adversarial, feature, mel, duration, and KL supervision
+
+| Component setting | Value |
+| --- | ---: |
+| Latent channels | 192 |
+| Text hidden channels | 96 |
+| Text encoder layers / heads | 3 / 2 |
+| Feed-forward channels | 768 |
+| Posterior encoder layers | 3 |
+| Initial decoder channels | 320 |
+| Upsample rates | 8, 8, 2, 2 |
+| Training segment | 16,384 samples |
+
+The downloadable checkpoint is inference-only. Training discriminators and optimizer state are excluded from both the package and deployable parameter count.
+
+## Runtime
+
+End-to-end warm inference, including the text frontend:
+
+| Device | RTF lower is better | Audio / wall time | p95 utterance latency | Load time |
+| --- | ---: | ---: | ---: | ---: |
+| CPU | 0.6969 | 1.4x | 4.398 s | 1.29 s |
+| CUDA | 0.0168 | 59.5x | 0.086 s | 1.36 s |
+
+Runtime depends strongly on CPU, PyTorch build, thread settings, and text length. Reproduce it from the JSON reports under `evaluation/runtime/`.
+
+## Limitations
+
+- English only, with one fixed male voice; this release does not support voice cloning.
+- Unfamiliar phrasing can become flatter or less expressive.
+- Numbers and uncommon names are the weakest intelligibility categories in the release evaluation; verify critical verbalizations.
+- Homograph pronunciation depends on context available to the frontend.
+- Long passages are synthesized in chunks rather than as one unbounded sequence.
+- This model is not validated for medical, legal, emergency, or accessibility-critical use.
+- Synthetic training speech can transfer teacher cadence and biases.
+
+## Integrity and License
+
+`release_manifest.json` records the source checkpoint, deployable parameter count, byte sizes, and SHA-256 hashes. Original Inflect work is licensed under Apache-2.0; bundled third-party portions retain the notices in `THIRD_PARTY_NOTICES.md`. Do not use the included voice to impersonate a real person or deceive listeners.
+
+## Citation
+
+```bibtex
+@software{song2026inflectmicrov2,
+  author = {Owen Song},
+  title = {Inflect-Micro-v2: Complete Local Text-to-Waveform TTS Under 10M Parameters},
+  year = {2026},
+  url = {https://huggingface.co/owensong/Inflect-Micro-v2}
+}
+```

THIRD_PARTY_NOTICES.md

@@ -0,0 +1,26 @@
+# Third-Party Notices
+
+Inflect v2 includes or adapts portions of the following open-source projects.
+Their licenses apply to the corresponding portions; the rest of the Inflect v2
+package is provided under the root Apache-2.0 license.
+
+## VITS
+
+- Project: VITS (`jaywalnut310/vits`)
+- Copyright: Copyright (c) 2021 Jaehyeon Kim
+- License: MIT
+- Packaged license: `third_party/VITS_LICENSE.txt`
+
+The compact model architecture and several inference runtime modules derive from
+VITS. The text frontend also retains its original Keith Ito MIT license at
+`runtime/text/LICENSE`.
+
+## BigVGAN
+
+- Project: BigVGAN (`NVIDIA/BigVGAN`)
+- Copyright: Copyright (c) 2024 NVIDIA CORPORATION
+- License: MIT
+- Packaged license: `third_party/BIGVGAN_LICENSE.txt`
+
+The lightweight alias-free waveform activation implementation derives from
+BigVGAN's alias-free design and was adapted for this compact runtime.

config.json

@@ -0,0 +1,108 @@
+{
+  "train": {
+    "log_interval": 25,
+    "eval_interval": 2000,
+    "max_steps": 188000,
+    "seed": 20260720,
+    "epochs": 1000,
+    "learning_rate": 5e-06,
+    "resume_learning_rate": 5e-06,
+    "betas": [
+      0.8,
+      0.99
+    ],
+    "eps": 1e-09,
+    "batch_size": 32,
+    "num_workers": 8,
+    "fp16_run": true,
+    "lr_decay": 0.9999,
+    "segment_size": 16384,
+    "init_lr_ratio": 1,
+    "warmup_epochs": 0,
+    "c_mel": 45,
+    "c_kl": 1.0,
+    "c_stft": 1.5,
+    "stft_resolutions": [
+      [
+        1024,
+        256,
+        1024
+      ],
+      [
+        512,
+        128,
+        512
+      ],
+      [
+        256,
+        64,
+        256
+      ]
+    ],
+    "freeze_linguistic": true
+  },
+  "data": {
+    "training_files": "filelists/inflect_qwenfull_train_divrepair_v1.txt",
+    "validation_files": "filelists/inflect_qwenfull_dev_inflect_frontend.txt",
+    "text_cleaners": [],
+    "max_wav_value": 32768.0,
+    "sampling_rate": 24000,
+    "filter_length": 1024,
+    "hop_length": 256,
+    "win_length": 1024,
+    "n_mel_channels": 80,
+    "mel_fmin": 0.0,
+    "mel_fmax": 12000.0,
+    "add_blank": true,
+    "n_speakers": 0,
+    "cleaned_text": true
+  },
+  "model": {
+    "inter_channels": 192,
+    "hidden_channels": 96,
+    "filter_channels": 768,
+    "n_heads": 2,
+    "n_layers": 3,
+    "kernel_size": 3,
+    "p_dropout": 0.1,
+    "resblock": "1",
+    "resblock_kernel_sizes": [
+      3,
+      7,
+      11
+    ],
+    "resblock_dilation_sizes": [
+      [
+        1,
+        3,
+        5
+      ],
+      [
+        1,
+        3,
+        5
+      ],
+      [
+        1,
+        3,
+        5
+      ]
+    ],
+    "upsample_rates": [
+      8,
+      8,
+      2,
+      2
+    ],
+    "upsample_initial_channel": 320,
+    "upsample_kernel_sizes": [
+      16,
+      16,
+      4,
+      4
+    ],
+    "n_layers_q": 3,
+    "use_spectral_norm": false,
+    "use_sdp": false
+  }
+}

evaluation/asr_failures.json

@@ -0,0 +1,1961 @@
+{
+  "format": "inflect_tts_asr_failure_taxonomy_v1",
+  "source_report": "release_polish_overnight_20260720/final_large200/asr/report.json",
+  "asr": "whisper-large-v3",
+  "systems": {
+    "inflect-micro-v2-finalist": {
+      "utterances": 200,
+      "reference_words": 2608,
+      "semantic_wer": 0.012269938650306749,
+      "substitution_rate": 0.008819018404907975,
+      "deletion_rate": 0.0011503067484662577,
+      "insertion_rate": 0.0023006134969325155,
+      "utterance_failure_rate": 0.105,
+      "severe_failure_rate": 0.02,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.02
+    },
+    "inflect-nano-v2-finalist": {
+      "utterances": 200,
+      "reference_words": 2608,
+      "semantic_wer": 0.015337423312883436,
+      "substitution_rate": 0.009585889570552147,
+      "deletion_rate": 0.0011503067484662577,
+      "insertion_rate": 0.004601226993865031,
+      "utterance_failure_rate": 0.115,
+      "severe_failure_rate": 0.03,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.005,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.02
+    },
+    "kitten-nano-bruno": {
+      "utterances": 200,
+      "reference_words": 2608,
+      "semantic_wer": 0.009585889570552147,
+      "substitution_rate": 0.005368098159509203,
+      "deletion_rate": 0.0023006134969325155,
+      "insertion_rate": 0.0019171779141104294,
+      "utterance_failure_rate": 0.075,
+      "severe_failure_rate": 0.01,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.005,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.02
+    },
+    "piper-ryan-low": {
+      "utterances": 200,
+      "reference_words": 2608,
+      "semantic_wer": 0.012269938650306749,
+      "substitution_rate": 0.009585889570552147,
+      "deletion_rate": 0.0011503067484662577,
+      "insertion_rate": 0.0015337423312883436,
+      "utterance_failure_rate": 0.095,
+      "severe_failure_rate": 0.02,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.02
+    }
+  },
+  "systems_by_category": {
+    "inflect-micro-v2-finalist::conversational": {
+      "utterances": 25,
+      "reference_words": 309,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-micro-v2-finalist::descriptive": {
+      "utterances": 16,
+      "reference_words": 226,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0625
+    },
+    "inflect-micro-v2-finalist::emotional": {
+      "utterances": 24,
+      "reference_words": 329,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.041666666666666664
+    },
+    "inflect-micro-v2-finalist::homographs": {
+      "utterances": 6,
+      "reference_words": 70,
+      "semantic_wer": 0.014285714285714285,
+      "substitution_rate": 0.014285714285714285,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.16666666666666666,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-micro-v2-finalist::long": {
+      "utterances": 15,
+      "reference_words": 432,
+      "semantic_wer": 0.0023148148148148147,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0023148148148148147,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.06666666666666667,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-micro-v2-finalist::long_range": {
+      "utterances": 6,
+      "reference_words": 121,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-micro-v2-finalist::names_places": {
+      "utterances": 6,
+      "reference_words": 57,
+      "semantic_wer": 0.07017543859649122,
+      "substitution_rate": 0.07017543859649122,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.5,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-micro-v2-finalist::narrative": {
+      "utterances": 24,
+      "reference_words": 352,
+      "semantic_wer": 0.002840909090909091,
+      "substitution_rate": 0.002840909090909091,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.041666666666666664,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-micro-v2-finalist::numbers": {
+      "utterances": 6,
+      "reference_words": 64,
+      "semantic_wer": 0.15625,
+      "substitution_rate": 0.125,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.03125,
+      "utterance_failure_rate": 0.8333333333333334,
+      "severe_failure_rate": 0.3333333333333333,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-micro-v2-finalist::punctuation": {
+      "utterances": 6,
+      "reference_words": 65,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-micro-v2-finalist::question": {
+      "utterances": 15,
+      "reference_words": 165,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-micro-v2-finalist::rare_words": {
+      "utterances": 6,
+      "reference_words": 57,
+      "semantic_wer": 0.05263157894736842,
+      "substitution_rate": 0.03508771929824561,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.017543859649122806,
+      "utterance_failure_rate": 0.3333333333333333,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-micro-v2-finalist::short": {
+      "utterances": 24,
+      "reference_words": 105,
+      "semantic_wer": 0.0380952380952381,
+      "substitution_rate": 0.01904761904761905,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.01904761904761905,
+      "utterance_failure_rate": 0.16666666666666666,
+      "severe_failure_rate": 0.041666666666666664,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.08333333333333333
+    },
+    "inflect-micro-v2-finalist::technical": {
+      "utterances": 21,
+      "reference_words": 256,
+      "semantic_wer": 0.03125,
+      "substitution_rate": 0.01953125,
+      "deletion_rate": 0.0078125,
+      "insertion_rate": 0.00390625,
+      "utterance_failure_rate": 0.19047619047619047,
+      "severe_failure_rate": 0.047619047619047616,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-nano-v2-finalist::conversational": {
+      "utterances": 25,
+      "reference_words": 309,
+      "semantic_wer": 0.003236245954692557,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.003236245954692557,
+      "utterance_failure_rate": 0.04,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-nano-v2-finalist::descriptive": {
+      "utterances": 16,
+      "reference_words": 226,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0625
+    },
+    "inflect-nano-v2-finalist::emotional": {
+      "utterances": 24,
+      "reference_words": 329,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.041666666666666664
+    },
+    "inflect-nano-v2-finalist::homographs": {
+      "utterances": 6,
+      "reference_words": 70,
+      "semantic_wer": 0.05714285714285714,
+      "substitution_rate": 0.05714285714285714,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.3333333333333333,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-nano-v2-finalist::long": {
+      "utterances": 15,
+      "reference_words": 432,
+      "semantic_wer": 0.004629629629629629,
+      "substitution_rate": 0.0023148148148148147,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0023148148148148147,
+      "utterance_failure_rate": 0.06666666666666667,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-nano-v2-finalist::long_range": {
+      "utterances": 6,
+      "reference_words": 121,
+      "semantic_wer": 0.008264462809917356,
+      "substitution_rate": 0.008264462809917356,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.16666666666666666,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-nano-v2-finalist::names_places": {
+      "utterances": 6,
+      "reference_words": 57,
+      "semantic_wer": 0.08771929824561403,
+      "substitution_rate": 0.08771929824561403,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.5,
+      "severe_failure_rate": 0.16666666666666666,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-nano-v2-finalist::narrative": {
+      "utterances": 24,
+      "reference_words": 352,
+      "semantic_wer": 0.011363636363636364,
+      "substitution_rate": 0.005681818181818182,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.005681818181818182,
+      "utterance_failure_rate": 0.125,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-nano-v2-finalist::numbers": {
+      "utterances": 6,
+      "reference_words": 64,
+      "semantic_wer": 0.125,
+      "substitution_rate": 0.078125,
+      "deletion_rate": 0.015625,
+      "insertion_rate": 0.03125,
+      "utterance_failure_rate": 0.5,
+      "severe_failure_rate": 0.3333333333333333,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.16666666666666666,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-nano-v2-finalist::punctuation": {
+      "utterances": 6,
+      "reference_words": 65,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-nano-v2-finalist::question": {
+      "utterances": 15,
+      "reference_words": 165,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-nano-v2-finalist::rare_words": {
+      "utterances": 6,
+      "reference_words": 57,
+      "semantic_wer": 0.03508771929824561,
+      "substitution_rate": 0.017543859649122806,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.017543859649122806,
+      "utterance_failure_rate": 0.16666666666666666,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "inflect-nano-v2-finalist::short": {
+      "utterances": 24,
+      "reference_words": 105,
+      "semantic_wer": 0.047619047619047616,
+      "substitution_rate": 0.009523809523809525,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0380952380952381,
+      "utterance_failure_rate": 0.16666666666666666,
+      "severe_failure_rate": 0.08333333333333333,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.08333333333333333
+    },
+    "inflect-nano-v2-finalist::technical": {
+      "utterances": 21,
+      "reference_words": 256,
+      "semantic_wer": 0.03125,
+      "substitution_rate": 0.01953125,
+      "deletion_rate": 0.0078125,
+      "insertion_rate": 0.00390625,
+      "utterance_failure_rate": 0.19047619047619047,
+      "severe_failure_rate": 0.047619047619047616,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "kitten-nano-bruno::conversational": {
+      "utterances": 25,
+      "reference_words": 309,
+      "semantic_wer": 0.003236245954692557,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.003236245954692557,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.04,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "kitten-nano-bruno::descriptive": {
+      "utterances": 16,
+      "reference_words": 226,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0625
+    },
+    "kitten-nano-bruno::emotional": {
+      "utterances": 24,
+      "reference_words": 329,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.041666666666666664
+    },
+    "kitten-nano-bruno::homographs": {
+      "utterances": 6,
+      "reference_words": 70,
+      "semantic_wer": 0.014285714285714285,
+      "substitution_rate": 0.014285714285714285,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.16666666666666666,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "kitten-nano-bruno::long": {
+      "utterances": 15,
+      "reference_words": 432,
+      "semantic_wer": 0.0023148148148148147,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0023148148148148147,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.06666666666666667,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "kitten-nano-bruno::long_range": {
+      "utterances": 6,
+      "reference_words": 121,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "kitten-nano-bruno::names_places": {
+      "utterances": 6,
+      "reference_words": 57,
+      "semantic_wer": 0.05263157894736842,
+      "substitution_rate": 0.05263157894736842,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.16666666666666666,
+      "severe_failure_rate": 0.16666666666666666,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "kitten-nano-bruno::narrative": {
+      "utterances": 24,
+      "reference_words": 352,
+      "semantic_wer": 0.011363636363636364,
+      "substitution_rate": 0.002840909090909091,
+      "deletion_rate": 0.005681818181818182,
+      "insertion_rate": 0.002840909090909091,
+      "utterance_failure_rate": 0.125,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.041666666666666664,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "kitten-nano-bruno::numbers": {
+      "utterances": 6,
+      "reference_words": 64,
+      "semantic_wer": 0.09375,
+      "substitution_rate": 0.0625,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.03125,
+      "utterance_failure_rate": 0.5,
+      "severe_failure_rate": 0.16666666666666666,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "kitten-nano-bruno::punctuation": {
+      "utterances": 6,
+      "reference_words": 65,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "kitten-nano-bruno::question": {
+      "utterances": 15,
+      "reference_words": 165,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "kitten-nano-bruno::rare_words": {
+      "utterances": 6,
+      "reference_words": 57,
+      "semantic_wer": 0.03508771929824561,
+      "substitution_rate": 0.017543859649122806,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.017543859649122806,
+      "utterance_failure_rate": 0.16666666666666666,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "kitten-nano-bruno::short": {
+      "utterances": 24,
+      "reference_words": 105,
+      "semantic_wer": 0.01904761904761905,
+      "substitution_rate": 0.01904761904761905,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.08333333333333333,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.08333333333333333
+    },
+    "kitten-nano-bruno::technical": {
+      "utterances": 21,
+      "reference_words": 256,
+      "semantic_wer": 0.01953125,
+      "substitution_rate": 0.0078125,
+      "deletion_rate": 0.0078125,
+      "insertion_rate": 0.00390625,
+      "utterance_failure_rate": 0.09523809523809523,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "piper-ryan-low::conversational": {
+      "utterances": 25,
+      "reference_words": 309,
+      "semantic_wer": 0.006472491909385114,
+      "substitution_rate": 0.006472491909385114,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.04,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "piper-ryan-low::descriptive": {
+      "utterances": 16,
+      "reference_words": 226,
+      "semantic_wer": 0.008849557522123894,
+      "substitution_rate": 0.008849557522123894,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.125,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0625
+    },
+    "piper-ryan-low::emotional": {
+      "utterances": 24,
+      "reference_words": 329,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.041666666666666664
+    },
+    "piper-ryan-low::homographs": {
+      "utterances": 6,
+      "reference_words": 70,
+      "semantic_wer": 0.014285714285714285,
+      "substitution_rate": 0.014285714285714285,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.16666666666666666,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "piper-ryan-low::long": {
+      "utterances": 15,
+      "reference_words": 432,
+      "semantic_wer": 0.004629629629629629,
+      "substitution_rate": 0.0023148148148148147,
+      "deletion_rate": 0.0023148148148148147,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.13333333333333333,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "piper-ryan-low::long_range": {
+      "utterances": 6,
+      "reference_words": 121,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "piper-ryan-low::names_places": {
+      "utterances": 6,
+      "reference_words": 57,
+      "semantic_wer": 0.07017543859649122,
+      "substitution_rate": 0.07017543859649122,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.3333333333333333,
+      "severe_failure_rate": 0.16666666666666666,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "piper-ryan-low::narrative": {
+      "utterances": 24,
+      "reference_words": 352,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "piper-ryan-low::numbers": {
+      "utterances": 6,
+      "reference_words": 64,
+      "semantic_wer": 0.078125,
+      "substitution_rate": 0.046875,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.03125,
+      "utterance_failure_rate": 0.3333333333333333,
+      "severe_failure_rate": 0.16666666666666666,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "piper-ryan-low::punctuation": {
+      "utterances": 6,
+      "reference_words": 65,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "piper-ryan-low::question": {
+      "utterances": 15,
+      "reference_words": 165,
+      "semantic_wer": 0.0,
+      "substitution_rate": 0.0,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.0,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "piper-ryan-low::rare_words": {
+      "utterances": 6,
+      "reference_words": 57,
+      "semantic_wer": 0.08771929824561403,
+      "substitution_rate": 0.07017543859649122,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.017543859649122806,
+      "utterance_failure_rate": 0.5,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    },
+    "piper-ryan-low::short": {
+      "utterances": 24,
+      "reference_words": 105,
+      "semantic_wer": 0.0380952380952381,
+      "substitution_rate": 0.0380952380952381,
+      "deletion_rate": 0.0,
+      "insertion_rate": 0.0,
+      "utterance_failure_rate": 0.08333333333333333,
+      "severe_failure_rate": 0.08333333333333333,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.08333333333333333
+    },
+    "piper-ryan-low::technical": {
+      "utterances": 21,
+      "reference_words": 256,
+      "semantic_wer": 0.02734375,
+      "substitution_rate": 0.015625,
+      "deletion_rate": 0.0078125,
+      "insertion_rate": 0.00390625,
+      "utterance_failure_rate": 0.19047619047619047,
+      "severe_failure_rate": 0.0,
+      "empty_hypothesis_rate": 0.0,
+      "leading_omission_rate": 0.0,
+      "ending_omission_rate": 0.0,
+      "insertion_burst_rate": 0.0,
+      "adjacent_phrase_repetition_rate": 0.0
+    }
+  },
+  "worst_utterances": [
+    {
+      "model": "inflect-micro-v2-finalist",
+      "id": "utmos200_176",
+      "category": "short",
+      "reference": "Noted.",
+      "hypothesis": "Noid.",
+      "reference_words": 1,
+      "hypothesis_words": 1,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 1.0,
+      "utterance_failure": true,
+      "severe_failure": true,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-micro-v2-finalist/utmos200_176.wav"
+    },
+    {
+      "model": "piper-ryan-low",
+      "id": "utmos200_041",
+      "category": "short",
+      "reference": "I mean. Yeah. Kind of.",
+      "hypothesis": "I mean, no, I know.",
+      "reference_words": 5,
+      "hypothesis_words": 5,
+      "substitutions": 3,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.6,
+      "utterance_failure": true,
+      "severe_failure": true,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/piper-ryan-low/utmos200_041.wav"
+    },
+    {
+      "model": "inflect-nano-v2-finalist",
+      "id": "utmos200_041",
+      "category": "short",
+      "reference": "I mean. Yeah. Kind of.",
+      "hypothesis": "I mean that, yeah, that kind of,",
+      "reference_words": 5,
+      "hypothesis_words": 7,
+      "substitutions": 0,
+      "deletions": 0,
+      "insertions": 2,
+      "semantic_wer": 0.4,
+      "utterance_failure": true,
+      "severe_failure": true,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-nano-v2-finalist/utmos200_041.wav"
+    },
+    {
+      "model": "inflect-micro-v2-finalist",
+      "id": "utmos200_065",
+      "category": "numbers",
+      "reference": "Call extension zero four nine after eight thirty in the morning.",
+      "hypothesis": "Call extension 049 after 8.30 in the morning.",
+      "reference_words": 8,
+      "hypothesis_words": 9,
+      "substitutions": 2,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.375,
+      "utterance_failure": true,
+      "severe_failure": true,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-micro-v2-finalist/utmos200_065.wav"
+    },
+    {
+      "model": "inflect-nano-v2-finalist",
+      "id": "utmos200_065",
+      "category": "numbers",
+      "reference": "Call extension zero four nine after eight thirty in the morning.",
+      "hypothesis": "Call extension 049 after 8.30 in the morning.",
+      "reference_words": 8,
+      "hypothesis_words": 9,
+      "substitutions": 2,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.375,
+      "utterance_failure": true,
+      "severe_failure": true,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-nano-v2-finalist/utmos200_065.wav"
+    },
+    {
+      "model": "piper-ryan-low",
+      "id": "utmos200_065",
+      "category": "numbers",
+      "reference": "Call extension zero four nine after eight thirty in the morning.",
+      "hypothesis": "Call extension 049 after 8.30 in the morning.",
+      "reference_words": 8,
+      "hypothesis_words": 9,
+      "substitutions": 2,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.375,
+      "utterance_failure": true,
+      "severe_failure": true,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/piper-ryan-low/utmos200_065.wav"
+    },
+    {
+      "model": "kitten-nano-bruno",
+      "id": "utmos200_065",
+      "category": "numbers",
+      "reference": "Call extension zero four nine after eight thirty in the morning.",
+      "hypothesis": "Call extension 049 after 8.30 in the morning.",
+      "reference_words": 8,
+      "hypothesis_words": 9,
+      "substitutions": 2,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.375,
+      "utterance_failure": true,
+      "severe_failure": true,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/kitten-nano-bruno/utmos200_065.wav"
+    },
+    {
+      "model": "inflect-nano-v2-finalist",
+      "id": "utmos200_035",
+      "category": "names_places",
+      "reference": "Aurelia and Joaquin crossed the Schuylkill River before noon.",
+      "hypothesis": "Uralaya and Jokwing crossed the Sialkal River before noon.",
+      "reference_words": 9,
+      "hypothesis_words": 9,
+      "substitutions": 3,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.3333333333333333,
+      "utterance_failure": true,
+      "severe_failure": true,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-nano-v2-finalist/utmos200_035.wav"
+    },
+    {
+      "model": "inflect-nano-v2-finalist",
+      "id": "utmos200_191",
+      "category": "short",
+      "reference": "Right. Moving on.",
+      "hypothesis": "Right at moving on.",
+      "reference_words": 3,
+      "hypothesis_words": 4,
+      "substitutions": 0,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.3333333333333333,
+      "utterance_failure": true,
+      "severe_failure": true,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-nano-v2-finalist/utmos200_191.wav"
+    },
+    {
+      "model": "piper-ryan-low",
+      "id": "utmos200_049",
+      "category": "names_places",
+      "reference": "Xiomara met Siobhan near the old library in Worcestershire.",
+      "hypothesis": "See Amaru Mechivan near the Old Library in Worcestershire.",
+      "reference_words": 9,
+      "hypothesis_words": 9,
+      "substitutions": 3,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.3333333333333333,
+      "utterance_failure": true,
+      "severe_failure": true,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/piper-ryan-low/utmos200_049.wav"
+    },
+    {
+      "model": "piper-ryan-low",
+      "id": "utmos200_091",
+      "category": "short",
+      "reference": "That's fair.",
+      "hypothesis": "That was fair.",
+      "reference_words": 3,
+      "hypothesis_words": 3,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.3333333333333333,
+      "utterance_failure": true,
+      "severe_failure": true,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/piper-ryan-low/utmos200_091.wav"
+    },
+    {
+      "model": "kitten-nano-bruno",
+      "id": "utmos200_035",
+      "category": "names_places",
+      "reference": "Aurelia and Joaquin crossed the Schuylkill River before noon.",
+      "hypothesis": "Auralaya and Jokwan crossed the Shilkal river before noon.",
+      "reference_words": 9,
+      "hypothesis_words": 9,
+      "substitutions": 3,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.3333333333333333,
+      "utterance_failure": true,
+      "severe_failure": true,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/kitten-nano-bruno/utmos200_035.wav"
+    },
+    {
+      "model": "inflect-micro-v2-finalist",
+      "id": "utmos200_051",
+      "category": "numbers",
+      "reference": "At exactly seven forty-five, Morgan placed eighteen dollars and fifty cents beside the notebook.",
+      "hypothesis": "At exactly $7.45, Morton placed $18.50 beside the notebook.",
+      "reference_words": 10,
+      "hypothesis_words": 11,
+      "substitutions": 2,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.3,
+      "utterance_failure": true,
+      "severe_failure": true,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-micro-v2-finalist/utmos200_051.wav"
+    },
+    {
+      "model": "inflect-nano-v2-finalist",
+      "id": "utmos200_051",
+      "category": "numbers",
+      "reference": "At exactly seven forty-five, Morgan placed eighteen dollars and fifty cents beside the notebook.",
+      "hypothesis": "Exactly $7.45, Morgan placed $18.50 beside the notebook.",
+      "reference_words": 10,
+      "hypothesis_words": 10,
+      "substitutions": 1,
+      "deletions": 1,
+      "insertions": 1,
+      "semantic_wer": 0.3,
+      "utterance_failure": true,
+      "severe_failure": true,
+      "empty_hypothesis": false,
+      "leading_omission": true,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-nano-v2-finalist/utmos200_051.wav"
+    },
+    {
+      "model": "inflect-micro-v2-finalist",
+      "id": "utmos200_028",
+      "category": "technical",
+      "reference": "Hash the UTF eight payload with S H A two fifty-six before opening the checkpoint.",
+      "hypothesis": "Cache the UTF-8 payload with SH-256 before opening the checkpoint.",
+      "reference_words": 14,
+      "hypothesis_words": 12,
+      "substitutions": 2,
+      "deletions": 2,
+      "insertions": 0,
+      "semantic_wer": 0.2857142857142857,
+      "utterance_failure": true,
+      "severe_failure": true,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-micro-v2-finalist/utmos200_028.wav"
+    },
+    {
+      "model": "inflect-nano-v2-finalist",
+      "id": "utmos200_028",
+      "category": "technical",
+      "reference": "Hash the UTF eight payload with S H A two fifty-six before opening the checkpoint.",
+      "hypothesis": "Ash the UTF-8 payload with SHA-256 before opening the checkpoint.",
+      "reference_words": 14,
+      "hypothesis_words": 12,
+      "substitutions": 2,
+      "deletions": 2,
+      "insertions": 0,
+      "semantic_wer": 0.2857142857142857,
+      "utterance_failure": true,
+      "severe_failure": true,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-nano-v2-finalist/utmos200_028.wav"
+    },
+    {
+      "model": "inflect-micro-v2-finalist",
+      "id": "utmos200_139",
+      "category": "short",
+      "reference": "No way. Absolutely not.",
+      "hypothesis": "No wait, absolutely not.",
+      "reference_words": 4,
+      "hypothesis_words": 4,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.25,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-micro-v2-finalist/utmos200_139.wav"
+    },
+    {
+      "model": "inflect-micro-v2-finalist",
+      "id": "utmos200_181",
+      "category": "short",
+      "reference": "Yeah, no. Hard pass.",
+      "hypothesis": "Yeah, no but hard pass.",
+      "reference_words": 4,
+      "hypothesis_words": 5,
+      "substitutions": 0,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.25,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-micro-v2-finalist/utmos200_181.wav"
+    },
+    {
+      "model": "inflect-nano-v2-finalist",
+      "id": "utmos200_139",
+      "category": "short",
+      "reference": "No way. Absolutely not.",
+      "hypothesis": "No way to absolutely not.",
+      "reference_words": 4,
+      "hypothesis_words": 5,
+      "substitutions": 0,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.25,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-nano-v2-finalist/utmos200_139.wav"
+    },
+    {
+      "model": "inflect-nano-v2-finalist",
+      "id": "utmos200_181",
+      "category": "short",
+      "reference": "Yeah, no. Hard pass.",
+      "hypothesis": "Yeah, no.Card pass.",
+      "reference_words": 4,
+      "hypothesis_words": 4,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.25,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-nano-v2-finalist/utmos200_181.wav"
+    },
+    {
+      "model": "kitten-nano-bruno",
+      "id": "utmos200_181",
+      "category": "short",
+      "reference": "Yeah, no. Hard pass.",
+      "hypothesis": "Yeth no. Hard pass.",
+      "reference_words": 4,
+      "hypothesis_words": 4,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.25,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/kitten-nano-bruno/utmos200_181.wav"
+    },
+    {
+      "model": "inflect-micro-v2-finalist",
+      "id": "utmos200_035",
+      "category": "names_places",
+      "reference": "Aurelia and Joaquin crossed the Schuylkill River before noon.",
+      "hypothesis": "Aurelia and Joachim crossed the Schalke river before noon.",
+      "reference_words": 9,
+      "hypothesis_words": 9,
+      "substitutions": 2,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.2222222222222222,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-micro-v2-finalist/utmos200_035.wav"
+    },
+    {
+      "model": "piper-ryan-low",
+      "id": "utmos200_028",
+      "category": "technical",
+      "reference": "Hash the UTF eight payload with S H A two fifty-six before opening the checkpoint.",
+      "hypothesis": "Hash the UTF-8 payload with SHA-256 before opening the checkpoint.",
+      "reference_words": 14,
+      "hypothesis_words": 12,
+      "substitutions": 1,
+      "deletions": 2,
+      "insertions": 0,
+      "semantic_wer": 0.21428571428571427,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/piper-ryan-low/utmos200_028.wav"
+    },
+    {
+      "model": "kitten-nano-bruno",
+      "id": "utmos200_028",
+      "category": "technical",
+      "reference": "Hash the UTF eight payload with S H A two fifty-six before opening the checkpoint.",
+      "hypothesis": "Hash the UTF-8 payload with SHA-256 before opening the checkpoint.",
+      "reference_words": 14,
+      "hypothesis_words": 12,
+      "substitutions": 1,
+      "deletions": 2,
+      "insertions": 0,
+      "semantic_wer": 0.21428571428571427,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/kitten-nano-bruno/utmos200_028.wav"
+    },
+    {
+      "model": "inflect-micro-v2-finalist",
+      "id": "utmos200_041",
+      "category": "short",
+      "reference": "I mean. Yeah. Kind of.",
+      "hypothesis": "I mean, yeah, I kind of.",
+      "reference_words": 5,
+      "hypothesis_words": 6,
+      "substitutions": 0,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.2,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-micro-v2-finalist/utmos200_041.wav"
+    },
+    {
+      "model": "inflect-micro-v2-finalist",
+      "id": "utmos200_084",
+      "category": "technical",
+      "reference": "The performance improvement is real, but it comes with tradeoffs.",
+      "hypothesis": "The performance improvement is real, but it comes with trade-offs.",
+      "reference_words": 10,
+      "hypothesis_words": 11,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.2,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-micro-v2-finalist/utmos200_084.wav"
+    },
+    {
+      "model": "inflect-nano-v2-finalist",
+      "id": "utmos200_074",
+      "category": "homographs",
+      "reference": "The dove dove beneath the branch while the wind began to wind the loose cable.",
+      "hypothesis": "The dub-dub beneath the branch wall the wind began to wind the loose cable.",
+      "reference_words": 15,
+      "hypothesis_words": 15,
+      "substitutions": 3,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.2,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-nano-v2-finalist/utmos200_074.wav"
+    },
+    {
+      "model": "inflect-nano-v2-finalist",
+      "id": "utmos200_084",
+      "category": "technical",
+      "reference": "The performance improvement is real, but it comes with tradeoffs.",
+      "hypothesis": "The performance improvement is real, but it comes with trade-offs.",
+      "reference_words": 10,
+      "hypothesis_words": 11,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.2,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-nano-v2-finalist/utmos200_084.wav"
+    },
+    {
+      "model": "piper-ryan-low",
+      "id": "utmos200_012",
+      "category": "rare_words",
+      "reference": "The archaeologist catalogued a Byzantine reliquary and several anachronistic fragments.",
+      "hypothesis": "the archaeologists cataloged a Byzantine reliquary and several anachronistic fragments.",
+      "reference_words": 10,
+      "hypothesis_words": 10,
+      "substitutions": 2,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.2,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/piper-ryan-low/utmos200_012.wav"
+    },
+    {
+      "model": "piper-ryan-low",
+      "id": "utmos200_051",
+      "category": "numbers",
+      "reference": "At exactly seven forty-five, Morgan placed eighteen dollars and fifty cents beside the notebook.",
+      "hypothesis": "At exactly 7.45, Morgan placed $18.50 beside the notebook.",
+      "reference_words": 10,
+      "hypothesis_words": 11,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.2,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/piper-ryan-low/utmos200_051.wav"
+    },
+    {
+      "model": "piper-ryan-low",
+      "id": "utmos200_084",
+      "category": "technical",
+      "reference": "The performance improvement is real, but it comes with tradeoffs.",
+      "hypothesis": "The performance improvement is real, but it comes with trade-offs.",
+      "reference_words": 10,
+      "hypothesis_words": 11,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.2,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/piper-ryan-low/utmos200_084.wav"
+    },
+    {
+      "model": "piper-ryan-low",
+      "id": "utmos200_188",
+      "category": "conversational",
+      "reference": "You got engaged? When? How? Tell me everything right now.",
+      "hypothesis": "You got engaged. What? Oh. Tell me everything right now.",
+      "reference_words": 10,
+      "hypothesis_words": 10,
+      "substitutions": 2,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.2,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/piper-ryan-low/utmos200_188.wav"
+    },
+    {
+      "model": "kitten-nano-bruno",
+      "id": "utmos200_041",
+      "category": "short",
+      "reference": "I mean. Yeah. Kind of.",
+      "hypothesis": "I mean. Yes. Kind of.",
+      "reference_words": 5,
+      "hypothesis_words": 5,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.2,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/kitten-nano-bruno/utmos200_041.wav"
+    },
+    {
+      "model": "kitten-nano-bruno",
+      "id": "utmos200_051",
+      "category": "numbers",
+      "reference": "At exactly seven forty-five, Morgan placed eighteen dollars and fifty cents beside the notebook.",
+      "hypothesis": "At exactly 7.45, Morgan placed $18.50 beside the notebook.",
+      "reference_words": 10,
+      "hypothesis_words": 11,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.2,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/kitten-nano-bruno/utmos200_051.wav"
+    },
+    {
+      "model": "kitten-nano-bruno",
+      "id": "utmos200_084",
+      "category": "technical",
+      "reference": "The performance improvement is real, but it comes with tradeoffs.",
+      "hypothesis": "The performance improvement is real, but it comes with trade-offs.",
+      "reference_words": 10,
+      "hypothesis_words": 11,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.2,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/kitten-nano-bruno/utmos200_084.wav"
+    },
+    {
+      "model": "inflect-micro-v2-finalist",
+      "id": "utmos200_009",
+      "category": "numbers",
+      "reference": "The package weighs twelve point six kilograms and arrived on July twenty-first.",
+      "hypothesis": "The package weighs 12.6kg and arrived on July 21.",
+      "reference_words": 11,
+      "hypothesis_words": 11,
+      "substitutions": 2,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.18181818181818182,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-micro-v2-finalist/utmos200_009.wav"
+    },
+    {
+      "model": "inflect-micro-v2-finalist",
+      "id": "utmos200_082",
+      "category": "rare_words",
+      "reference": "The physician recommended a noninvasive procedure for the recurring neurological symptoms.",
+      "hypothesis": "The physician recommended a non-invasive procedure for the recurring neurological symptoms.",
+      "reference_words": 11,
+      "hypothesis_words": 12,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.18181818181818182,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-micro-v2-finalist/utmos200_082.wav"
+    },
+    {
+      "model": "inflect-nano-v2-finalist",
+      "id": "utmos200_009",
+      "category": "numbers",
+      "reference": "The package weighs twelve point six kilograms and arrived on July twenty-first.",
+      "hypothesis": "The package weighs 12.6kg and arrived on July 21.",
+      "reference_words": 11,
+      "hypothesis_words": 11,
+      "substitutions": 2,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.18181818181818182,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-nano-v2-finalist/utmos200_009.wav"
+    },
+    {
+      "model": "inflect-nano-v2-finalist",
+      "id": "utmos200_082",
+      "category": "rare_words",
+      "reference": "The physician recommended a noninvasive procedure for the recurring neurological symptoms.",
+      "hypothesis": "The physician recommended a non-invasive procedure for the recurring neurological symptoms.",
+      "reference_words": 11,
+      "hypothesis_words": 12,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.18181818181818182,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-nano-v2-finalist/utmos200_082.wav"
+    },
+    {
+      "model": "piper-ryan-low",
+      "id": "utmos200_082",
+      "category": "rare_words",
+      "reference": "The physician recommended a noninvasive procedure for the recurring neurological symptoms.",
+      "hypothesis": "the physician recommended a non-invasive procedure for the recurring neurological symptoms.",
+      "reference_words": 11,
+      "hypothesis_words": 12,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.18181818181818182,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/piper-ryan-low/utmos200_082.wav"
+    },
+    {
+      "model": "kitten-nano-bruno",
+      "id": "utmos200_082",
+      "category": "rare_words",
+      "reference": "The physician recommended a noninvasive procedure for the recurring neurological symptoms.",
+      "hypothesis": "The physician recommended a non-invasive procedure for the recurring neurological symptoms.",
+      "reference_words": 11,
+      "hypothesis_words": 12,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.18181818181818182,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/kitten-nano-bruno/utmos200_082.wav"
+    },
+    {
+      "model": "inflect-nano-v2-finalist",
+      "id": "utmos200_180",
+      "category": "narrative",
+      "reference": "She knew. She had always known. She just waited for the right moment.",
+      "hypothesis": "Ji knew but she had always known, she just waited for the right moment.",
+      "reference_words": 13,
+      "hypothesis_words": 14,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.15384615384615385,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-nano-v2-finalist/utmos200_180.wav"
+    },
+    {
+      "model": "kitten-nano-bruno",
+      "id": "utmos200_175",
+      "category": "narrative",
+      "reference": "The note was short, which was somehow more frightening than a long one would have been.",
+      "hypothesis": "The note was short, which was somehow more fright, ing than a long one would have been.",
+      "reference_words": 16,
+      "hypothesis_words": 17,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.125,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/kitten-nano-bruno/utmos200_175.wav"
+    },
+    {
+      "model": "inflect-micro-v2-finalist",
+      "id": "utmos200_049",
+      "category": "names_places",
+      "reference": "Xiomara met Siobhan near the old library in Worcestershire.",
+      "hypothesis": "Zamara met Siobhan near the old library in Worcestershire.",
+      "reference_words": 9,
+      "hypothesis_words": 9,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.1111111111111111,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-micro-v2-finalist/utmos200_049.wav"
+    },
+    {
+      "model": "inflect-micro-v2-finalist",
+      "id": "utmos200_077",
+      "category": "names_places",
+      "reference": "Saskatchewan looked unusually bright beneath the pale February sunrise.",
+      "hypothesis": "Saskatchewan looked unusually bright beneath that pale February sunrise.",
+      "reference_words": 9,
+      "hypothesis_words": 9,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.1111111111111111,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-micro-v2-finalist/utmos200_077.wav"
+    },
+    {
+      "model": "inflect-nano-v2-finalist",
+      "id": "utmos200_022",
+      "category": "narrative",
+      "reference": "The plan was simple. The execution was anything but.",
+      "hypothesis": "The plan was simple. The execution was anything but perfect.",
+      "reference_words": 9,
+      "hypothesis_words": 10,
+      "substitutions": 0,
+      "deletions": 0,
+      "insertions": 1,
+      "semantic_wer": 0.1111111111111111,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-nano-v2-finalist/utmos200_022.wav"
+    },
+    {
+      "model": "inflect-nano-v2-finalist",
+      "id": "utmos200_049",
+      "category": "names_places",
+      "reference": "Xiomara met Siobhan near the old library in Worcestershire.",
+      "hypothesis": "Damara met Siobhan near the old library in Worcestershire.",
+      "reference_words": 9,
+      "hypothesis_words": 9,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.1111111111111111,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-nano-v2-finalist/utmos200_049.wav"
+    },
+    {
+      "model": "piper-ryan-low",
+      "id": "utmos200_035",
+      "category": "names_places",
+      "reference": "Aurelia and Joaquin crossed the Schuylkill River before noon.",
+      "hypothesis": "Aurelia and Joaquin crossed the Chiocul River before noon.",
+      "reference_words": 9,
+      "hypothesis_words": 9,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.1111111111111111,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "utmos22_large200_v1/audio/piper-ryan-low/utmos200_035.wav"
+    },
+    {
+      "model": "inflect-micro-v2-finalist",
+      "id": "utmos200_018",
+      "category": "homographs",
+      "reference": "They refuse to place the refuse beside the produce aisle.",
+      "hypothesis": "They refused to place the refuse beside the produce aisle.",
+      "reference_words": 10,
+      "hypothesis_words": 10,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.1,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-micro-v2-finalist/utmos200_018.wav"
+    },
+    {
+      "model": "inflect-micro-v2-finalist",
+      "id": "utmos200_023",
+      "category": "numbers",
+      "reference": "Her score improved from eighty-two percent to ninety-seven percent in two weeks.",
+      "hypothesis": "GERS score improved from 82% to 97% in two weeks.",
+      "reference_words": 10,
+      "hypothesis_words": 10,
+      "substitutions": 1,
+      "deletions": 0,
+      "insertions": 0,
+      "semantic_wer": 0.1,
+      "utterance_failure": true,
+      "severe_failure": false,
+      "empty_hypothesis": false,
+      "leading_omission": false,
+      "ending_omission": false,
+      "insertion_burst": false,
+      "adjacent_phrase_repetition": false,
+      "audio": "release_polish_overnight_20260720/final_large200/audio/inflect-micro-v2-finalist/utmos200_023.wav"
+    }
+  ]
+}

evaluation/long_form.json

@@ -0,0 +1,9 @@
+{
+  "text_characters": 738,
+  "durations_seconds": {
+    "inflect-micro-v2-finalist": 44.37266666666667,
+    "inflect-nano-v2-finalist": 43.967333333333336
+  },
+  "asr_report": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/long_form/asr/report.json",
+  "signal_report": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/long_form/signal/report.json"
+}

evaluation/long_form_asr.json

@@ -0,0 +1,149 @@
+{
+  "format": "inflect_tts_multi_asr_wer_v1",
+  "manifest": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/long_form/combined_manifest.jsonl",
+  "asr_models": {
+    "whisper-large-v3": "openai/whisper-large-v3"
+  },
+  "normalization": {
+    "raw": "lowercase_punctuation_stripped",
+    "semantic": "transformers_whisper_EnglishTextNormalizer_then_lexical"
+  },
+  "systems": {
+    "inflect-micro-v2-finalist": {
+      "asr": {
+        "whisper-large-v3": {
+          "raw": {
+            "utterances": 1,
+            "errors": 16,
+            "words": 110,
+            "wer": 0.14545454545454545
+          },
+          "raw_95ci": [
+            0.14545454545454545,
+            0.14545454545454545
+          ],
+          "semantic": {
+            "utterances": 1,
+            "errors": 4,
+            "words": 103,
+            "wer": 0.038834951456310676
+          },
+          "semantic_95ci": [
+            0.038834951456310676,
+            0.038834951456310676
+          ],
+          "categories": {
+            "long_form": {
+              "raw": {
+                "utterances": 1,
+                "errors": 16,
+                "words": 110,
+                "wer": 0.14545454545454545
+              },
+              "semantic": {
+                "utterances": 1,
+                "errors": 4,
+                "words": 103,
+                "wer": 0.038834951456310676
+              }
+            }
+          }
+        }
+      },
+      "cross_asr_mean_semantic_wer": 0.038834951456310676,
+      "cross_asr_worst_semantic_wer": 0.038834951456310676
+    },
+    "inflect-nano-v2-finalist": {
+      "asr": {
+        "whisper-large-v3": {
+          "raw": {
+            "utterances": 1,
+            "errors": 18,
+            "words": 110,
+            "wer": 0.16363636363636364
+          },
+          "raw_95ci": [
+            0.16363636363636364,
+            0.16363636363636364
+          ],
+          "semantic": {
+            "utterances": 1,
+            "errors": 6,
+            "words": 103,
+            "wer": 0.05825242718446602
+          },
+          "semantic_95ci": [
+            0.05825242718446602,
+            0.05825242718446602
+          ],
+          "categories": {
+            "long_form": {
+              "raw": {
+                "utterances": 1,
+                "errors": 18,
+                "words": 110,
+                "wer": 0.16363636363636364
+              },
+              "semantic": {
+                "utterances": 1,
+                "errors": 6,
+                "words": 103,
+                "wer": 0.05825242718446602
+              }
+            }
+          }
+        }
+      },
+      "cross_asr_mean_semantic_wer": 0.05825242718446602,
+      "cross_asr_worst_semantic_wer": 0.05825242718446602
+    }
+  },
+  "rows": [
+    {
+      "id": "release_long_form",
+      "category": "long_form",
+      "text": "At seven forty-five on Tuesday morning, Morgan checked the maintenance report before the first delivery arrived. The document listed three unresolved items: a fluorescent lamp near the loading bay, a pressure sensor marked B seventeen, and a package weighing twelve point six kilograms. Although the weather changed without warning, the technicians finished their inspection, called Saskatchewan headquarters, and sent the revised PDF before noon. Could the replacement parts arrive by Friday? The coordinator replied, yes, but only if invoice eight eight nine five three was approved before five P.M. After one final review, everyone agreed that clear punctuation, stable pacing, and accurate numbers mattered more than speaking quickly.",
+      "model": "inflect-micro-v2-finalist",
+      "audio": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/long_form/audio/inflect-micro-v2-finalist.wav",
+      "asr": "whisper-large-v3",
+      "hypothesis": "At 7.45 on Tuesday morning, Morgan checked the maintenance report before the first delivery arrived. The document listed three unresolved items, a fluorescent lamp near the loading bay, a pressure sensor marked B17, and a package weighing 12.6 kilograms. Although the weather changed without warning, the technicians finished their inspection, called Saskatchewan headquarters, and sent the revised PDF before noon. Could the replacement parts arrive by Friday? The coordinator replied, yes, but only if invoice 88953 was approved before 5pm. After one final review, everyone agreed that clear punctuation, stable pacing, and accurate numbers mattered more than speaking quickly.",
+      "raw": {
+        "normalized_reference": "at seven forty-five on tuesday morning, morgan checked the maintenance report before the first delivery arrived. the document listed three unresolved items: a fluorescent lamp near the loading bay, a pressure sensor marked b seventeen, and a package weighing twelve point six kilograms. although the weather changed without warning, the technicians finished their inspection, called saskatchewan headquarters, and sent the revised pdf before noon. could the replacement parts arrive by friday? the coordinator replied, yes, but only if invoice eight eight nine five three was approved before five p.m. after one final review, everyone agreed that clear punctuation, stable pacing, and accurate numbers mattered more than speaking quickly.",
+        "normalized_hypothesis": "at 7.45 on tuesday morning, morgan checked the maintenance report before the first delivery arrived. the document listed three unresolved items, a fluorescent lamp near the loading bay, a pressure sensor marked b17, and a package weighing 12.6 kilograms. although the weather changed without warning, the technicians finished their inspection, called saskatchewan headquarters, and sent the revised pdf before noon. could the replacement parts arrive by friday? the coordinator replied, yes, but only if invoice 88953 was approved before 5pm. after one final review, everyone agreed that clear punctuation, stable pacing, and accurate numbers mattered more than speaking quickly.",
+        "errors": 16,
+        "words": 110,
+        "wer": 0.14545454545454545
+      },
+      "semantic": {
+        "normalized_reference": "at 745 on tuesday morning morgan checked the maintenance report before the 1st delivery arrived the document listed 3 unresolved items a fluorescent lamp near the loading bay a pressure sensor marked b 17 and a package weighing 12.6 kilograms although the weather changed without warning the technicians finished their inspection called saskatchewan headquarters and sent the revised pdf before noon could the replacement parts arrive by friday the coordinator replied yes but only if invoice 88953 was approved before 5 p m after one final review everyone agreed that clear punctuation stable pacing and accurate numbers mattered more than speaking quickly",
+        "normalized_hypothesis": "at 7.45 on tuesday morning morgan checked the maintenance report before the 1st delivery arrived the document listed 3 unresolved items a fluorescent lamp near the loading bay a pressure sensor marked b 17 and a package weighing 12.6 kilograms although the weather changed without warning the technicians finished their inspection called saskatchewan headquarters and sent the revised pdf before noon could the replacement parts arrive by friday the coordinator replied yes but only if invoice 88953 was approved before 5 pm after one final review everyone agreed that clear punctuation stable pacing and accurate numbers mattered more than speaking quickly",
+        "errors": 4,
+        "words": 103,
+        "wer": 0.038834951456310676
+      }
+    },
+    {
+      "id": "release_long_form",
+      "category": "long_form",
+      "text": "At seven forty-five on Tuesday morning, Morgan checked the maintenance report before the first delivery arrived. The document listed three unresolved items: a fluorescent lamp near the loading bay, a pressure sensor marked B seventeen, and a package weighing twelve point six kilograms. Although the weather changed without warning, the technicians finished their inspection, called Saskatchewan headquarters, and sent the revised PDF before noon. Could the replacement parts arrive by Friday? The coordinator replied, yes, but only if invoice eight eight nine five three was approved before five P.M. After one final review, everyone agreed that clear punctuation, stable pacing, and accurate numbers mattered more than speaking quickly.",
+      "model": "inflect-nano-v2-finalist",
+      "audio": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/long_form/audio/inflect-nano-v2-finalist.wav",
+      "asr": "whisper-large-v3",
+      "hypothesis": "At 7.45 on Tuesday morning, Morgan checked the maintenance report before the first delivery arrived. The document listed three unresolved items, a fluorescent lamp near the loading bay, a pressure sensor marked B17, and a package weighing 12.6 kg. Although the weather changed without warning, the technicians finished their inspection, called Saskatchewan headquarters, and sent the revised PDEFF before noon. Could the replacement parts arrive by Friday? The coordinator replied, yes, but only if invoice 88953 was approved before 5pm. After one final review, everyone agreed that clear punctuation, stable pacing, and accurate numbers mattered more than speaking quickly.",
+      "raw": {
+        "normalized_reference": "at seven forty-five on tuesday morning, morgan checked the maintenance report before the first delivery arrived. the document listed three unresolved items: a fluorescent lamp near the loading bay, a pressure sensor marked b seventeen, and a package weighing twelve point six kilograms. although the weather changed without warning, the technicians finished their inspection, called saskatchewan headquarters, and sent the revised pdf before noon. could the replacement parts arrive by friday? the coordinator replied, yes, but only if invoice eight eight nine five three was approved before five p.m. after one final review, everyone agreed that clear punctuation, stable pacing, and accurate numbers mattered more than speaking quickly.",
+        "normalized_hypothesis": "at 7.45 on tuesday morning, morgan checked the maintenance report before the first delivery arrived. the document listed three unresolved items, a fluorescent lamp near the loading bay, a pressure sensor marked b17, and a package weighing 12.6 kg. although the weather changed without warning, the technicians finished their inspection, called saskatchewan headquarters, and sent the revised pdeff before noon. could the replacement parts arrive by friday? the coordinator replied, yes, but only if invoice 88953 was approved before 5pm. after one final review, everyone agreed that clear punctuation, stable pacing, and accurate numbers mattered more than speaking quickly.",
+        "errors": 18,
+        "words": 110,
+        "wer": 0.16363636363636364
+      },
+      "semantic": {
+        "normalized_reference": "at 745 on tuesday morning morgan checked the maintenance report before the 1st delivery arrived the document listed 3 unresolved items a fluorescent lamp near the loading bay a pressure sensor marked b 17 and a package weighing 12.6 kilograms although the weather changed without warning the technicians finished their inspection called saskatchewan headquarters and sent the revised pdf before noon could the replacement parts arrive by friday the coordinator replied yes but only if invoice 88953 was approved before 5 p m after one final review everyone agreed that clear punctuation stable pacing and accurate numbers mattered more than speaking quickly",
+        "normalized_hypothesis": "at 7.45 on tuesday morning morgan checked the maintenance report before the 1st delivery arrived the document listed 3 unresolved items a fluorescent lamp near the loading bay a pressure sensor marked b 17 and a package weighing 12.6 kg although the weather changed without warning the technicians finished their inspection called saskatchewan headquarters and sent the revised pdeff before noon could the replacement parts arrive by friday the coordinator replied yes but only if invoice 88953 was approved before 5 pm after one final review everyone agreed that clear punctuation stable pacing and accurate numbers mattered more than speaking quickly",
+        "errors": 6,
+        "words": 103,
+        "wer": 0.05825242718446602
+      }
+    }
+  ]
+}

evaluation/long_form_signal.json

@@ -0,0 +1,485 @@
+{
+  "format": "inflect_tts_signal_diagnostics_v1",
+  "manifest": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/long_form/combined_manifest.jsonl",
+  "metrics_are_diagnostic_not_mos": true,
+  "systems": {
+    "inflect-micro-v2-finalist": {
+      "count": {
+        "value": 1
+      },
+      "seconds": {
+        "mean": 44.37266666666667,
+        "median": 44.37266666666667,
+        "p05": 44.37266666666667,
+        "p95": 44.37266666666667
+      },
+      "words": {
+        "mean": 108.0,
+        "median": 108.0,
+        "p05": 108.0,
+        "p95": 108.0
+      },
+      "words_per_second": {
+        "mean": 2.4339307982391563,
+        "median": 2.4339307982391563,
+        "p05": 2.4339307982391563,
+        "p95": 2.4339307982391563
+      },
+      "peak_dbfs": {
+        "mean": -5.35886221619594,
+        "median": -5.35886221619594,
+        "p05": -5.35886221619594,
+        "p95": -5.35886221619594
+      },
+      "clip_fraction": {
+        "mean": 0.0,
+        "median": 0.0,
+        "p05": 0.0,
+        "p95": 0.0
+      },
+      "dc_offset": {
+        "mean": 3.754604676942108e-06,
+        "median": 3.754604676942108e-06,
+        "p05": 3.754604676942108e-06,
+        "p95": 3.754604676942108e-06
+      },
+      "leading_silence_seconds": {
+        "mean": 0.30933333333333335,
+        "median": 0.30933333333333335,
+        "p05": 0.30933333333333335,
+        "p95": 0.30933333333333335
+      },
+      "trailing_silence_seconds": {
+        "mean": 0.021333333333333333,
+        "median": 0.021333333333333333,
+        "p05": 0.021333333333333333,
+        "p95": 0.021333333333333333
+      },
+      "internal_silence_fraction": {
+        "mean": 0.13581730769230768,
+        "median": 0.13581730769230768,
+        "p05": 0.13581730769230768,
+        "p95": 0.13581730769230768
+      },
+      "tail_to_overall_rms_db": {
+        "mean": -48.11232409407665,
+        "median": -48.11232409407665,
+        "p05": -48.11232409407665,
+        "p95": -48.11232409407665
+      },
+      "active_hf_ratio_median": {
+        "mean": 7.739095963188447e-06,
+        "median": 7.739095963188447e-06,
+        "p05": 7.739095963188447e-06,
+        "p95": 7.739095963188447e-06
+      },
+      "active_flatness_median": {
+        "mean": 0.0004935810575261712,
+        "median": 0.0004935810575261712,
+        "p05": 0.0004935810575261712,
+        "p95": 0.0004935810575261712
+      },
+      "active_centroid_hz_median": {
+        "mean": 1304.1968200062613,
+        "median": 1304.1968200062613,
+        "p05": 1304.1968200062613,
+        "p95": 1304.1968200062613
+      },
+      "rms_jump_db_p95": {
+        "mean": 7.320098400115967,
+        "median": 7.320098400115967,
+        "p05": 7.320098400115967,
+        "p95": 7.320098400115967
+      },
+      "voiced_fraction": {
+        "mean": 0.854326923076923,
+        "median": 0.854326923076923,
+        "p05": 0.854326923076923,
+        "p95": 0.854326923076923
+      },
+      "f0_median_hz": {
+        "mean": 107.51896117937795,
+        "median": 107.51896117937795,
+        "p05": 107.51896117937795,
+        "p95": 107.51896117937795
+      },
+      "f0_range_semitones_p05_p95": {
+        "mean": 24.84172183582185,
+        "median": 24.84172183582185,
+        "p05": 24.84172183582185,
+        "p95": 24.84172183582185
+      },
+      "f0_jump_cents_p95": {
+        "mean": 1821.961141844264,
+        "median": 1821.961141844264,
+        "p05": 1821.961141844264,
+        "p95": 1821.961141844264
+      },
+      "f0_large_jump_fraction": {
+        "mean": 0.2225425194580571,
+        "median": 0.2225425194580571,
+        "p05": 0.2225425194580571,
+        "p95": 0.2225425194580571
+      }
+    },
+    "inflect-nano-v2-finalist": {
+      "count": {
+        "value": 1
+      },
+      "seconds": {
+        "mean": 43.967333333333336,
+        "median": 43.967333333333336,
+        "p05": 43.967333333333336,
+        "p95": 43.967333333333336
+      },
+      "words": {
+        "mean": 108.0,
+        "median": 108.0,
+        "p05": 108.0,
+        "p95": 108.0
+      },
+      "words_per_second": {
+        "mean": 2.456369122530363,
+        "median": 2.456369122530363,
+        "p05": 2.456369122530363,
+        "p95": 2.456369122530363
+      },
+      "peak_dbfs": {
+        "mean": -3.501703011153014,
+        "median": -3.501703011153014,
+        "p05": -3.501703011153014,
+        "p95": -3.501703011153014
+      },
+      "clip_fraction": {
+        "mean": 0.0,
+        "median": 0.0,
+        "p05": 0.0,
+        "p95": 0.0
+      },
+      "dc_offset": {
+        "mean": 5.1231389079475775e-05,
+        "median": 5.1231389079475775e-05,
+        "p05": 5.1231389079475775e-05,
+        "p95": 5.1231389079475775e-05
+      },
+      "leading_silence_seconds": {
+        "mean": 0.33066666666666666,
+        "median": 0.33066666666666666,
+        "p05": 0.33066666666666666,
+        "p95": 0.33066666666666666
+      },
+      "trailing_silence_seconds": {
+        "mean": 0.064,
+        "median": 0.064,
+        "p05": 0.064,
+        "p95": 0.064
+      },
+      "internal_silence_fraction": {
+        "mean": 0.13221737020863658,
+        "median": 0.13221737020863658,
+        "p05": 0.13221737020863658,
+        "p95": 0.13221737020863658
+      },
+      "tail_to_overall_rms_db": {
+        "mean": -59.80073796381536,
+        "median": -59.80073796381536,
+        "p05": -59.80073796381536,
+        "p95": -59.80073796381536
+      },
+      "active_hf_ratio_median": {
+        "mean": 7.782135980960447e-06,
+        "median": 7.782135980960447e-06,
+        "p05": 7.782135980960447e-06,
+        "p95": 7.782135980960447e-06
+      },
+      "active_flatness_median": {
+        "mean": 0.0004650004848372191,
+        "median": 0.0004650004848372191,
+        "p05": 0.0004650004848372191,
+        "p95": 0.0004650004848372191
+      },
+      "active_centroid_hz_median": {
+        "mean": 1315.4845530140253,
+        "median": 1315.4845530140253,
+        "p05": 1315.4845530140253,
+        "p95": 1315.4845530140253
+      },
+      "rms_jump_db_p95": {
+        "mean": 7.494379043579102,
+        "median": 7.494379043579102,
+        "p05": 7.494379043579102,
+        "p95": 7.494379043579102
+      },
+      "voiced_fraction": {
+        "mean": 0.8561377971858322,
+        "median": 0.8561377971858322,
+        "p05": 0.8561377971858322,
+        "p95": 0.8561377971858322
+      },
+      "f0_median_hz": {
+        "mean": 107.39862455247871,
+        "median": 107.39862455247871,
+        "p05": 107.39862455247871,
+        "p95": 107.39862455247871
+      },
+      "f0_range_semitones_p05_p95": {
+        "mean": 24.13550808891641,
+        "median": 24.13550808891641,
+        "p05": 24.13550808891641,
+        "p95": 24.13550808891641
+      },
+      "f0_jump_cents_p95": {
+        "mean": 1733.9331321194263,
+        "median": 1733.9331321194263,
+        "p05": 1733.9331321194263,
+        "p95": 1733.9331321194263
+      },
+      "f0_large_jump_fraction": {
+        "mean": 0.23409817020040663,
+        "median": 0.23409817020040663,
+        "p05": 0.23409817020040663,
+        "p95": 0.23409817020040663
+      }
+    }
+  },
+  "systems_by_category": {
+    "inflect-micro-v2-finalist::long_form": {
+      "count": {
+        "value": 1
+      },
+      "seconds": {
+        "mean": 44.37266666666667,
+        "median": 44.37266666666667,
+        "p05": 44.37266666666667,
+        "p95": 44.37266666666667
+      },
+      "words": {
+        "mean": 108.0,
+        "median": 108.0,
+        "p05": 108.0,
+        "p95": 108.0
+      },
+      "words_per_second": {
+        "mean": 2.4339307982391563,
+        "median": 2.4339307982391563,
+        "p05": 2.4339307982391563,
+        "p95": 2.4339307982391563
+      },
+      "peak_dbfs": {
+        "mean": -5.35886221619594,
+        "median": -5.35886221619594,
+        "p05": -5.35886221619594,
+        "p95": -5.35886221619594
+      },
+      "clip_fraction": {
+        "mean": 0.0,
+        "median": 0.0,
+        "p05": 0.0,
+        "p95": 0.0
+      },
+      "dc_offset": {
+        "mean": 3.754604676942108e-06,
+        "median": 3.754604676942108e-06,
+        "p05": 3.754604676942108e-06,
+        "p95": 3.754604676942108e-06
+      },
+      "leading_silence_seconds": {
+        "mean": 0.30933333333333335,
+        "median": 0.30933333333333335,
+        "p05": 0.30933333333333335,
+        "p95": 0.30933333333333335
+      },
+      "trailing_silence_seconds": {
+        "mean": 0.021333333333333333,
+        "median": 0.021333333333333333,
+        "p05": 0.021333333333333333,
+        "p95": 0.021333333333333333
+      },
+      "internal_silence_fraction": {
+        "mean": 0.13581730769230768,
+        "median": 0.13581730769230768,
+        "p05": 0.13581730769230768,
+        "p95": 0.13581730769230768
+      },
+      "tail_to_overall_rms_db": {
+        "mean": -48.11232409407665,
+        "median": -48.11232409407665,
+        "p05": -48.11232409407665,
+        "p95": -48.11232409407665
+      },
+      "active_hf_ratio_median": {
+        "mean": 7.739095963188447e-06,
+        "median": 7.739095963188447e-06,
+        "p05": 7.739095963188447e-06,
+        "p95": 7.739095963188447e-06
+      },
+      "active_flatness_median": {
+        "mean": 0.0004935810575261712,
+        "median": 0.0004935810575261712,
+        "p05": 0.0004935810575261712,
+        "p95": 0.0004935810575261712
+      },
+      "active_centroid_hz_median": {
+        "mean": 1304.1968200062613,
+        "median": 1304.1968200062613,
+        "p05": 1304.1968200062613,
+        "p95": 1304.1968200062613
+      },
+      "rms_jump_db_p95": {
+        "mean": 7.320098400115967,
+        "median": 7.320098400115967,
+        "p05": 7.320098400115967,
+        "p95": 7.320098400115967
+      },
+      "voiced_fraction": {
+        "mean": 0.854326923076923,
+        "median": 0.854326923076923,
+        "p05": 0.854326923076923,
+        "p95": 0.854326923076923
+      },
+      "f0_median_hz": {
+        "mean": 107.51896117937795,
+        "median": 107.51896117937795,
+        "p05": 107.51896117937795,
+        "p95": 107.51896117937795
+      },
+      "f0_range_semitones_p05_p95": {
+        "mean": 24.84172183582185,
+        "median": 24.84172183582185,
+        "p05": 24.84172183582185,
+        "p95": 24.84172183582185
+      },
+      "f0_jump_cents_p95": {
+        "mean": 1821.961141844264,
+        "median": 1821.961141844264,
+        "p05": 1821.961141844264,
+        "p95": 1821.961141844264
+      },
+      "f0_large_jump_fraction": {
+        "mean": 0.2225425194580571,
+        "median": 0.2225425194580571,
+        "p05": 0.2225425194580571,
+        "p95": 0.2225425194580571
+      }
+    },
+    "inflect-nano-v2-finalist::long_form": {
+      "count": {
+        "value": 1
+      },
+      "seconds": {
+        "mean": 43.967333333333336,
+        "median": 43.967333333333336,
+        "p05": 43.967333333333336,
+        "p95": 43.967333333333336
+      },
+      "words": {
+        "mean": 108.0,
+        "median": 108.0,
+        "p05": 108.0,
+        "p95": 108.0
+      },
+      "words_per_second": {
+        "mean": 2.456369122530363,
+        "median": 2.456369122530363,
+        "p05": 2.456369122530363,
+        "p95": 2.456369122530363
+      },
+      "peak_dbfs": {
+        "mean": -3.501703011153014,
+        "median": -3.501703011153014,
+        "p05": -3.501703011153014,
+        "p95": -3.501703011153014
+      },
+      "clip_fraction": {
+        "mean": 0.0,
+        "median": 0.0,
+        "p05": 0.0,
+        "p95": 0.0
+      },
+      "dc_offset": {
+        "mean": 5.1231389079475775e-05,
+        "median": 5.1231389079475775e-05,
+        "p05": 5.1231389079475775e-05,
+        "p95": 5.1231389079475775e-05
+      },
+      "leading_silence_seconds": {
+        "mean": 0.33066666666666666,
+        "median": 0.33066666666666666,
+        "p05": 0.33066666666666666,
+        "p95": 0.33066666666666666
+      },
+      "trailing_silence_seconds": {
+        "mean": 0.064,
+        "median": 0.064,
+        "p05": 0.064,
+        "p95": 0.064
+      },
+      "internal_silence_fraction": {
+        "mean": 0.13221737020863658,
+        "median": 0.13221737020863658,
+        "p05": 0.13221737020863658,
+        "p95": 0.13221737020863658
+      },
+      "tail_to_overall_rms_db": {
+        "mean": -59.80073796381536,
+        "median": -59.80073796381536,
+        "p05": -59.80073796381536,
+        "p95": -59.80073796381536
+      },
+      "active_hf_ratio_median": {
+        "mean": 7.782135980960447e-06,
+        "median": 7.782135980960447e-06,
+        "p05": 7.782135980960447e-06,
+        "p95": 7.782135980960447e-06
+      },
+      "active_flatness_median": {
+        "mean": 0.0004650004848372191,
+        "median": 0.0004650004848372191,
+        "p05": 0.0004650004848372191,
+        "p95": 0.0004650004848372191
+      },
+      "active_centroid_hz_median": {
+        "mean": 1315.4845530140253,
+        "median": 1315.4845530140253,
+        "p05": 1315.4845530140253,
+        "p95": 1315.4845530140253
+      },
+      "rms_jump_db_p95": {
+        "mean": 7.494379043579102,
+        "median": 7.494379043579102,
+        "p05": 7.494379043579102,
+        "p95": 7.494379043579102
+      },
+      "voiced_fraction": {
+        "mean": 0.8561377971858322,
+        "median": 0.8561377971858322,
+        "p05": 0.8561377971858322,
+        "p95": 0.8561377971858322
+      },
+      "f0_median_hz": {
+        "mean": 107.39862455247871,
+        "median": 107.39862455247871,
+        "p05": 107.39862455247871,
+        "p95": 107.39862455247871
+      },
+      "f0_range_semitones_p05_p95": {
+        "mean": 24.13550808891641,
+        "median": 24.13550808891641,
+        "p05": 24.13550808891641,
+        "p95": 24.13550808891641
+      },
+      "f0_jump_cents_p95": {
+        "mean": 1733.9331321194263,
+        "median": 1733.9331321194263,
+        "p05": 1733.9331321194263,
+        "p95": 1733.9331321194263
+      },
+      "f0_large_jump_fraction": {
+        "mean": 0.23409817020040663,
+        "median": 0.23409817020040663,
+        "p05": 0.23409817020040663,
+        "p95": 0.23409817020040663
+      }
+    }
+  }
+}

evaluation/multiseed_robustness.json

@@ -0,0 +1,780 @@
+{
+  "format": "inflect_v2_multiseed_robustness_v1",
+  "created_utc": "2026-07-21T08:20:18.083821+00:00",
+  "systems": {
+    "inflect-micro-v2-finalist": {
+      "prompts": 24,
+      "seeds_per_prompt": [
+        3
+      ],
+      "mean_seed_utterance_wer": 0.01917989417989418,
+      "mean_prompt_worst_seed_wer": 0.024735449735449735,
+      "every_seed_exact_prompt_rate": 0.8333333333333334,
+      "any_seed_exact_prompt_rate": 0.875,
+      "utmos22": {
+        "count": 72,
+        "mean": 4.4135655826992455,
+        "mean_bootstrap_95_ci": [
+          4.382303592231538,
+          4.439516715705395
+        ],
+        "median": 4.458602428436279,
+        "population_stddev": 0.12306422365741825,
+        "minimum": 3.8417863845825195,
+        "p10": 4.2751964092254635,
+        "p90": 4.491524410247803,
+        "maximum": 4.5073137283325195,
+        "rows": [
+          {
+            "id": "utmos200_001__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_001__seed1.wav",
+            "predicted_mos": 4.479095935821533
+          },
+          {
+            "id": "utmos200_015__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_015__seed1.wav",
+            "predicted_mos": 4.464743614196777
+          },
+          {
+            "id": "utmos200_029__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_029__seed1.wav",
+            "predicted_mos": 4.442873477935791
+          },
+          {
+            "id": "utmos200_002__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_002__seed1.wav",
+            "predicted_mos": 4.465550422668457
+          },
+          {
+            "id": "utmos200_016__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_016__seed1.wav",
+            "predicted_mos": 4.437236785888672
+          },
+          {
+            "id": "utmos200_030__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_030__seed1.wav",
+            "predicted_mos": 4.459320068359375
+          },
+          {
+            "id": "utmos200_003__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_003__seed1.wav",
+            "predicted_mos": 4.466009140014648
+          },
+          {
+            "id": "utmos200_017__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_017__seed1.wav",
+            "predicted_mos": 4.453671455383301
+          },
+          {
+            "id": "utmos200_031__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_031__seed1.wav",
+            "predicted_mos": 4.477704048156738
+          },
+          {
+            "id": "utmos200_004__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_004__seed1.wav",
+            "predicted_mos": 4.462135314941406
+          },
+          {
+            "id": "utmos200_018__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_018__seed1.wav",
+            "predicted_mos": 4.452676296234131
+          },
+          {
+            "id": "utmos200_032__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_032__seed1.wav",
+            "predicted_mos": 4.496460914611816
+          },
+          {
+            "id": "utmos200_005__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_005__seed1.wav",
+            "predicted_mos": 4.472134113311768
+          },
+          {
+            "id": "utmos200_019__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_019__seed1.wav",
+            "predicted_mos": 4.5073137283325195
+          },
+          {
+            "id": "utmos200_033__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_033__seed1.wav",
+            "predicted_mos": 4.449575424194336
+          },
+          {
+            "id": "utmos200_006__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_006__seed1.wav",
+            "predicted_mos": 4.430253505706787
+          },
+          {
+            "id": "utmos200_020__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_020__seed1.wav",
+            "predicted_mos": 4.480719566345215
+          },
+          {
+            "id": "utmos200_034__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_034__seed1.wav",
+            "predicted_mos": 4.179777145385742
+          },
+          {
+            "id": "utmos200_007__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_007__seed1.wav",
+            "predicted_mos": 4.272939205169678
+          },
+          {
+            "id": "utmos200_021__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_021__seed1.wav",
+            "predicted_mos": 4.438465118408203
+          },
+          {
+            "id": "utmos200_035__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_035__seed1.wav",
+            "predicted_mos": 4.481887340545654
+          },
+          {
+            "id": "utmos200_008__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_008__seed1.wav",
+            "predicted_mos": 4.464724063873291
+          },
+          {
+            "id": "utmos200_022__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_022__seed1.wav",
+            "predicted_mos": 4.458331108093262
+          },
+          {
+            "id": "utmos200_036__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_036__seed1.wav",
+            "predicted_mos": 4.492228031158447
+          },
+          {
+            "id": "utmos200_001__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_001__seed2.wav",
+            "predicted_mos": 4.4713592529296875
+          },
+          {
+            "id": "utmos200_015__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_015__seed2.wav",
+            "predicted_mos": 4.431799411773682
+          },
+          {
+            "id": "utmos200_029__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_029__seed2.wav",
+            "predicted_mos": 4.388747692108154
+          },
+          {
+            "id": "utmos200_002__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_002__seed2.wav",
+            "predicted_mos": 4.44416618347168
+          },
+          {
+            "id": "utmos200_016__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_016__seed2.wav",
+            "predicted_mos": 4.376438140869141
+          },
+          {
+            "id": "utmos200_030__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_030__seed2.wav",
+            "predicted_mos": 4.410470485687256
+          },
+          {
+            "id": "utmos200_003__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_003__seed2.wav",
+            "predicted_mos": 4.48537540435791
+          },
+          {
+            "id": "utmos200_017__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_017__seed2.wav",
+            "predicted_mos": 4.378164291381836
+          },
+          {
+            "id": "utmos200_031__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_031__seed2.wav",
+            "predicted_mos": 4.458873748779297
+          },
+          {
+            "id": "utmos200_004__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_004__seed2.wav",
+            "predicted_mos": 4.491084098815918
+          },
+          {
+            "id": "utmos200_018__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_018__seed2.wav",
+            "predicted_mos": 4.481396675109863
+          },
+          {
+            "id": "utmos200_032__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_032__seed2.wav",
+            "predicted_mos": 4.491573333740234
+          },
+          {
+            "id": "utmos200_005__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_005__seed2.wav",
+            "predicted_mos": 4.211251735687256
+          },
+          {
+            "id": "utmos200_019__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_019__seed2.wav",
+            "predicted_mos": 4.444100379943848
+          },
+          {
+            "id": "utmos200_033__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_033__seed2.wav",
+            "predicted_mos": 4.394369125366211
+          },
+          {
+            "id": "utmos200_006__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_006__seed2.wav",
+            "predicted_mos": 4.440512657165527
+          },
+          {
+            "id": "utmos200_020__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_020__seed2.wav",
+            "predicted_mos": 4.295511245727539
+          },
+          {
+            "id": "utmos200_034__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_034__seed2.wav",
+            "predicted_mos": 4.480815887451172
+          },
+          {
+            "id": "utmos200_007__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_007__seed2.wav",
+            "predicted_mos": 4.1143903732299805
+          },
+          {
+            "id": "utmos200_021__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_021__seed2.wav",
+            "predicted_mos": 4.363529205322266
+          },
+          {
+            "id": "utmos200_035__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_035__seed2.wav",
+            "predicted_mos": 4.498285293579102
+          },
+          {
+            "id": "utmos200_008__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_008__seed2.wav",
+            "predicted_mos": 4.207676887512207
+          },
+          {
+            "id": "utmos200_022__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_022__seed2.wav",
+            "predicted_mos": 4.459481239318848
+          },
+          {
+            "id": "utmos200_036__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_036__seed2.wav",
+            "predicted_mos": 4.419037342071533
+          },
+          {
+            "id": "utmos200_001__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_001__seed3.wav",
+            "predicted_mos": 4.466662406921387
+          },
+          {
+            "id": "utmos200_015__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_015__seed3.wav",
+            "predicted_mos": 4.47134256362915
+          },
+          {
+            "id": "utmos200_029__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_029__seed3.wav",
+            "predicted_mos": 4.4083757400512695
+          },
+          {
+            "id": "utmos200_002__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_002__seed3.wav",
+            "predicted_mos": 4.474235534667969
+          },
+          {
+            "id": "utmos200_016__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_016__seed3.wav",
+            "predicted_mos": 4.422906875610352
+          },
+          {
+            "id": "utmos200_030__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_030__seed3.wav",
+            "predicted_mos": 3.8417863845825195
+          },
+          {
+            "id": "utmos200_003__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_003__seed3.wav",
+            "predicted_mos": 4.448728561401367
+          },
+          {
+            "id": "utmos200_017__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_017__seed3.wav",
+            "predicted_mos": 4.433065414428711
+          },
+          {
+            "id": "utmos200_031__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_031__seed3.wav",
+            "predicted_mos": 4.487939834594727
+          },
+          {
+            "id": "utmos200_004__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_004__seed3.wav",
+            "predicted_mos": 4.489143371582031
+          },
+          {
+            "id": "utmos200_018__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_018__seed3.wav",
+            "predicted_mos": 4.462905406951904
+          },
+          {
+            "id": "utmos200_032__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_032__seed3.wav",
+            "predicted_mos": 4.034055233001709
+          },
+          {
+            "id": "utmos200_005__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_005__seed3.wav",
+            "predicted_mos": 4.493489742279053
+          },
+          {
+            "id": "utmos200_019__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_019__seed3.wav",
+            "predicted_mos": 4.3289031982421875
+          },
+          {
+            "id": "utmos200_033__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_033__seed3.wav",
+            "predicted_mos": 4.384283542633057
+          },
+          {
+            "id": "utmos200_006__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_006__seed3.wav",
+            "predicted_mos": 4.450536251068115
+          },
+          {
+            "id": "utmos200_020__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_020__seed3.wav",
+            "predicted_mos": 4.48296594619751
+          },
+          {
+            "id": "utmos200_034__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_034__seed3.wav",
+            "predicted_mos": 4.5004472732543945
+          },
+          {
+            "id": "utmos200_007__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_007__seed3.wav",
+            "predicted_mos": 4.473709583282471
+          },
+          {
+            "id": "utmos200_021__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_021__seed3.wav",
+            "predicted_mos": 4.41731595993042
+          },
+          {
+            "id": "utmos200_035__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_035__seed3.wav",
+            "predicted_mos": 4.485398292541504
+          },
+          {
+            "id": "utmos200_008__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_008__seed3.wav",
+            "predicted_mos": 4.00593900680542
+          },
+          {
+            "id": "utmos200_022__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_022__seed3.wav",
+            "predicted_mos": 4.498015403747559
+          },
+          {
+            "id": "utmos200_036__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-micro-v2-finalist/utmos200_036__seed3.wav",
+            "predicted_mos": 4.490334510803223
+          }
+        ]
+      }
+    },
+    "inflect-nano-v2-finalist": {
+      "prompts": 24,
+      "seeds_per_prompt": [
+        3
+      ],
+      "mean_seed_utterance_wer": 0.023434744268077602,
+      "mean_prompt_worst_seed_wer": 0.03574735449735449,
+      "every_seed_exact_prompt_rate": 0.7083333333333334,
+      "any_seed_exact_prompt_rate": 0.875,
+      "utmos22": {
+        "count": 72,
+        "mean": 4.400220692157745,
+        "mean_bootstrap_95_ci": [
+          4.364260655475987,
+          4.432265907691584
+        ],
+        "median": 4.45754075050354,
+        "population_stddev": 0.14739345703662027,
+        "minimum": 3.8460259437561035,
+        "p10": 4.210692977905273,
+        "p90": 4.497595262527466,
+        "maximum": 4.517411231994629,
+        "rows": [
+          {
+            "id": "utmos200_001__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_001__seed1.wav",
+            "predicted_mos": 4.470656394958496
+          },
+          {
+            "id": "utmos200_015__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_015__seed1.wav",
+            "predicted_mos": 4.481729030609131
+          },
+          {
+            "id": "utmos200_029__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_029__seed1.wav",
+            "predicted_mos": 4.423151969909668
+          },
+          {
+            "id": "utmos200_002__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_002__seed1.wav",
+            "predicted_mos": 4.426032066345215
+          },
+          {
+            "id": "utmos200_016__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_016__seed1.wav",
+            "predicted_mos": 4.440719127655029
+          },
+          {
+            "id": "utmos200_030__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_030__seed1.wav",
+            "predicted_mos": 4.457940101623535
+          },
+          {
+            "id": "utmos200_003__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_003__seed1.wav",
+            "predicted_mos": 4.460284233093262
+          },
+          {
+            "id": "utmos200_017__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_017__seed1.wav",
+            "predicted_mos": 4.345141410827637
+          },
+          {
+            "id": "utmos200_031__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_031__seed1.wav",
+            "predicted_mos": 4.147719383239746
+          },
+          {
+            "id": "utmos200_004__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_004__seed1.wav",
+            "predicted_mos": 4.4947710037231445
+          },
+          {
+            "id": "utmos200_018__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_018__seed1.wav",
+            "predicted_mos": 4.453065395355225
+          },
+          {
+            "id": "utmos200_032__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_032__seed1.wav",
+            "predicted_mos": 4.252050399780273
+          },
+          {
+            "id": "utmos200_005__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_005__seed1.wav",
+            "predicted_mos": 4.499449729919434
+          },
+          {
+            "id": "utmos200_019__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_019__seed1.wav",
+            "predicted_mos": 4.0041351318359375
+          },
+          {
+            "id": "utmos200_033__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_033__seed1.wav",
+            "predicted_mos": 4.467413902282715
+          },
+          {
+            "id": "utmos200_006__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_006__seed1.wav",
+            "predicted_mos": 4.457907199859619
+          },
+          {
+            "id": "utmos200_020__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_020__seed1.wav",
+            "predicted_mos": 4.439480304718018
+          },
+          {
+            "id": "utmos200_034__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_034__seed1.wav",
+            "predicted_mos": 4.427018165588379
+          },
+          {
+            "id": "utmos200_007__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_007__seed1.wav",
+            "predicted_mos": 4.497909069061279
+          },
+          {
+            "id": "utmos200_021__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_021__seed1.wav",
+            "predicted_mos": 3.8460259437561035
+          },
+          {
+            "id": "utmos200_035__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_035__seed1.wav",
+            "predicted_mos": 4.462929725646973
+          },
+          {
+            "id": "utmos200_008__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_008__seed1.wav",
+            "predicted_mos": 4.5093841552734375
+          },
+          {
+            "id": "utmos200_022__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_022__seed1.wav",
+            "predicted_mos": 4.457174301147461
+          },
+          {
+            "id": "utmos200_036__seed1",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_036__seed1.wav",
+            "predicted_mos": 4.4644694328308105
+          },
+          {
+            "id": "utmos200_001__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_001__seed2.wav",
+            "predicted_mos": 4.474266052246094
+          },
+          {
+            "id": "utmos200_015__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_015__seed2.wav",
+            "predicted_mos": 4.483236789703369
+          },
+          {
+            "id": "utmos200_029__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_029__seed2.wav",
+            "predicted_mos": 4.366954326629639
+          },
+          {
+            "id": "utmos200_002__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_002__seed2.wav",
+            "predicted_mos": 4.471959590911865
+          },
+          {
+            "id": "utmos200_016__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_016__seed2.wav",
+            "predicted_mos": 4.386756896972656
+          },
+          {
+            "id": "utmos200_030__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_030__seed2.wav",
+            "predicted_mos": 4.451584815979004
+          },
+          {
+            "id": "utmos200_003__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_003__seed2.wav",
+            "predicted_mos": 4.479098320007324
+          },
+          {
+            "id": "utmos200_017__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_017__seed2.wav",
+            "predicted_mos": 4.388742923736572
+          },
+          {
+            "id": "utmos200_031__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_031__seed2.wav",
+            "predicted_mos": 4.422547340393066
+          },
+          {
+            "id": "utmos200_004__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_004__seed2.wav",
+            "predicted_mos": 4.492393970489502
+          },
+          {
+            "id": "utmos200_018__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_018__seed2.wav",
+            "predicted_mos": 4.430690288543701
+          },
+          {
+            "id": "utmos200_032__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_032__seed2.wav",
+            "predicted_mos": 4.508492946624756
+          },
+          {
+            "id": "utmos200_005__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_005__seed2.wav",
+            "predicted_mos": 4.468206882476807
+          },
+          {
+            "id": "utmos200_019__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_019__seed2.wav",
+            "predicted_mos": 4.489481449127197
+          },
+          {
+            "id": "utmos200_033__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_033__seed2.wav",
+            "predicted_mos": 4.323983669281006
+          },
+          {
+            "id": "utmos200_006__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_006__seed2.wav",
+            "predicted_mos": 4.459604263305664
+          },
+          {
+            "id": "utmos200_020__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_020__seed2.wav",
+            "predicted_mos": 4.206554412841797
+          },
+          {
+            "id": "utmos200_034__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_034__seed2.wav",
+            "predicted_mos": 4.489374160766602
+          },
+          {
+            "id": "utmos200_007__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_007__seed2.wav",
+            "predicted_mos": 4.48223876953125
+          },
+          {
+            "id": "utmos200_021__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_021__seed2.wav",
+            "predicted_mos": 4.320951461791992
+          },
+          {
+            "id": "utmos200_035__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_035__seed2.wav",
+            "predicted_mos": 4.4568071365356445
+          },
+          {
+            "id": "utmos200_008__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_008__seed2.wav",
+            "predicted_mos": 4.50920295715332
+          },
+          {
+            "id": "utmos200_022__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_022__seed2.wav",
+            "predicted_mos": 3.9364013671875
+          },
+          {
+            "id": "utmos200_036__seed2",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_036__seed2.wav",
+            "predicted_mos": 4.484243392944336
+          },
+          {
+            "id": "utmos200_001__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_001__seed3.wav",
+            "predicted_mos": 4.443660736083984
+          },
+          {
+            "id": "utmos200_015__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_015__seed3.wav",
+            "predicted_mos": 4.5000810623168945
+          },
+          {
+            "id": "utmos200_029__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_029__seed3.wav",
+            "predicted_mos": 4.402558326721191
+          },
+          {
+            "id": "utmos200_002__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_002__seed3.wav",
+            "predicted_mos": 4.412972450256348
+          },
+          {
+            "id": "utmos200_016__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_016__seed3.wav",
+            "predicted_mos": 4.423831939697266
+          },
+          {
+            "id": "utmos200_030__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_030__seed3.wav",
+            "predicted_mos": 4.4557976722717285
+          },
+          {
+            "id": "utmos200_003__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_003__seed3.wav",
+            "predicted_mos": 4.485733985900879
+          },
+          {
+            "id": "utmos200_017__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_017__seed3.wav",
+            "predicted_mos": 4.366256237030029
+          },
+          {
+            "id": "utmos200_031__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_031__seed3.wav",
+            "predicted_mos": 4.488124847412109
+          },
+          {
+            "id": "utmos200_004__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_004__seed3.wav",
+            "predicted_mos": 4.489302158355713
+          },
+          {
+            "id": "utmos200_018__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_018__seed3.wav",
+            "predicted_mos": 4.480629920959473
+          },
+          {
+            "id": "utmos200_032__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_032__seed3.wav",
+            "predicted_mos": 4.489974498748779
+          },
+          {
+            "id": "utmos200_005__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_005__seed3.wav",
+            "predicted_mos": 4.097378730773926
+          },
+          {
+            "id": "utmos200_019__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_019__seed3.wav",
+            "predicted_mos": 4.392618179321289
+          },
+          {
+            "id": "utmos200_033__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_033__seed3.wav",
+            "predicted_mos": 4.492916584014893
+          },
+          {
+            "id": "utmos200_006__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_006__seed3.wav",
+            "predicted_mos": 4.4339423179626465
+          },
+          {
+            "id": "utmos200_020__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_020__seed3.wav",
+            "predicted_mos": 4.09462308883667
+          },
+          {
+            "id": "utmos200_034__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_034__seed3.wav",
+            "predicted_mos": 4.2479400634765625
+          },
+          {
+            "id": "utmos200_007__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_007__seed3.wav",
+            "predicted_mos": 4.473614692687988
+          },
+          {
+            "id": "utmos200_021__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_021__seed3.wav",
+            "predicted_mos": 4.390239238739014
+          },
+          {
+            "id": "utmos200_035__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_035__seed3.wav",
+            "predicted_mos": 4.501547813415527
+          },
+          {
+            "id": "utmos200_008__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_008__seed3.wav",
+            "predicted_mos": 4.517411231994629
+          },
+          {
+            "id": "utmos200_022__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_022__seed3.wav",
+            "predicted_mos": 3.8979973793029785
+          },
+          {
+            "id": "utmos200_036__seed3",
+            "file_path": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/audio/inflect-nano-v2-finalist/utmos200_036__seed3.wav",
+            "predicted_mos": 4.466404914855957
+          }
+        ]
+      }
+    }
+  },
+  "asr_report": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/asr/report.json",
+  "utmos_report": "release_polish_overnight_20260720/extended_benchmarks/multiseed_robustness/utmos22/report.json"
+}

evaluation/nisqa.json

@@ -0,0 +1,1997 @@
+{
+  "format": "inflect_tts_nisqa_v2",
+  "manifest": "release_polish_overnight_20260720/final_large200/combined_manifest.jsonl",
+  "sample_rate": 48000,
+  "dimensions": [
+    "mos",
+    "noisiness",
+    "discontinuity",
+    "coloration",
+    "loudness"
+  ],
+  "scope_note": "NISQA is an automated diagnostic and does not replace human TTS MOS.",
+  "systems": {
+    "inflect-micro-v2-finalist": {
+      "count": 200,
+      "mos": {
+        "mean": 5.066970911026001,
+        "median": 5.078833341598511,
+        "p05": 4.961752128601074,
+        "p95": 5.1485018014907835
+      },
+      "noisiness": {
+        "mean": 4.676522171497345,
+        "median": 4.6842615604400635,
+        "p05": 4.595613098144531,
+        "p95": 4.7530937671661375
+      },
+      "discontinuity": {
+        "mean": 4.893943982124329,
+        "median": 4.903385639190674,
+        "p05": 4.811065721511841,
+        "p95": 4.94633162021637
+      },
+      "coloration": {
+        "mean": 4.627148995399475,
+        "median": 4.634049654006958,
+        "p05": 4.566201853752136,
+        "p95": 4.674837756156921
+      },
+      "loudness": {
+        "mean": 4.75942345380783,
+        "median": 4.76876163482666,
+        "p05": 4.671535444259644,
+        "p95": 4.816184568405151
+      }
+    },
+    "inflect-nano-v2-finalist": {
+      "count": 200,
+      "mos": {
+        "mean": 5.05525309085846,
+        "median": 5.075676441192627,
+        "p05": 4.897062230110168,
+        "p95": 5.164716577529907
+      },
+      "noisiness": {
+        "mean": 4.667827854156494,
+        "median": 4.6796605587005615,
+        "p05": 4.581105875968933,
+        "p95": 4.7437347888946535
+      },
+      "discontinuity": {
+        "mean": 4.877564997673034,
+        "median": 4.899646997451782,
+        "p05": 4.796159672737121,
+        "p95": 4.948808050155639
+      },
+      "coloration": {
+        "mean": 4.615852819681168,
+        "median": 4.630635738372803,
+        "p05": 4.554482650756836,
+        "p95": 4.682383918762207
+      },
+      "loudness": {
+        "mean": 4.752314190864563,
+        "median": 4.773461818695068,
+        "p05": 4.634638261795044,
+        "p95": 4.819168114662171
+      }
+    },
+    "kitten-nano-bruno": {
+      "count": 200,
+      "mos": {
+        "mean": 4.296343866586685,
+        "median": 4.382935047149658,
+        "p05": 3.6946617722511292,
+        "p95": 4.641563630104065
+      },
+      "noisiness": {
+        "mean": 4.217127414941788,
+        "median": 4.269944429397583,
+        "p05": 3.7426937341690065,
+        "p95": 4.475915861129761
+      },
+      "discontinuity": {
+        "mean": 4.458463145494461,
+        "median": 4.521968603134155,
+        "p05": 4.075259256362915,
+        "p95": 4.675390267372131
+      },
+      "coloration": {
+        "mean": 4.04601433634758,
+        "median": 4.130239486694336,
+        "p05": 3.3240675568580627,
+        "p95": 4.339664316177368
+      },
+      "loudness": {
+        "mean": 4.311446378231048,
+        "median": 4.352288484573364,
+        "p05": 3.933194947242737,
+        "p95": 4.506394481658935
+      }
+    },
+    "piper-ryan-low": {
+      "count": 200,
+      "mos": {
+        "mean": 4.677609285116196,
+        "median": 4.746471166610718,
+        "p05": 4.175184011459351,
+        "p95": 4.969823598861694
+      },
+      "noisiness": {
+        "mean": 4.21767549276352,
+        "median": 4.272319793701172,
+        "p05": 3.704395222663879,
+        "p95": 4.5846161365509035
+      },
+      "discontinuity": {
+        "mean": 4.74571726322174,
+        "median": 4.798771619796753,
+        "p05": 4.451419758796692,
+        "p95": 4.914640951156616
+      },
+      "coloration": {
+        "mean": 4.336950881481171,
+        "median": 4.390848875045776,
+        "p05": 3.937146234512329,
+        "p95": 4.5602212905883786
+      },
+      "loudness": {
+        "mean": 4.334026387929916,
+        "median": 4.40546989440918,
+        "p05": 3.8327621936798097,
+        "p95": 4.62636170387268
+      }
+    }
+  },
+  "systems_by_category": {
+    "inflect-micro-v2-finalist::conversational": {
+      "count": 25,
+      "mos": {
+        "mean": 5.084625129699707,
+        "median": 5.088410377502441,
+        "p05": 5.009137058258057,
+        "p95": 5.160932826995849
+      },
+      "noisiness": {
+        "mean": 4.681895160675049,
+        "median": 4.686830997467041,
+        "p05": 4.587554836273194,
+        "p95": 4.76609468460083
+      },
+      "discontinuity": {
+        "mean": 4.904483070373535,
+        "median": 4.916714668273926,
+        "p05": 4.835331153869629,
+        "p95": 4.954121780395508
+      },
+      "coloration": {
+        "mean": 4.632850875854492,
+        "median": 4.62752628326416,
+        "p05": 4.59243221282959,
+        "p95": 4.674702262878418
+      },
+      "loudness": {
+        "mean": 4.751417484283447,
+        "median": 4.757721424102783,
+        "p05": 4.657126331329346,
+        "p95": 4.8176648139953615
+      }
+    },
+    "inflect-micro-v2-finalist::descriptive": {
+      "count": 16,
+      "mos": {
+        "mean": 5.080087423324585,
+        "median": 5.08701229095459,
+        "p05": 4.986552119255066,
+        "p95": 5.145286917686462
+      },
+      "noisiness": {
+        "mean": 4.681358724832535,
+        "median": 4.681260824203491,
+        "p05": 4.6196911334991455,
+        "p95": 4.759960651397705
+      },
+      "discontinuity": {
+        "mean": 4.892662018537521,
+        "median": 4.897983551025391,
+        "p05": 4.813859820365906,
+        "p95": 4.935849905014038
+      },
+      "coloration": {
+        "mean": 4.645889520645142,
+        "median": 4.651722431182861,
+        "p05": 4.59220552444458,
+        "p95": 4.679877758026123
+      },
+      "loudness": {
+        "mean": 4.780287712812424,
+        "median": 4.789277791976929,
+        "p05": 4.726202368736267,
+        "p95": 4.8107863664627075
+      }
+    },
+    "inflect-micro-v2-finalist::emotional": {
+      "count": 24,
+      "mos": {
+        "mean": 5.088000257809957,
+        "median": 5.0938146114349365,
+        "p05": 5.01492748260498,
+        "p95": 5.164720225334167
+      },
+      "noisiness": {
+        "mean": 4.688186625639598,
+        "median": 4.704174041748047,
+        "p05": 4.627037310600281,
+        "p95": 4.7350623369216915
+      },
+      "discontinuity": {
+        "mean": 4.915786584218343,
+        "median": 4.918968439102173,
+        "p05": 4.872726726531982,
+        "p95": 4.954227328300476
+      },
+      "coloration": {
+        "mean": 4.645064155260722,
+        "median": 4.647671222686768,
+        "p05": 4.613062524795533,
+        "p95": 4.6709935665130615
+      },
+      "loudness": {
+        "mean": 4.77682501077652,
+        "median": 4.781891107559204,
+        "p05": 4.7116050004959105,
+        "p95": 4.816433620452881
+      }
+    },
+    "inflect-micro-v2-finalist::homographs": {
+      "count": 6,
+      "mos": {
+        "mean": 5.071512937545776,
+        "median": 5.063185214996338,
+        "p05": 5.032201647758484,
+        "p95": 5.133068561553955
+      },
+      "noisiness": {
+        "mean": 4.6850950717926025,
+        "median": 4.690298557281494,
+        "p05": 4.648426294326782,
+        "p95": 4.7215656042099
+      },
+      "discontinuity": {
+        "mean": 4.8697404861450195,
+        "median": 4.881919860839844,
+        "p05": 4.785102367401123,
+        "p95": 4.932027578353882
+      },
+      "coloration": {
+        "mean": 4.63677438100179,
+        "median": 4.637820243835449,
+        "p05": 4.597443103790283,
+        "p95": 4.6673994064331055
+      },
+      "loudness": {
+        "mean": 4.751783529917399,
+        "median": 4.763185262680054,
+        "p05": 4.691569805145264,
+        "p95": 4.788424849510193
+      }
+    },
+    "inflect-micro-v2-finalist::long": {
+      "count": 15,
+      "mos": {
+        "mean": 5.060258356730143,
+        "median": 5.054421901702881,
+        "p05": 5.0101508617401125,
+        "p95": 5.117084503173828
+      },
+      "noisiness": {
+        "mean": 4.654302851359049,
+        "median": 4.660305976867676,
+        "p05": 4.600639009475708,
+        "p95": 4.6997123718261715
+      },
+      "discontinuity": {
+        "mean": 4.8977326075236,
+        "median": 4.895064830780029,
+        "p05": 4.8771138191223145,
+        "p95": 4.929110908508301
+      },
+      "coloration": {
+        "mean": 4.637342166900635,
+        "median": 4.638521671295166,
+        "p05": 4.607419109344482,
+        "p95": 4.663017654418946
+      },
+      "loudness": {
+        "mean": 4.76460796991984,
+        "median": 4.760266304016113,
+        "p05": 4.734184217453003,
+        "p95": 4.801061058044434
+      }
+    },
+    "inflect-micro-v2-finalist::long_range": {
+      "count": 6,
+      "mos": {
+        "mean": 5.100514650344849,
+        "median": 5.100477695465088,
+        "p05": 5.071790814399719,
+        "p95": 5.1220245361328125
+      },
+      "noisiness": {
+        "mean": 4.6899623076121015,
+        "median": 4.685410261154175,
+        "p05": 4.653762936592102,
+        "p95": 4.731876730918884
+      },
+      "discontinuity": {
+        "mean": 4.905282735824585,
+        "median": 4.907660722732544,
+        "p05": 4.873275279998779,
+        "p95": 4.9310362339019775
+      },
+      "coloration": {
+        "mean": 4.64597765604655,
+        "median": 4.651540756225586,
+        "p05": 4.622016668319702,
+        "p95": 4.662797212600708
+      },
+      "loudness": {
+        "mean": 4.777456760406494,
+        "median": 4.7702436447143555,
+        "p05": 4.7586283683776855,
+        "p95": 4.809190273284912
+      }
+    },
+    "inflect-micro-v2-finalist::names_places": {
+      "count": 6,
+      "mos": {
+        "mean": 5.066026131312053,
+        "median": 5.075019598007202,
+        "p05": 4.996745944023132,
+        "p95": 5.114146590232849
+      },
+      "noisiness": {
+        "mean": 4.686958074569702,
+        "median": 4.696692228317261,
+        "p05": 4.6366798877716064,
+        "p95": 4.726277112960815
+      },
+      "discontinuity": {
+        "mean": 4.901586612065633,
+        "median": 4.896145343780518,
+        "p05": 4.8806774616241455,
+        "p95": 4.926874995231628
+      },
+      "coloration": {
+        "mean": 4.630800088246663,
+        "median": 4.639263868331909,
+        "p05": 4.6018781661987305,
+        "p95": 4.64354145526886
+      },
+      "loudness": {
+        "mean": 4.764280796051025,
+        "median": 4.76242208480835,
+        "p05": 4.734339952468872,
+        "p95": 4.793020725250244
+      }
+    },
+    "inflect-micro-v2-finalist::narrative": {
+      "count": 24,
+      "mos": {
+        "mean": 5.087787350018819,
+        "median": 5.085803747177124,
+        "p05": 5.032889986038208,
+        "p95": 5.146023249626159
+      },
+      "noisiness": {
+        "mean": 4.690195957819621,
+        "median": 4.686437606811523,
+        "p05": 4.6295722961425785,
+        "p95": 4.769657254219055
+      },
+      "discontinuity": {
+        "mean": 4.902777910232544,
+        "median": 4.909550666809082,
+        "p05": 4.856229662895203,
+        "p95": 4.940158390998841
+      },
+      "coloration": {
+        "mean": 4.635043104489644,
+        "median": 4.634787321090698,
+        "p05": 4.597193908691406,
+        "p95": 4.667031073570252
+      },
+      "loudness": {
+        "mean": 4.767695367336273,
+        "median": 4.772120952606201,
+        "p05": 4.696179842948913,
+        "p95": 4.811451125144958
+      }
+    },
+    "inflect-micro-v2-finalist::numbers": {
+      "count": 6,
+      "mos": {
+        "mean": 5.027142604192098,
+        "median": 5.0750086307525635,
+        "p05": 4.768957138061523,
+        "p95": 5.158122301101685
+      },
+      "noisiness": {
+        "mean": 4.676971356074016,
+        "median": 4.707132339477539,
+        "p05": 4.529569625854492,
+        "p95": 4.754891514778137
+      },
+      "discontinuity": {
+        "mean": 4.855656226476033,
+        "median": 4.8860182762146,
+        "p05": 4.728206396102905,
+        "p95": 4.909510493278503
+      },
+      "coloration": {
+        "mean": 4.622106154759725,
+        "median": 4.644842147827148,
+        "p05": 4.52558708190918,
+        "p95": 4.668639302253723
+      },
+      "loudness": {
+        "mean": 4.763143301010132,
+        "median": 4.784612417221069,
+        "p05": 4.658777713775635,
+        "p95": 4.814141750335693
+      }
+    },
+    "inflect-micro-v2-finalist::punctuation": {
+      "count": 6,
+      "mos": {
+        "mean": 5.088138818740845,
+        "median": 5.094591856002808,
+        "p05": 5.051167130470276,
+        "p95": 5.1222697496414185
+      },
+      "noisiness": {
+        "mean": 4.704684654871623,
+        "median": 4.7150719165802,
+        "p05": 4.6764914989471436,
+        "p95": 4.72073769569397
+      },
+      "discontinuity": {
+        "mean": 4.90296729405721,
+        "median": 4.898035764694214,
+        "p05": 4.857134699821472,
+        "p95": 4.952345132827759
+      },
+      "coloration": {
+        "mean": 4.64106289545695,
+        "median": 4.6401262283325195,
+        "p05": 4.619646906852722,
+        "p95": 4.660913825035095
+      },
+      "loudness": {
+        "mean": 4.788663466771443,
+        "median": 4.781702518463135,
+        "p05": 4.768738746643066,
+        "p95": 4.814196944236755
+      }
+    },
+    "inflect-micro-v2-finalist::question": {
+      "count": 15,
+      "mos": {
+        "mean": 5.032029183705648,
+        "median": 5.039788722991943,
+        "p05": 4.93597674369812,
+        "p95": 5.123143720626831
+      },
+      "noisiness": {
+        "mean": 4.651882966359456,
+        "median": 4.659022331237793,
+        "p05": 4.59098072052002,
+        "p95": 4.698281526565552
+      },
+      "discontinuity": {
+        "mean": 4.8836393038431805,
+        "median": 4.8888983726501465,
+        "p05": 4.814854621887207,
+        "p95": 4.928069686889648
+      },
+      "coloration": {
+        "mean": 4.622194321950277,
+        "median": 4.621955871582031,
+        "p05": 4.573028326034546,
+        "p95": 4.69516134262085
+      },
+      "loudness": {
+        "mean": 4.736126931508382,
+        "median": 4.732144355773926,
+        "p05": 4.695382118225098,
+        "p95": 4.786236143112182
+      }
+    },
+    "inflect-micro-v2-finalist::rare_words": {
+      "count": 6,
+      "mos": {
+        "mean": 5.094142913818359,
+        "median": 5.094696283340454,
+        "p05": 5.068721175193787,
+        "p95": 5.115187883377075
+      },
+      "noisiness": {
+        "mean": 4.659267028172811,
+        "median": 4.673528671264648,
+        "p05": 4.602387070655823,
+        "p95": 4.70691704750061
+      },
+      "discontinuity": {
+        "mean": 4.917672395706177,
+        "median": 4.92179536819458,
+        "p05": 4.896824479103088,
+        "p95": 4.933650255203247
+      },
+      "coloration": {
+        "mean": 4.63420049349467,
+        "median": 4.6322290897369385,
+        "p05": 4.610973834991455,
+        "p95": 4.656691789627075
+      },
+      "loudness": {
+        "mean": 4.7646942138671875,
+        "median": 4.759293794631958,
+        "p05": 4.7546844482421875,
+        "p95": 4.784469723701477
+      }
+    },
+    "inflect-micro-v2-finalist::short": {
+      "count": 24,
+      "mos": {
+        "mean": 5.00382645924886,
+        "median": 5.0558905601501465,
+        "p05": 4.7033586502075195,
+        "p95": 5.122596263885498
+      },
+      "noisiness": {
+        "mean": 4.648739496866862,
+        "median": 4.684486627578735,
+        "p05": 4.342646741867066,
+        "p95": 4.755645680427551
+      },
+      "discontinuity": {
+        "mean": 4.857406497001648,
+        "median": 4.895836591720581,
+        "p05": 4.6450162649154665,
+        "p95": 4.954412913322448
+      },
+      "coloration": {
+        "mean": 4.56047276655833,
+        "median": 4.580313682556152,
+        "p05": 4.3241647481918335,
+        "p95": 4.665554308891297
+      },
+      "loudness": {
+        "mean": 4.7179668347040815,
+        "median": 4.747915506362915,
+        "p05": 4.540653777122498,
+        "p95": 4.818449997901917
+      }
+    },
+    "inflect-micro-v2-finalist::technical": {
+      "count": 21,
+      "mos": {
+        "mean": 5.077011244637625,
+        "median": 5.080066680908203,
+        "p05": 5.010731220245361,
+        "p95": 5.1484761238098145
+      },
+      "noisiness": {
+        "mean": 4.69018893014817,
+        "median": 4.693188190460205,
+        "p05": 4.631587505340576,
+        "p95": 4.7528228759765625
+      },
+      "discontinuity": {
+        "mean": 4.896800495329357,
+        "median": 4.901136875152588,
+        "p05": 4.8606648445129395,
+        "p95": 4.933905601501465
+      },
+      "coloration": {
+        "mean": 4.635323660714286,
+        "median": 4.639402389526367,
+        "p05": 4.605711936950684,
+        "p95": 4.683533668518066
+      },
+      "loudness": {
+        "mean": 4.768752438681466,
+        "median": 4.7783637046813965,
+        "p05": 4.717595100402832,
+        "p95": 4.8210554122924805
+      }
+    },
+    "inflect-nano-v2-finalist::conversational": {
+      "count": 25,
+      "mos": {
+        "mean": 5.096280899047851,
+        "median": 5.107004642486572,
+        "p05": 4.997963905334473,
+        "p95": 5.15976095199585
+      },
+      "noisiness": {
+        "mean": 4.687820911407471,
+        "median": 4.695287227630615,
+        "p05": 4.617572975158692,
+        "p95": 4.746697044372558
+      },
+      "discontinuity": {
+        "mean": 4.9022172164916995,
+        "median": 4.909136772155762,
+        "p05": 4.848705577850342,
+        "p95": 4.953138160705566
+      },
+      "coloration": {
+        "mean": 4.638227558135986,
+        "median": 4.635822296142578,
+        "p05": 4.595337581634522,
+        "p95": 4.6981409072875975
+      },
+      "loudness": {
+        "mean": 4.770219554901123,
+        "median": 4.775851726531982,
+        "p05": 4.677781009674073,
+        "p95": 4.831088256835938
+      }
+    },
+    "inflect-nano-v2-finalist::descriptive": {
+      "count": 16,
+      "mos": {
+        "mean": 5.080873131752014,
+        "median": 5.076714754104614,
+        "p05": 4.968343496322632,
+        "p95": 5.1620402336120605
+      },
+      "noisiness": {
+        "mean": 4.676590383052826,
+        "median": 4.682937383651733,
+        "p05": 4.589970350265503,
+        "p95": 4.762011528015137
+      },
+      "discontinuity": {
+        "mean": 4.890090584754944,
+        "median": 4.88960075378418,
+        "p05": 4.825415372848511,
+        "p95": 4.93328583240509
+      },
+      "coloration": {
+        "mean": 4.645678818225861,
+        "median": 4.652896165847778,
+        "p05": 4.596825480461121,
+        "p95": 4.675348997116089
+      },
+      "loudness": {
+        "mean": 4.7767637968063354,
+        "median": 4.792876958847046,
+        "p05": 4.6829081773757935,
+        "p95": 4.8231565952301025
+      }
+    },
+    "inflect-nano-v2-finalist::emotional": {
+      "count": 24,
+      "mos": {
+        "mean": 5.086991290251414,
+        "median": 5.082931995391846,
+        "p05": 5.011798572540283,
+        "p95": 5.184224057197571
+      },
+      "noisiness": {
+        "mean": 4.687529106934865,
+        "median": 4.690556764602661,
+        "p05": 4.652610564231873,
+        "p95": 4.7341481924057005
+      },
+      "discontinuity": {
+        "mean": 4.905580421288808,
+        "median": 4.907684564590454,
+        "p05": 4.844750595092774,
+        "p95": 4.955935835838318
+      },
+      "coloration": {
+        "mean": 4.6388090054194135,
+        "median": 4.638544797897339,
+        "p05": 4.582046365737915,
+        "p95": 4.678829598426819
+      },
+      "loudness": {
+        "mean": 4.771471639474233,
+        "median": 4.7630956172943115,
+        "p05": 4.732297706604004,
+        "p95": 4.81437463760376
+      }
+    },
+    "inflect-nano-v2-finalist::homographs": {
+      "count": 6,
+      "mos": {
+        "mean": 5.105078140894572,
+        "median": 5.101407051086426,
+        "p05": 5.05771005153656,
+        "p95": 5.166211366653442
+      },
+      "noisiness": {
+        "mean": 4.680643399556478,
+        "median": 4.714547395706177,
+        "p05": 4.599217414855957,
+        "p95": 4.730640530586243
+      },
+      "discontinuity": {
+        "mean": 4.894648313522339,
+        "median": 4.91188645362854,
+        "p05": 4.819316625595093,
+        "p95": 4.931559801101685
+      },
+      "coloration": {
+        "mean": 4.641849120457967,
+        "median": 4.6365134716033936,
+        "p05": 4.605284333229065,
+        "p95": 4.69017231464386
+      },
+      "loudness": {
+        "mean": 4.751724720001221,
+        "median": 4.740878582000732,
+        "p05": 4.723875880241394,
+        "p95": 4.797453761100769
+      }
+    },
+    "inflect-nano-v2-finalist::long": {
+      "count": 15,
+      "mos": {
+        "mean": 5.069772402445476,
+        "median": 5.073033809661865,
+        "p05": 5.036215829849243,
+        "p95": 5.104828643798828
+      },
+      "noisiness": {
+        "mean": 4.658925914764405,
+        "median": 4.658283233642578,
+        "p05": 4.625529336929321,
+        "p95": 4.685673522949219
+      },
+      "discontinuity": {
+        "mean": 4.893540032704672,
+        "median": 4.8910698890686035,
+        "p05": 4.875516033172607,
+        "p95": 4.91028470993042
+      },
+      "coloration": {
+        "mean": 4.634574890136719,
+        "median": 4.63601016998291,
+        "p05": 4.614579534530639,
+        "p95": 4.657125854492188
+      },
+      "loudness": {
+        "mean": 4.767716916402181,
+        "median": 4.775877475738525,
+        "p05": 4.734948587417603,
+        "p95": 4.792448711395264
+      }
+    },
+    "inflect-nano-v2-finalist::long_range": {
+      "count": 6,
+      "mos": {
+        "mean": 5.098030010859172,
+        "median": 5.107767343521118,
+        "p05": 5.051172614097595,
+        "p95": 5.140203595161438
+      },
+      "noisiness": {
+        "mean": 4.674017906188965,
+        "median": 4.665566921234131,
+        "p05": 4.6271902322769165,
+        "p95": 4.730582594871521
+      },
+      "discontinuity": {
+        "mean": 4.900173107783,
+        "median": 4.902195453643799,
+        "p05": 4.852349042892456,
+        "p95": 4.937187314033508
+      },
+      "coloration": {
+        "mean": 4.644658168156941,
+        "median": 4.6470630168914795,
+        "p05": 4.5962419509887695,
+        "p95": 4.689153432846069
+      },
+      "loudness": {
+        "mean": 4.755328257878621,
+        "median": 4.753864049911499,
+        "p05": 4.712657570838928,
+        "p95": 4.808353066444397
+      }
+    },
+    "inflect-nano-v2-finalist::names_places": {
+      "count": 6,
+      "mos": {
+        "mean": 5.064853111902873,
+        "median": 5.065268039703369,
+        "p05": 4.982016205787659,
+        "p95": 5.145555019378662
+      },
+      "noisiness": {
+        "mean": 4.667881488800049,
+        "median": 4.679285049438477,
+        "p05": 4.631013512611389,
+        "p95": 4.693566679954529
+      },
+      "discontinuity": {
+        "mean": 4.90459680557251,
+        "median": 4.908649206161499,
+        "p05": 4.873905539512634,
+        "p95": 4.930853366851807
+      },
+      "coloration": {
+        "mean": 4.628645420074463,
+        "median": 4.624447822570801,
+        "p05": 4.599899768829346,
+        "p95": 4.672908306121826
+      },
+      "loudness": {
+        "mean": 4.75518012046814,
+        "median": 4.750007629394531,
+        "p05": 4.713192820549011,
+        "p95": 4.8045361042022705
+      }
+    },
+    "inflect-nano-v2-finalist::narrative": {
+      "count": 24,
+      "mos": {
+        "mean": 5.045566161473592,
+        "median": 5.0831968784332275,
+        "p05": 5.014375329017639,
+        "p95": 5.1646970987319945
+      },
+      "noisiness": {
+        "mean": 4.664055486520131,
+        "median": 4.690380573272705,
+        "p05": 4.606567311286926,
+        "p95": 4.73886616230011
+      },
+      "discontinuity": {
+        "mean": 4.82183160384496,
+        "median": 4.902927398681641,
+        "p05": 4.848869848251343,
+        "p95": 4.942612719535828
+      },
+      "coloration": {
+        "mean": 4.604200392961502,
+        "median": 4.628083944320679,
+        "p05": 4.590324902534485,
+        "p95": 4.672721028327942
+      },
+      "loudness": {
+        "mean": 4.749283274014791,
+        "median": 4.772930860519409,
+        "p05": 4.7082852602005,
+        "p95": 4.806641840934754
+      }
+    },
+    "inflect-nano-v2-finalist::numbers": {
+      "count": 6,
+      "mos": {
+        "mean": 5.072862148284912,
+        "median": 5.077731609344482,
+        "p05": 5.019850969314575,
+        "p95": 5.106603622436523
+      },
+      "noisiness": {
+        "mean": 4.686765670776367,
+        "median": 4.68359637260437,
+        "p05": 4.616840839385986,
+        "p95": 4.766324877738953
+      },
+      "discontinuity": {
+        "mean": 4.886892000834147,
+        "median": 4.892426252365112,
+        "p05": 4.855258941650391,
+        "p95": 4.911754012107849
+      },
+      "coloration": {
+        "mean": 4.63871709505717,
+        "median": 4.637871742248535,
+        "p05": 4.6176005601882935,
+        "p95": 4.6588969230651855
+      },
+      "loudness": {
+        "mean": 4.781726042429606,
+        "median": 4.787912607192993,
+        "p05": 4.751368880271912,
+        "p95": 4.795805215835571
+      }
+    },
+    "inflect-nano-v2-finalist::punctuation": {
+      "count": 6,
+      "mos": {
+        "mean": 5.085851351420085,
+        "median": 5.0631513595581055,
+        "p05": 5.024835228919983,
+        "p95": 5.171986937522888
+      },
+      "noisiness": {
+        "mean": 4.703969558080037,
+        "median": 4.703524351119995,
+        "p05": 4.657694220542908,
+        "p95": 4.747106671333313
+      },
+      "discontinuity": {
+        "mean": 4.903408765792847,
+        "median": 4.898129224777222,
+        "p05": 4.876668572425842,
+        "p95": 4.934781432151794
+      },
+      "coloration": {
+        "mean": 4.637060165405273,
+        "median": 4.634654521942139,
+        "p05": 4.605042099952698,
+        "p95": 4.669329524040222
+      },
+      "loudness": {
+        "mean": 4.78837776184082,
+        "median": 4.79640531539917,
+        "p05": 4.761211514472961,
+        "p95": 4.803242802619934
+      }
+    },
+    "inflect-nano-v2-finalist::question": {
+      "count": 15,
+      "mos": {
+        "mean": 4.979924805959066,
+        "median": 5.0182952880859375,
+        "p05": 4.741870784759522,
+        "p95": 5.1213373184204105
+      },
+      "noisiness": {
+        "mean": 4.628494071960449,
+        "median": 4.635858058929443,
+        "p05": 4.521486711502075,
+        "p95": 4.701609420776367
+      },
+      "discontinuity": {
+        "mean": 4.85793924331665,
+        "median": 4.885674476623535,
+        "p05": 4.738079690933228,
+        "p95": 4.917272090911865
+      },
+      "coloration": {
+        "mean": 4.5775071144104,
+        "median": 4.617051601409912,
+        "p05": 4.444387197494507,
+        "p95": 4.643242597579956
+      },
+      "loudness": {
+        "mean": 4.703190549214681,
+        "median": 4.73654317855835,
+        "p05": 4.535270166397095,
+        "p95": 4.78399429321289
+      }
+    },
+    "inflect-nano-v2-finalist::rare_words": {
+      "count": 6,
+      "mos": {
+        "mean": 5.107121388117473,
+        "median": 5.117914915084839,
+        "p05": 5.063977837562561,
+        "p95": 5.129244446754456
+      },
+      "noisiness": {
+        "mean": 4.673150142033895,
+        "median": 4.663188695907593,
+        "p05": 4.640535593032837,
+        "p95": 4.726982116699219
+      },
+      "discontinuity": {
+        "mean": 4.929641803105672,
+        "median": 4.927665710449219,
+        "p05": 4.906194806098938,
+        "p95": 4.954469442367554
+      },
+      "coloration": {
+        "mean": 4.630059321721395,
+        "median": 4.638365030288696,
+        "p05": 4.594162583351135,
+        "p95": 4.654105544090271
+      },
+      "loudness": {
+        "mean": 4.7749260266621905,
+        "median": 4.776717901229858,
+        "p05": 4.744807958602905,
+        "p95": 4.799864172935486
+      }
+    },
+    "inflect-nano-v2-finalist::short": {
+      "count": 24,
+      "mos": {
+        "mean": 4.935086131095886,
+        "median": 5.009885549545288,
+        "p05": 4.485897254943848,
+        "p95": 5.139957475662231
+      },
+      "noisiness": {
+        "mean": 4.620879312356313,
+        "median": 4.650858163833618,
+        "p05": 4.42651948928833,
+        "p95": 4.743213105201721
+      },
+      "discontinuity": {
+        "mean": 4.819106638431549,
+        "median": 4.871162176132202,
+        "p05": 4.628170323371887,
+        "p95": 4.954963111877442
+      },
+      "coloration": {
+        "mean": 4.527366419633229,
+        "median": 4.593937635421753,
+        "p05": 4.329446601867676,
+        "p95": 4.632969546318054
+      },
+      "loudness": {
+        "mean": 4.681465268135071,
+        "median": 4.726132154464722,
+        "p05": 4.3697028160095215,
+        "p95": 4.821282315254211
+      }
+    },
+    "inflect-nano-v2-finalist::technical": {
+      "count": 21,
+      "mos": {
+        "mean": 5.084664026896159,
+        "median": 5.1049346923828125,
+        "p05": 5.009692192077637,
+        "p95": 5.1404852867126465
+      },
+      "noisiness": {
+        "mean": 4.684552351633708,
+        "median": 4.689305305480957,
+        "p05": 4.618290901184082,
+        "p95": 4.734136581420898
+      },
+      "discontinuity": {
+        "mean": 4.895777043842134,
+        "median": 4.899458408355713,
+        "p05": 4.872351169586182,
+        "p95": 4.929137229919434
+      },
+      "coloration": {
+        "mean": 4.632753735496884,
+        "median": 4.645188331604004,
+        "p05": 4.5678019523620605,
+        "p95": 4.675919532775879
+      },
+      "loudness": {
+        "mean": 4.77231688726516,
+        "median": 4.777512073516846,
+        "p05": 4.6978583335876465,
+        "p95": 4.826922416687012
+      }
+    },
+    "kitten-nano-bruno::conversational": {
+      "count": 25,
+      "mos": {
+        "mean": 4.268109455108642,
+        "median": 4.365891933441162,
+        "p05": 3.724268579483032,
+        "p95": 4.5813354492187495
+      },
+      "noisiness": {
+        "mean": 4.206264238357544,
+        "median": 4.268477439880371,
+        "p05": 3.781935453414917,
+        "p95": 4.451886940002441
+      },
+      "discontinuity": {
+        "mean": 4.43425760269165,
+        "median": 4.476420879364014,
+        "p05": 4.12585220336914,
+        "p95": 4.643235874176026
+      },
+      "coloration": {
+        "mean": 3.9630802345275877,
+        "median": 4.025370121002197,
+        "p05": 3.352046060562134,
+        "p95": 4.303129863739014
+      },
+      "loudness": {
+        "mean": 4.284803619384766,
+        "median": 4.315358638763428,
+        "p05": 3.9501728057861327,
+        "p95": 4.480254650115967
+      }
+    },
+    "kitten-nano-bruno::descriptive": {
+      "count": 16,
+      "mos": {
+        "mean": 4.337514206767082,
+        "median": 4.481396675109863,
+        "p05": 3.7846506237983704,
+        "p95": 4.6030203104019165
+      },
+      "noisiness": {
+        "mean": 4.246744692325592,
+        "median": 4.249030590057373,
+        "p05": 3.911725401878357,
+        "p95": 4.49030876159668
+      },
+      "discontinuity": {
+        "mean": 4.498975157737732,
+        "median": 4.531914949417114,
+        "p05": 4.22223973274231,
+        "p95": 4.714626669883728
+      },
+      "coloration": {
+        "mean": 4.064449533820152,
+        "median": 4.194642543792725,
+        "p05": 3.42701518535614,
+        "p95": 4.360018372535706
+      },
+      "loudness": {
+        "mean": 4.313280254602432,
+        "median": 4.363183498382568,
+        "p05": 4.054816246032715,
+        "p95": 4.472912549972534
+      }
+    },
+    "kitten-nano-bruno::emotional": {
+      "count": 24,
+      "mos": {
+        "mean": 4.236473699410756,
+        "median": 4.346853256225586,
+        "p05": 3.704057478904724,
+        "p95": 4.672033452987671
+      },
+      "noisiness": {
+        "mean": 4.175979008277257,
+        "median": 4.2271201610565186,
+        "p05": 3.8793925404548646,
+        "p95": 4.410620450973511
+      },
+      "discontinuity": {
+        "mean": 4.341713180144628,
+        "median": 4.4773268699646,
+        "p05": 3.63242826461792,
+        "p95": 4.699916076660156
+      },
+      "coloration": {
+        "mean": 4.048004696766536,
+        "median": 4.0794007778167725,
+        "p05": 3.7543776750564577,
+        "p95": 4.3167815685272215
+      },
+      "loudness": {
+        "mean": 4.310027033090591,
+        "median": 4.314358949661255,
+        "p05": 4.0880649328231815,
+        "p95": 4.5249312162399296
+      }
+    },
+    "kitten-nano-bruno::homographs": {
+      "count": 6,
+      "mos": {
+        "mean": 4.515197118123372,
+        "median": 4.531957626342773,
+        "p05": 4.369951367378235,
+        "p95": 4.628189921379089
+      },
+      "noisiness": {
+        "mean": 4.410176515579224,
+        "median": 4.436787366867065,
+        "p05": 4.270847797393799,
+        "p95": 4.505735397338867
+      },
+      "discontinuity": {
+        "mean": 4.575532118479411,
+        "median": 4.645455598831177,
+        "p05": 4.4101516008377075,
+        "p95": 4.667983055114746
+      },
+      "coloration": {
+        "mean": 4.233264247576396,
+        "median": 4.293304681777954,
+        "p05": 3.989873945713043,
+        "p95": 4.366684794425964
+      },
+      "loudness": {
+        "mean": 4.433592478434245,
+        "median": 4.456871032714844,
+        "p05": 4.31841254234314,
+        "p95": 4.521695613861084
+      }
+    },
+    "kitten-nano-bruno::long": {
+      "count": 15,
+      "mos": {
+        "mean": 4.36791795094808,
+        "median": 4.38620138168335,
+        "p05": 4.132020378112793,
+        "p95": 4.565616655349731
+      },
+      "noisiness": {
+        "mean": 4.249105930328369,
+        "median": 4.253856182098389,
+        "p05": 4.114522647857666,
+        "p95": 4.375577020645141
+      },
+      "discontinuity": {
+        "mean": 4.520841630299886,
+        "median": 4.542236804962158,
+        "p05": 4.3042370796203615,
+        "p95": 4.662956047058105
+      },
+      "coloration": {
+        "mean": 4.070761378606161,
+        "median": 4.074793815612793,
+        "p05": 3.8232643604278564,
+        "p95": 4.270243358612061
+      },
+      "loudness": {
+        "mean": 4.31005490620931,
+        "median": 4.333577632904053,
+        "p05": 4.141245031356812,
+        "p95": 4.459735298156738
+      }
+    },
+    "kitten-nano-bruno::long_range": {
+      "count": 6,
+      "mos": {
+        "mean": 4.384116888046265,
+        "median": 4.331946134567261,
+        "p05": 4.258500933647156,
+        "p95": 4.620538115501404
+      },
+      "noisiness": {
+        "mean": 4.247405370076497,
+        "median": 4.247298717498779,
+        "p05": 4.1250319480896,
+        "p95": 4.392456293106079
+      },
+      "discontinuity": {
+        "mean": 4.519463777542114,
+        "median": 4.500611782073975,
+        "p05": 4.430265426635742,
+        "p95": 4.656807541847229
+      },
+      "coloration": {
+        "mean": 4.125454783439636,
+        "median": 4.145297288894653,
+        "p05": 3.961834490299225,
+        "p95": 4.301606893539429
+      },
+      "loudness": {
+        "mean": 4.354299783706665,
+        "median": 4.348464012145996,
+        "p05": 4.246163487434387,
+        "p95": 4.494070291519165
+      }
+    },
+    "kitten-nano-bruno::names_places": {
+      "count": 6,
+      "mos": {
+        "mean": 4.3377305666605634,
+        "median": 4.296597480773926,
+        "p05": 4.132095217704773,
+        "p95": 4.623205065727234
+      },
+      "noisiness": {
+        "mean": 4.2723612785339355,
+        "median": 4.21521782875061,
+        "p05": 4.174685120582581,
+        "p95": 4.449698209762573
+      },
+      "discontinuity": {
+        "mean": 4.490620136260986,
+        "median": 4.47933554649353,
+        "p05": 4.330509066581726,
+        "p95": 4.68049693107605
+      },
+      "coloration": {
+        "mean": 4.108516176541646,
+        "median": 4.118963241577148,
+        "p05": 3.8788992762565613,
+        "p95": 4.328705430030823
+      },
+      "loudness": {
+        "mean": 4.319932619730632,
+        "median": 4.335159778594971,
+        "p05": 4.163914203643799,
+        "p95": 4.487408638000488
+      }
+    },
+    "kitten-nano-bruno::narrative": {
+      "count": 24,
+      "mos": {
+        "mean": 4.341499249140422,
+        "median": 4.40898585319519,
+        "p05": 3.928095757961273,
+        "p95": 4.632351756095886
+      },
+      "noisiness": {
+        "mean": 4.276511361201604,
+        "median": 4.305336952209473,
+        "p05": 4.077909183502197,
+        "p95": 4.424390602111816
+      },
+      "discontinuity": {
+        "mean": 4.492856323719025,
+        "median": 4.543771982192993,
+        "p05": 4.328080081939698,
+        "p95": 4.634611916542053
+      },
+      "coloration": {
+        "mean": 4.059429377317429,
+        "median": 4.1291587352752686,
+        "p05": 3.869328033924103,
+        "p95": 4.288502359390259
+      },
+      "loudness": {
+        "mean": 4.317196319500606,
+        "median": 4.3635547161102295,
+        "p05": 4.1294492244720455,
+        "p95": 4.472176361083984
+      }
+    },
+    "kitten-nano-bruno::numbers": {
+      "count": 6,
+      "mos": {
+        "mean": 4.490991989771525,
+        "median": 4.500324726104736,
+        "p05": 4.334450006484985,
+        "p95": 4.601772665977478
+      },
+      "noisiness": {
+        "mean": 4.35084851582845,
+        "median": 4.356521129608154,
+        "p05": 4.223582983016968,
+        "p95": 4.4524102210998535
+      },
+      "discontinuity": {
+        "mean": 4.5802295207977295,
+        "median": 4.584545850753784,
+        "p05": 4.5087326765060425,
+        "p95": 4.6444796323776245
+      },
+      "coloration": {
+        "mean": 4.193106333414714,
+        "median": 4.179413318634033,
+        "p05": 4.095273852348328,
+        "p95": 4.313513875007629
+      },
+      "loudness": {
+        "mean": 4.3892598152160645,
+        "median": 4.393868446350098,
+        "p05": 4.314112424850464,
+        "p95": 4.456575393676758
+      }
+    },
+    "kitten-nano-bruno::punctuation": {
+      "count": 6,
+      "mos": {
+        "mean": 4.444781303405762,
+        "median": 4.470665216445923,
+        "p05": 4.245614409446716,
+        "p95": 4.597299814224243
+      },
+      "noisiness": {
+        "mean": 4.302037676175435,
+        "median": 4.359004020690918,
+        "p05": 3.9859930872917175,
+        "p95": 4.48966372013092
+      },
+      "discontinuity": {
+        "mean": 4.546312729517619,
+        "median": 4.569594860076904,
+        "p05": 4.3923163414001465,
+        "p95": 4.6802650690078735
+      },
+      "coloration": {
+        "mean": 4.184681256612142,
+        "median": 4.184524297714233,
+        "p05": 4.094428181648254,
+        "p95": 4.266529321670532
+      },
+      "loudness": {
+        "mean": 4.397162278493245,
+        "median": 4.407179594039917,
+        "p05": 4.298161149024963,
+        "p95": 4.486984729766846
+      }
+    },
+    "kitten-nano-bruno::question": {
+      "count": 15,
+      "mos": {
+        "mean": 4.332177464167277,
+        "median": 4.3222737312316895,
+        "p05": 4.021507430076599,
+        "p95": 4.56958212852478
+      },
+      "noisiness": {
+        "mean": 4.289537938435872,
+        "median": 4.275187969207764,
+        "p05": 4.046637487411499,
+        "p95": 4.495771837234497
+      },
+      "discontinuity": {
+        "mean": 4.521408176422119,
+        "median": 4.539500713348389,
+        "p05": 4.275354957580566,
+        "p95": 4.666641426086426
+      },
+      "coloration": {
+        "mean": 4.084719880421956,
+        "median": 4.127988338470459,
+        "p05": 3.6590033531188966,
+        "p95": 4.30285677909851
+      },
+      "loudness": {
+        "mean": 4.327378718058268,
+        "median": 4.333405017852783,
+        "p05": 4.084225273132324,
+        "p95": 4.475044012069702
+      }
+    },
+    "kitten-nano-bruno::rare_words": {
+      "count": 6,
+      "mos": {
+        "mean": 4.481317122777303,
+        "median": 4.5381152629852295,
+        "p05": 4.20686674118042,
+        "p95": 4.621187210083008
+      },
+      "noisiness": {
+        "mean": 4.321086645126343,
+        "median": 4.344420909881592,
+        "p05": 4.198361873626709,
+        "p95": 4.388173937797546
+      },
+      "discontinuity": {
+        "mean": 4.587893724441528,
+        "median": 4.613213777542114,
+        "p05": 4.445345520973206,
+        "p95": 4.675289750099182
+      },
+      "coloration": {
+        "mean": 4.175944407780965,
+        "median": 4.221080780029297,
+        "p05": 3.944853186607361,
+        "p95": 4.305307149887085
+      },
+      "loudness": {
+        "mean": 4.419040679931641,
+        "median": 4.457917213439941,
+        "p05": 4.27720308303833,
+        "p95": 4.492039918899536
+      }
+    },
+    "kitten-nano-bruno::short": {
+      "count": 24,
+      "mos": {
+        "mean": 3.9046946863333383,
+        "median": 4.064854860305786,
+        "p05": 3.0507349491119387,
+        "p95": 4.409884333610535
+      },
+      "noisiness": {
+        "mean": 3.8803602953751883,
+        "median": 3.9190186262130737,
+        "p05": 3.4029449939727785,
+        "p95": 4.2835283279418945
+      },
+      "discontinuity": {
+        "mean": 4.235527475674947,
+        "median": 4.321232318878174,
+        "p05": 3.500464606285095,
+        "p95": 4.551818752288819
+      },
+      "coloration": {
+        "mean": 3.7684244414170585,
+        "median": 4.013571739196777,
+        "p05": 2.789597249031067,
+        "p95": 4.2459431171417235
+      },
+      "loudness": {
+        "mean": 4.146879871686299,
+        "median": 4.253454923629761,
+        "p05": 3.5685810327529905,
+        "p95": 4.415100407600403
+      }
+    },
+    "kitten-nano-bruno::technical": {
+      "count": 21,
+      "mos": {
+        "mean": 4.435978344508579,
+        "median": 4.519176006317139,
+        "p05": 4.158945083618164,
+        "p95": 4.643588066101074
+      },
+      "noisiness": {
+        "mean": 4.325209140777588,
+        "median": 4.366355895996094,
+        "p05": 4.123382091522217,
+        "p95": 4.441181182861328
+      },
+      "discontinuity": {
+        "mean": 4.558866659800212,
+        "median": 4.5804924964904785,
+        "p05": 4.4042816162109375,
+        "p95": 4.658003330230713
+      },
+      "coloration": {
+        "mean": 4.172192494074504,
+        "median": 4.218852519989014,
+        "p05": 3.8767950534820557,
+        "p95": 4.370019912719727
+      },
+      "loudness": {
+        "mean": 4.387475876581101,
+        "median": 4.432989597320557,
+        "p05": 4.208746433258057,
+        "p95": 4.494574546813965
+      }
+    },
+    "piper-ryan-low::conversational": {
+      "count": 25,
+      "mos": {
+        "mean": 4.654668464660644,
+        "median": 4.692073345184326,
+        "p05": 4.184446239471436,
+        "p95": 5.024080848693847
+      },
+      "noisiness": {
+        "mean": 4.193416728973388,
+        "median": 4.196901321411133,
+        "p05": 3.7953391075134277,
+        "p95": 4.59900016784668
+      },
+      "discontinuity": {
+        "mean": 4.751036472320557,
+        "median": 4.779134750366211,
+        "p05": 4.466301727294922,
+        "p95": 4.940692138671875
+      },
+      "coloration": {
+        "mean": 4.321975021362305,
+        "median": 4.331526756286621,
+        "p05": 3.9654213905334474,
+        "p95": 4.594906425476074
+      },
+      "loudness": {
+        "mean": 4.316827392578125,
+        "median": 4.353581428527832,
+        "p05": 3.9170026779174805,
+        "p95": 4.622183227539062
+      }
+    },
+    "piper-ryan-low::descriptive": {
+      "count": 16,
+      "mos": {
+        "mean": 4.697682768106461,
+        "median": 4.654010534286499,
+        "p05": 4.431982517242432,
+        "p95": 4.934651255607605
+      },
+      "noisiness": {
+        "mean": 4.193892315030098,
+        "median": 4.202427864074707,
+        "p05": 3.7581629157066345,
+        "p95": 4.605378746986389
+      },
+      "discontinuity": {
+        "mean": 4.752126306295395,
+        "median": 4.771055698394775,
+        "p05": 4.598942399024963,
+        "p95": 4.876826524734497
+      },
+      "coloration": {
+        "mean": 4.328488498926163,
+        "median": 4.333678245544434,
+        "p05": 4.077860951423645,
+        "p95": 4.560897469520569
+      },
+      "loudness": {
+        "mean": 4.340002968907356,
+        "median": 4.398552417755127,
+        "p05": 3.837493121623993,
+        "p95": 4.593310475349426
+      }
+    },
+    "piper-ryan-low::emotional": {
+      "count": 24,
+      "mos": {
+        "mean": 4.719764590263367,
+        "median": 4.709111452102661,
+        "p05": 4.455312633514405,
+        "p95": 4.959488725662231
+      },
+      "noisiness": {
+        "mean": 4.243843754132588,
+        "median": 4.317485332489014,
+        "p05": 3.901565968990326,
+        "p95": 4.546926045417786
+      },
+      "discontinuity": {
+        "mean": 4.782371163368225,
+        "median": 4.806643724441528,
+        "p05": 4.639827466011047,
+        "p95": 4.879727530479431
+      },
+      "coloration": {
+        "mean": 4.361620903015137,
+        "median": 4.40766978263855,
+        "p05": 4.126323080062866,
+        "p95": 4.530363917350769
+      },
+      "loudness": {
+        "mean": 4.347679088513057,
+        "median": 4.38455867767334,
+        "p05": 4.039763450622559,
+        "p95": 4.618425297737121
+      }
+    },
+    "piper-ryan-low::homographs": {
+      "count": 6,
+      "mos": {
+        "mean": 4.692897717158,
+        "median": 4.758767604827881,
+        "p05": 4.383318662643433,
+        "p95": 4.899083256721497
+      },
+      "noisiness": {
+        "mean": 4.185900688171387,
+        "median": 4.152200698852539,
+        "p05": 3.9394330978393555,
+        "p95": 4.451714992523193
+      },
+      "discontinuity": {
+        "mean": 4.782031615575154,
+        "median": 4.801535129547119,
+        "p05": 4.609615445137024,
+        "p95": 4.92097544670105
+      },
+      "coloration": {
+        "mean": 4.329089641571045,
+        "median": 4.369808197021484,
+        "p05": 4.0782365798950195,
+        "p95": 4.477747917175293
+      },
+      "loudness": {
+        "mean": 4.376166303952535,
+        "median": 4.473013401031494,
+        "p05": 4.039314329624176,
+        "p95": 4.565522193908691
+      }
+    },
+    "piper-ryan-low::long": {
+      "count": 15,
+      "mos": {
+        "mean": 4.80869197845459,
+        "median": 4.848377704620361,
+        "p05": 4.483882761001587,
+        "p95": 4.940733575820923
+      },
+      "noisiness": {
+        "mean": 4.2974871635437015,
+        "median": 4.354221343994141,
+        "p05": 3.9808048725128176,
+        "p95": 4.518116760253906
+      },
+      "discontinuity": {
+        "mean": 4.840127786000569,
+        "median": 4.840728759765625,
+        "p05": 4.760170793533325,
+        "p95": 4.909868764877319
+      },
+      "coloration": {
+        "mean": 4.4312187194824215,
+        "median": 4.4727349281311035,
+        "p05": 4.18483304977417,
+        "p95": 4.54548921585083
+      },
+      "loudness": {
+        "mean": 4.4669421195983885,
+        "median": 4.497457027435303,
+        "p05": 4.2082109451293945,
+        "p95": 4.63002724647522
+      }
+    },
+    "piper-ryan-low::long_range": {
+      "count": 6,
+      "mos": {
+        "mean": 4.930895169576009,
+        "median": 4.944577932357788,
+        "p05": 4.8467124700546265,
+        "p95": 4.971932411193848
+      },
+      "noisiness": {
+        "mean": 4.457035144170125,
+        "median": 4.476535320281982,
+        "p05": 4.3066534996032715,
+        "p95": 4.588317275047302
+      },
+      "discontinuity": {
+        "mean": 4.8793595631917315,
+        "median": 4.877874135971069,
+        "p05": 4.84382426738739,
+        "p95": 4.915496110916138
+      },
+      "coloration": {
+        "mean": 4.535036166508992,
+        "median": 4.538049221038818,
+        "p05": 4.491033673286438,
+        "p95": 4.584642171859741
+      },
+      "loudness": {
+        "mean": 4.5655380090077715,
+        "median": 4.5574610233306885,
+        "p05": 4.4656243324279785,
+        "p95": 4.68453574180603
+      }
+    },
+    "piper-ryan-low::names_places": {
+      "count": 6,
+      "mos": {
+        "mean": 4.7072140375773115,
+        "median": 4.796809196472168,
+        "p05": 4.30429208278656,
+        "p95": 4.9243165254592896
+      },
+      "noisiness": {
+        "mean": 4.2477126121521,
+        "median": 4.33577036857605,
+        "p05": 3.7818233966827393,
+        "p95": 4.517517924308777
+      },
+      "discontinuity": {
+        "mean": 4.770273605982463,
+        "median": 4.815858602523804,
+        "p05": 4.568361163139343,
+        "p95": 4.864709258079529
+      },
+      "coloration": {
+        "mean": 4.356080810228984,
+        "median": 4.430967807769775,
+        "p05": 4.043894469738007,
+        "p95": 4.509103417396545
+      },
+      "loudness": {
+        "mean": 4.3984503746032715,
+        "median": 4.465851306915283,
+        "p05": 4.12934935092926,
+        "p95": 4.550064206123352
+      }
+    },
+    "piper-ryan-low::narrative": {
+      "count": 24,
+      "mos": {
+        "mean": 4.748206456502278,
+        "median": 4.777817249298096,
+        "p05": 4.467332792282105,
+        "p95": 4.990657329559326
+      },
+      "noisiness": {
+        "mean": 4.260942389567693,
+        "median": 4.391161203384399,
+        "p05": 3.7142462134361267,
+        "p95": 4.530977749824524
+      },
+      "discontinuity": {
+        "mean": 4.759474148352941,
+        "median": 4.784025430679321,
+        "p05": 4.53602397441864,
+        "p95": 4.931457829475403
+      },
+      "coloration": {
+        "mean": 4.399409870306651,
+        "median": 4.427166938781738,
+        "p05": 4.16161630153656,
+        "p95": 4.595574378967285
+      },
+      "loudness": {
+        "mean": 4.404503921667735,
+        "median": 4.426621675491333,
+        "p05": 4.116325545310974,
+        "p95": 4.6394977331161495
+      }
+    },
+    "piper-ryan-low::numbers": {
+      "count": 6,
+      "mos": {
+        "mean": 4.82223637898763,
+        "median": 4.839604377746582,
+        "p05": 4.713255524635315,
+        "p95": 4.917005658149719
+      },
+      "noisiness": {
+        "mean": 4.235751112302144,
+        "median": 4.258164644241333,
+        "p05": 3.9398967623710632,
+        "p95": 4.51856803894043
+      },
+      "discontinuity": {
+        "mean": 4.80777891476949,
+        "median": 4.8165178298950195,
+        "p05": 4.730862617492676,
+        "p95": 4.8688424825668335
+      },
+      "coloration": {
+        "mean": 4.416639963785808,
+        "median": 4.4333577156066895,
+        "p05": 4.317197561264038,
+        "p95": 4.476765394210815
+      },
+      "loudness": {
+        "mean": 4.532079696655273,
+        "median": 4.558504343032837,
+        "p05": 4.4038039445877075,
+        "p95": 4.630873441696167
+      }
+    },
+    "piper-ryan-low::punctuation": {
+      "count": 6,
+      "mos": {
+        "mean": 4.782399574915568,
+        "median": 4.794174432754517,
+        "p05": 4.696298837661743,
+        "p95": 4.836727142333984
+      },
+      "noisiness": {
+        "mean": 4.349801619847615,
+        "median": 4.398617506027222,
+        "p05": 4.114583492279053,
+        "p95": 4.491045713424683
+      },
+      "discontinuity": {
+        "mean": 4.786481936772664,
+        "median": 4.811150312423706,
+        "p05": 4.700279116630554,
+        "p95": 4.841194748878479
+      },
+      "coloration": {
+        "mean": 4.436877568562825,
+        "median": 4.442990779876709,
+        "p05": 4.381941556930542,
+        "p95": 4.476019382476807
+      },
+      "loudness": {
+        "mean": 4.424544016520183,
+        "median": 4.472904443740845,
+        "p05": 4.189082622528076,
+        "p95": 4.53561532497406
+      }
+    },
+    "piper-ryan-low::question": {
+      "count": 15,
+      "mos": {
+        "mean": 4.661548042297364,
+        "median": 4.661092281341553,
+        "p05": 4.441859722137451,
+        "p95": 4.874482297897339
+      },
+      "noisiness": {
+        "mean": 4.0935604254404705,
+        "median": 4.123035430908203,
+        "p05": 3.751862072944641,
+        "p95": 4.468990182876587
+      },
+      "discontinuity": {
+        "mean": 4.797711404164632,
+        "median": 4.798983097076416,
+        "p05": 4.698680639266968,
+        "p95": 4.89758996963501
+      },
+      "coloration": {
+        "mean": 4.329330412546794,
+        "median": 4.322445869445801,
+        "p05": 4.124875593185425,
+        "p95": 4.51252236366272
+      },
+      "loudness": {
+        "mean": 4.3486167907714846,
+        "median": 4.344979763031006,
+        "p05": 4.18383936882019,
+        "p95": 4.517670488357544
+      }
+    },
+    "piper-ryan-low::rare_words": {
+      "count": 6,
+      "mos": {
+        "mean": 4.762843290964763,
+        "median": 4.740841388702393,
+        "p05": 4.658040523529053,
+        "p95": 4.8980982303619385
+      },
+      "noisiness": {
+        "mean": 4.2261199951171875,
+        "median": 4.204313039779663,
+        "p05": 4.032719373703003,
+        "p95": 4.419894814491272
+      },
+      "discontinuity": {
+        "mean": 4.827534755071004,
+        "median": 4.844349384307861,
+        "p05": 4.766901850700378,
+        "p95": 4.858636140823364
+      },
+      "coloration": {
+        "mean": 4.37746540705363,
+        "median": 4.3757643699646,
+        "p05": 4.299429416656494,
+        "p95": 4.457099676132202
+      },
+      "loudness": {
+        "mean": 4.444287300109863,
+        "median": 4.441998720169067,
+        "p05": 4.357715368270874,
+        "p95": 4.536978363990784
+      }
+    },
+    "piper-ryan-low::short": {
+      "count": 24,
+      "mos": {
+        "mean": 4.247917244831721,
+        "median": 4.321672201156616,
+        "p05": 3.153019607067108,
+        "p95": 4.8218111276626585
+      },
+      "noisiness": {
+        "mean": 3.9844569265842438,
+        "median": 3.9660487174987793,
+        "p05": 3.3954840779304503,
+        "p95": 4.581504940986633
+      },
+      "discontinuity": {
+        "mean": 4.4376387894153595,
+        "median": 4.600477457046509,
+        "p05": 3.160592067241669,
+        "p95": 4.861507606506348
+      },
+      "coloration": {
+        "mean": 4.042863676945369,
+        "median": 4.179455518722534,
+        "p05": 3.3848636388778686,
+        "p95": 4.436772871017456
+      },
+      "loudness": {
+        "mean": 3.869281530380249,
+        "median": 3.918291926383972,
+        "p05": 3.2730985641479493,
+        "p95": 4.5648229122161865
+      }
+    },
+    "piper-ryan-low::technical": {
+      "count": 21,
+      "mos": {
+        "mean": 4.788875466301327,
+        "median": 4.829463958740234,
+        "p05": 4.677733421325684,
+        "p95": 4.932690143585205
+      },
+      "noisiness": {
+        "mean": 4.370282320749192,
+        "median": 4.401300430297852,
+        "p05": 4.043017387390137,
+        "p95": 4.61635160446167
+      },
+      "discontinuity": {
+        "mean": 4.816073871794201,
+        "median": 4.8270368576049805,
+        "p05": 4.7223334312438965,
+        "p95": 4.894781112670898
+      },
+      "coloration": {
+        "mean": 4.413149697440011,
+        "median": 4.438601970672607,
+        "p05": 4.2930908203125,
+        "p95": 4.545473098754883
+      },
+      "loudness": {
+        "mean": 4.469029449281239,
+        "median": 4.480682849884033,
+        "p05": 4.3372321128845215,
+        "p95": 4.625380992889404
+      }
+    }
+  }
+}

evaluation/package_qa.json

@@ -0,0 +1,18 @@
+{
+  "format": "inflect_v2_release_package_qa_v1",
+  "package": "Inflect-Micro-v2",
+  "device": "cpu",
+  "sample_rate": 24000,
+  "manifest_files_verified": 27,
+  "load_seconds": 2.602567399997497,
+  "deterministic_same_seed": true,
+  "different_seed_changes_output": true,
+  "default_seconds": 4.096,
+  "slow_seconds": 2.912,
+  "fast_seconds": 2.1013333333333333,
+  "chunked_characters": 551,
+  "chunked_seconds": 34.90533333333333,
+  "default_waveform_sha256": "1e661f8d434358f243f926c2ad9c354cbd67552730be82b7000b4039718dcca7",
+  "peak": 0.6209497451782227,
+  "rms": 0.06188126638795305
+}

evaluation/runtime/cpu.json

@@ -0,0 +1,400 @@
+{
+  "format": "inflect_tts_runtime_v1",
+  "system": "micro180",
+  "provider": "cpu",
+  "utterances": 48,
+  "load_seconds": 1.2913346689892933,
+  "total_latency_seconds": 160.6842352389358,
+  "total_audio_seconds": 230.57066666666665,
+  "rtf": 0.6968979947099478,
+  "realtime_multiple": 1.4349302302357818,
+  "latency_mean_seconds": 3.347588234144496,
+  "latency_median_seconds": 3.2065962009946816,
+  "latency_p95_seconds": 4.397561740543461,
+  "rows": [
+    {
+      "id": "ood_001",
+      "category": "conversational",
+      "characters": 83,
+      "audio_seconds": 4.874666666666666,
+      "latency_seconds": 3.701560112996958,
+      "rtf": 0.7593463032679756
+    },
+    {
+      "id": "ood_002",
+      "category": "conversational",
+      "characters": 66,
+      "audio_seconds": 4.352,
+      "latency_seconds": 2.9957513550762087,
+      "rtf": 0.6883619841627318
+    },
+    {
+      "id": "ood_003",
+      "category": "conversational",
+      "characters": 81,
+      "audio_seconds": 4.682666666666667,
+      "latency_seconds": 4.003008865052834,
+      "rtf": 0.8548566767624218
+    },
+    {
+      "id": "ood_004",
+      "category": "conversational",
+      "characters": 77,
+      "audio_seconds": 4.426666666666667,
+      "latency_seconds": 2.894576665945351,
+      "rtf": 0.6538953311623534
+    },
+    {
+      "id": "ood_005",
+      "category": "conversational",
+      "characters": 63,
+      "audio_seconds": 3.925333333333333,
+      "latency_seconds": 2.8976474850205705,
+      "rtf": 0.738191444893148
+    },
+    {
+      "id": "ood_006",
+      "category": "conversational",
+      "characters": 65,
+      "audio_seconds": 3.8826666666666667,
+      "latency_seconds": 3.413119429955259,
+      "rtf": 0.8790657872480921
+    },
+    {
+      "id": "ood_007",
+      "category": "punctuation",
+      "characters": 94,
+      "audio_seconds": 6.272,
+      "latency_seconds": 4.897859386052005,
+      "rtf": 0.7809087031332916
+    },
+    {
+      "id": "ood_008",
+      "category": "punctuation",
+      "characters": 75,
+      "audio_seconds": 5.312,
+      "latency_seconds": 3.100503197987564,
+      "rtf": 0.5836790658862131
+    },
+    {
+      "id": "ood_009",
+      "category": "punctuation",
+      "characters": 63,
+      "audio_seconds": 3.904,
+      "latency_seconds": 2.68794972600881,
+      "rtf": 0.6885117126047157
+    },
+    {
+      "id": "ood_010",
+      "category": "punctuation",
+      "characters": 68,
+      "audio_seconds": 3.498666666666667,
+      "latency_seconds": 2.8092662420822307,
+      "rtf": 0.8029533847414912
+    },
+    {
+      "id": "ood_011",
+      "category": "punctuation",
+      "characters": 67,
+      "audio_seconds": 4.522666666666667,
+      "latency_seconds": 2.895203602965921,
+      "rtf": 0.6401540985331489
+    },
+    {
+      "id": "ood_012",
+      "category": "punctuation",
+      "characters": 73,
+      "audio_seconds": 4.469333333333333,
+      "latency_seconds": 2.89386862993706,
+      "rtf": 0.6474944726887814
+    },
+    {
+      "id": "ood_013",
+      "category": "numbers",
+      "characters": 96,
+      "audio_seconds": 5.834666666666666,
+      "latency_seconds": 3.6112114218994975,
+      "rtf": 0.6189233469891735
+    },
+    {
+      "id": "ood_014",
+      "category": "numbers",
+      "characters": 79,
+      "audio_seconds": 4.789333333333333,
+      "latency_seconds": 3.0953570218989626,
+      "rtf": 0.6463022735034025
+    },
+    {
+      "id": "ood_015",
+      "category": "numbers",
+      "characters": 80,
+      "audio_seconds": 4.288,
+      "latency_seconds": 2.897466129041277,
+      "rtf": 0.6757150487502978
+    },
+    {
+      "id": "ood_016",
+      "category": "numbers",
+      "characters": 78,
+      "audio_seconds": 4.864,
+      "latency_seconds": 3.198832436930388,
+      "rtf": 0.6576546950925963
+    },
+    {
+      "id": "ood_017",
+      "category": "numbers",
+      "characters": 80,
+      "audio_seconds": 4.576,
+      "latency_seconds": 3.2904248139820993,
+      "rtf": 0.7190613666918924
+    },
+    {
+      "id": "ood_018",
+      "category": "numbers",
+      "characters": 64,
+      "audio_seconds": 3.6053333333333333,
+      "latency_seconds": 3.0192351549630985,
+      "rtf": 0.837435786324824
+    },
+    {
+      "id": "ood_019",
+      "category": "technical",
+      "characters": 86,
+      "audio_seconds": 5.034666666666666,
+      "latency_seconds": 3.982155621983111,
+      "rtf": 0.7909472236460099
+    },
+    {
+      "id": "ood_020",
+      "category": "technical",
+      "characters": 93,
+      "audio_seconds": 5.514666666666667,
+      "latency_seconds": 3.6152250439627096,
+      "rtf": 0.6555654697708008
+    },
+    {
+      "id": "ood_021",
+      "category": "technical",
+      "characters": 89,
+      "audio_seconds": 5.44,
+      "latency_seconds": 3.5953551220009103,
+      "rtf": 0.6609108680148732
+    },
+    {
+      "id": "ood_022",
+      "category": "technical",
+      "characters": 86,
+      "audio_seconds": 5.354666666666667,
+      "latency_seconds": 3.214359965058975,
+      "rtf": 0.6002913281360137
+    },
+    {
+      "id": "ood_023",
+      "category": "technical",
+      "characters": 101,
+      "audio_seconds": 5.728,
+      "latency_seconds": 4.094791212002747,
+      "rtf": 0.7148727674585801
+    },
+    {
+      "id": "ood_024",
+      "category": "technical",
+      "characters": 82,
+      "audio_seconds": 5.397333333333333,
+      "latency_seconds": 3.7957619690569118,
+      "rtf": 0.703266175097007
+    },
+    {
+      "id": "ood_025",
+      "category": "names_places",
+      "characters": 71,
+      "audio_seconds": 4.074666666666666,
+      "latency_seconds": 2.8004974860232323,
+      "rtf": 0.6872948673159112
+    },
+    {
+      "id": "ood_026",
+      "category": "names_places",
+      "characters": 59,
+      "audio_seconds": 3.296,
+      "latency_seconds": 2.4975598730379716,
+      "rtf": 0.7577548158488991
+    },
+    {
+      "id": "ood_027",
+      "category": "names_places",
+      "characters": 72,
+      "audio_seconds": 3.936,
+      "latency_seconds": 2.894138294039294,
+      "rtf": 0.7352993633229914
+    },
+    {
+      "id": "ood_028",
+      "category": "names_places",
+      "characters": 61,
+      "audio_seconds": 3.616,
+      "latency_seconds": 2.8202893619891256,
+      "rtf": 0.7799472793111519
+    },
+    {
+      "id": "ood_029",
+      "category": "names_places",
+      "characters": 78,
+      "audio_seconds": 4.405333333333333,
+      "latency_seconds": 3.183753780904226,
+      "rtf": 0.7227043994183322
+    },
+    {
+      "id": "ood_030",
+      "category": "names_places",
+      "characters": 62,
+      "audio_seconds": 3.7973333333333334,
+      "latency_seconds": 3.5951710789231583,
+      "rtf": 0.9467620467669834
+    },
+    {
+      "id": "ood_031",
+      "category": "homographs",
+      "characters": 73,
+      "audio_seconds": 4.330666666666667,
+      "latency_seconds": 3.4116947089787573,
+      "rtf": 0.7877989629723116
+    },
+    {
+      "id": "ood_032",
+      "category": "homographs",
+      "characters": 78,
+      "audio_seconds": 3.968,
+      "latency_seconds": 3.682019405066967,
+      "rtf": 0.9279282774866349
+    },
+    {
+      "id": "ood_033",
+      "category": "homographs",
+      "characters": 60,
+      "audio_seconds": 3.36,
+      "latency_seconds": 2.7138323379913345,
+      "rtf": 0.8076881958307544
+    },
+    {
+      "id": "ood_034",
+      "category": "homographs",
+      "characters": 57,
+      "audio_seconds": 3.328,
+      "latency_seconds": 2.791520216036588,
+      "rtf": 0.8387981418379171
+    },
+    {
+      "id": "ood_035",
+      "category": "homographs",
+      "characters": 63,
+      "audio_seconds": 3.6266666666666665,
+      "latency_seconds": 2.8089981969678774,
+      "rtf": 0.774539944016878
+    },
+    {
+      "id": "ood_036",
+      "category": "homographs",
+      "characters": 51,
+      "audio_seconds": 2.9546666666666668,
+      "latency_seconds": 2.4995835940353572,
+      "rtf": 0.8459782019523998
+    },
+    {
+      "id": "ood_037",
+      "category": "long_range",
+      "characters": 122,
+      "audio_seconds": 6.496,
+      "latency_seconds": 4.397026888094842,
+      "rtf": 0.6768822179948957
+    },
+    {
+      "id": "ood_038",
+      "category": "long_range",
+      "characters": 141,
+      "audio_seconds": 7.8933333333333335,
+      "latency_seconds": 4.405088089988567,
+      "rtf": 0.5580770384276056
+    },
+    {
+      "id": "ood_039",
+      "category": "long_range",
+      "characters": 132,
+      "audio_seconds": 7.1146666666666665,
+      "latency_seconds": 4.397849738015793,
+      "rtf": 0.6181385501334042
+    },
+    {
+      "id": "ood_040",
+      "category": "long_range",
+      "characters": 122,
+      "audio_seconds": 6.176,
+      "latency_seconds": 3.894374215975404,
+      "rtf": 0.6305657733120796
+    },
+    {
+      "id": "ood_041",
+      "category": "long_range",
+      "characters": 124,
+      "audio_seconds": 7.424,
+      "latency_seconds": 4.009714661980979,
+      "rtf": 0.5401016516676965
+    },
+    {
+      "id": "ood_042",
+      "category": "long_range",
+      "characters": 124,
+      "audio_seconds": 6.261333333333333,
+      "latency_seconds": 4.010825427016243,
+      "rtf": 0.6405705004817255
+    },
+    {
+      "id": "ood_043",
+      "category": "rare_words",
+      "characters": 82,
+      "audio_seconds": 4.810666666666667,
+      "latency_seconds": 3.0846171140437946,
+      "rtf": 0.641203668384935
+    },
+    {
+      "id": "ood_044",
+      "category": "rare_words",
+      "characters": 78,
+      "audio_seconds": 4.512,
+      "latency_seconds": 2.8810565500753,
+      "rtf": 0.6385320368074691
+    },
+    {
+      "id": "ood_045",
+      "category": "rare_words",
+      "characters": 85,
+      "audio_seconds": 5.301333333333333,
+      "latency_seconds": 3.5225995020009577,
+      "rtf": 0.6644742521380077
+    },
+    {
+      "id": "ood_046",
+      "category": "rare_words",
+      "characters": 87,
+      "audio_seconds": 5.184,
+      "latency_seconds": 3.2946218359284103,
+      "rtf": 0.6355366195849557
+    },
+    {
+      "id": "ood_047",
+      "category": "rare_words",
+      "characters": 84,
+      "audio_seconds": 5.12,
+      "latency_seconds": 3.112685492960736,
+      "rtf": 0.6079463853438938
+    },
+    {
+      "id": "ood_048",
+      "category": "rare_words",
+      "characters": 90,
+      "audio_seconds": 5.034666666666666,
+      "latency_seconds": 3.3742267769994214,
+      "rtf": 0.6701986447959656
+    }
+  ]
+}

evaluation/runtime/cuda.json

@@ -0,0 +1,400 @@
+{
+  "format": "inflect_tts_runtime_v1",
+  "system": "micro180",
+  "provider": "cuda",
+  "utterances": 48,
+  "load_seconds": 1.362942800973542,
+  "total_latency_seconds": 3.877639473997988,
+  "total_audio_seconds": 230.59199999999998,
+  "rtf": 0.016816019089985725,
+  "realtime_multiple": 59.46710661119076,
+  "latency_mean_seconds": 0.08078415570829141,
+  "latency_median_seconds": 0.07884107751306146,
+  "latency_p95_seconds": 0.08621161602204666,
+  "rows": [
+    {
+      "id": "ood_001",
+      "category": "conversational",
+      "characters": 83,
+      "audio_seconds": 4.874666666666666,
+      "latency_seconds": 0.04850306594744325,
+      "rtf": 0.009950027204754497
+    },
+    {
+      "id": "ood_002",
+      "category": "conversational",
+      "characters": 66,
+      "audio_seconds": 4.352,
+      "latency_seconds": 0.04680907307192683,
+      "rtf": 0.01075576127571848
+    },
+    {
+      "id": "ood_003",
+      "category": "conversational",
+      "characters": 81,
+      "audio_seconds": 4.682666666666667,
+      "latency_seconds": 0.23080367501825094,
+      "rtf": 0.04928893971061737
+    },
+    {
+      "id": "ood_004",
+      "category": "conversational",
+      "characters": 77,
+      "audio_seconds": 4.426666666666667,
+      "latency_seconds": 0.07867441000416875,
+      "rtf": 0.017772833585279084
+    },
+    {
+      "id": "ood_005",
+      "category": "conversational",
+      "characters": 63,
+      "audio_seconds": 3.925333333333333,
+      "latency_seconds": 0.0762201170437038,
+      "rtf": 0.019417489056650084
+    },
+    {
+      "id": "ood_006",
+      "category": "conversational",
+      "characters": 65,
+      "audio_seconds": 3.8826666666666667,
+      "latency_seconds": 0.07715387898497283,
+      "rtf": 0.019871363062750558
+    },
+    {
+      "id": "ood_007",
+      "category": "punctuation",
+      "characters": 94,
+      "audio_seconds": 6.272,
+      "latency_seconds": 0.0797702909912914,
+      "rtf": 0.012718477517744164
+    },
+    {
+      "id": "ood_008",
+      "category": "punctuation",
+      "characters": 75,
+      "audio_seconds": 5.312,
+      "latency_seconds": 0.0779865289805457,
+      "rtf": 0.014681198979771404
+    },
+    {
+      "id": "ood_009",
+      "category": "punctuation",
+      "characters": 63,
+      "audio_seconds": 3.904,
+      "latency_seconds": 0.078226821962744,
+      "rtf": 0.02003760808471926
+    },
+    {
+      "id": "ood_010",
+      "category": "punctuation",
+      "characters": 68,
+      "audio_seconds": 3.498666666666667,
+      "latency_seconds": 0.07884166098665446,
+      "rtf": 0.022534773528959925
+    },
+    {
+      "id": "ood_011",
+      "category": "punctuation",
+      "characters": 67,
+      "audio_seconds": 4.522666666666667,
+      "latency_seconds": 0.07451220601797104,
+      "rtf": 0.01647528140137921
+    },
+    {
+      "id": "ood_012",
+      "category": "punctuation",
+      "characters": 73,
+      "audio_seconds": 4.469333333333333,
+      "latency_seconds": 0.07824564096517861,
+      "rtf": 0.01750722873624223
+    },
+    {
+      "id": "ood_013",
+      "category": "numbers",
+      "characters": 96,
+      "audio_seconds": 5.834666666666666,
+      "latency_seconds": 0.0768459300743416,
+      "rtf": 0.013170577595008273
+    },
+    {
+      "id": "ood_014",
+      "category": "numbers",
+      "characters": 79,
+      "audio_seconds": 4.789333333333333,
+      "latency_seconds": 0.07908053207211196,
+      "rtf": 0.016511803745569036
+    },
+    {
+      "id": "ood_015",
+      "category": "numbers",
+      "characters": 80,
+      "audio_seconds": 4.288,
+      "latency_seconds": 0.08663382101804018,
+      "rtf": 0.020203782886669817
+    },
+    {
+      "id": "ood_016",
+      "category": "numbers",
+      "characters": 78,
+      "audio_seconds": 4.864,
+      "latency_seconds": 0.07772980094887316,
+      "rtf": 0.01598063341876504
+    },
+    {
+      "id": "ood_017",
+      "category": "numbers",
+      "characters": 80,
+      "audio_seconds": 4.576,
+      "latency_seconds": 0.07818299101199955,
+      "rtf": 0.017085443840034867
+    },
+    {
+      "id": "ood_018",
+      "category": "numbers",
+      "characters": 64,
+      "audio_seconds": 3.6053333333333333,
+      "latency_seconds": 0.07488526694942266,
+      "rtf": 0.02077069164647448
+    },
+    {
+      "id": "ood_019",
+      "category": "technical",
+      "characters": 86,
+      "audio_seconds": 5.034666666666666,
+      "latency_seconds": 0.07327058492228389,
+      "rtf": 0.014553214695898549
+    },
+    {
+      "id": "ood_020",
+      "category": "technical",
+      "characters": 93,
+      "audio_seconds": 5.514666666666667,
+      "latency_seconds": 0.08311335602775216,
+      "rtf": 0.015071329066927978
+    },
+    {
+      "id": "ood_021",
+      "category": "technical",
+      "characters": 89,
+      "audio_seconds": 5.44,
+      "latency_seconds": 0.0815777350217104,
+      "rtf": 0.014995907173108527
+    },
+    {
+      "id": "ood_022",
+      "category": "technical",
+      "characters": 86,
+      "audio_seconds": 5.354666666666667,
+      "latency_seconds": 0.08467689098324627,
+      "rtf": 0.015813662409719795
+    },
+    {
+      "id": "ood_023",
+      "category": "technical",
+      "characters": 101,
+      "audio_seconds": 5.728,
+      "latency_seconds": 0.08542752102948725,
+      "rtf": 0.014914022526097635
+    },
+    {
+      "id": "ood_024",
+      "category": "technical",
+      "characters": 82,
+      "audio_seconds": 5.397333333333333,
+      "latency_seconds": 0.07980514410883188,
+      "rtf": 0.014786032134788516
+    },
+    {
+      "id": "ood_025",
+      "category": "names_places",
+      "characters": 71,
+      "audio_seconds": 4.074666666666666,
+      "latency_seconds": 0.08384539990220219,
+      "rtf": 0.02057724146814517
+    },
+    {
+      "id": "ood_026",
+      "category": "names_places",
+      "characters": 59,
+      "audio_seconds": 3.296,
+      "latency_seconds": 0.08088593499269336,
+      "rtf": 0.024540635616715222
+    },
+    {
+      "id": "ood_027",
+      "category": "names_places",
+      "characters": 72,
+      "audio_seconds": 3.936,
+      "latency_seconds": 0.08298896800260991,
+      "rtf": 0.021084595529118372
+    },
+    {
+      "id": "ood_028",
+      "category": "names_places",
+      "characters": 61,
+      "audio_seconds": 3.616,
+      "latency_seconds": 0.08673658303450793,
+      "rtf": 0.023986886901136043
+    },
+    {
+      "id": "ood_029",
+      "category": "names_places",
+      "characters": 78,
+      "audio_seconds": 4.405333333333333,
+      "latency_seconds": 0.08367316902149469,
+      "rtf": 0.018993606769407086
+    },
+    {
+      "id": "ood_030",
+      "category": "names_places",
+      "characters": 62,
+      "audio_seconds": 3.808,
+      "latency_seconds": 0.08070876495912671,
+      "rtf": 0.021194528613216048
+    },
+    {
+      "id": "ood_031",
+      "category": "homographs",
+      "characters": 73,
+      "audio_seconds": 4.330666666666667,
+      "latency_seconds": 0.0784687219420448,
+      "rtf": 0.0181193169509032
+    },
+    {
+      "id": "ood_032",
+      "category": "homographs",
+      "characters": 78,
+      "audio_seconds": 3.968,
+      "latency_seconds": 0.07884049403946847,
+      "rtf": 0.019869076118817657
+    },
+    {
+      "id": "ood_033",
+      "category": "homographs",
+      "characters": 60,
+      "audio_seconds": 3.36,
+      "latency_seconds": 0.08036918507423252,
+      "rtf": 0.02391940031971206
+    },
+    {
+      "id": "ood_034",
+      "category": "homographs",
+      "characters": 57,
+      "audio_seconds": 3.328,
+      "latency_seconds": 0.07696239301003516,
+      "rtf": 0.02312571905349614
+    },
+    {
+      "id": "ood_035",
+      "category": "homographs",
+      "characters": 63,
+      "audio_seconds": 3.6266666666666665,
+      "latency_seconds": 0.07384208799339831,
+      "rtf": 0.020360869851120858
+    },
+    {
+      "id": "ood_036",
+      "category": "homographs",
+      "characters": 51,
+      "audio_seconds": 2.9546666666666668,
+      "latency_seconds": 0.07758337398990989,
+      "rtf": 0.02625791087203629
+    },
+    {
+      "id": "ood_037",
+      "category": "long_range",
+      "characters": 122,
+      "audio_seconds": 6.496,
+      "latency_seconds": 0.08224317897111177,
+      "rtf": 0.012660587895799226
+    },
+    {
+      "id": "ood_038",
+      "category": "long_range",
+      "characters": 141,
+      "audio_seconds": 7.8933333333333335,
+      "latency_seconds": 0.08327149995602667,
+      "rtf": 0.010549598811996623
+    },
+    {
+      "id": "ood_039",
+      "category": "long_range",
+      "characters": 132,
+      "audio_seconds": 7.1146666666666665,
+      "latency_seconds": 0.08146983699407429,
+      "rtf": 0.011450970342120637
+    },
+    {
+      "id": "ood_040",
+      "category": "long_range",
+      "characters": 122,
+      "audio_seconds": 6.176,
+      "latency_seconds": 0.07936642598360777,
+      "rtf": 0.012850781409262916
+    },
+    {
+      "id": "ood_041",
+      "category": "long_range",
+      "characters": 124,
+      "audio_seconds": 7.424,
+      "latency_seconds": 0.07699621201027185,
+      "rtf": 0.010371257005693944
+    },
+    {
+      "id": "ood_042",
+      "category": "long_range",
+      "characters": 124,
+      "audio_seconds": 6.261333333333333,
+      "latency_seconds": 0.08013356593437493,
+      "rtf": 0.012798163213539437
+    },
+    {
+      "id": "ood_043",
+      "category": "rare_words",
+      "characters": 82,
+      "audio_seconds": 4.810666666666667,
+      "latency_seconds": 0.08174426795449108,
+      "rtf": 0.01699229516792359
+    },
+    {
+      "id": "ood_044",
+      "category": "rare_words",
+      "characters": 78,
+      "audio_seconds": 4.512,
+      "latency_seconds": 0.07456261909101158,
+      "rtf": 0.01652540316733413
+    },
+    {
+      "id": "ood_045",
+      "category": "rare_words",
+      "characters": 85,
+      "audio_seconds": 5.301333333333333,
+      "latency_seconds": 0.07449744909536093,
+      "rtf": 0.014052587228752692
+    },
+    {
+      "id": "ood_046",
+      "category": "rare_words",
+      "characters": 87,
+      "audio_seconds": 5.184,
+      "latency_seconds": 0.08111306896898896,
+      "rtf": 0.015646811143709288
+    },
+    {
+      "id": "ood_047",
+      "category": "rare_words",
+      "characters": 84,
+      "audio_seconds": 5.1306666666666665,
+      "latency_seconds": 0.07533228199463338,
+      "rtf": 0.014682747270263783
+    },
+    {
+      "id": "ood_048",
+      "category": "rare_words",
+      "characters": 90,
+      "audio_seconds": 5.034666666666666,
+      "latency_seconds": 0.055027046939358115,
+      "rtf": 0.010929630615603439
+    }
+  ]
+}

evaluation/speaker_consistency.json

@@ -0,0 +1,281 @@
+{
+  "format": "inflect_tts_speaker_consistency_ecapa_v1",
+  "encoder": "speechbrain/spkrec-ecapa-voxceleb",
+  "reference_filelist": "vits/filelists/inflect_qwenfull_dev_inflect_frontend.txt",
+  "reference_count": 32,
+  "scope_note": "Reference similarity is meaningful only for Inflect systems targeting this Mark teacher voice.",
+  "systems": {
+    "inflect-micro-v2-finalist": {
+      "count": 200,
+      "reference_cosine_mean": 0.904384995251894,
+      "reference_cosine_median": 0.922785758972168,
+      "reference_cosine_p05": 0.7935155183076859,
+      "reference_cosine_p95": 0.9520808905363083,
+      "system_centroid_cosine_mean": 0.9255932751297951,
+      "system_centroid_cosine_p05": 0.8265869051218033
+    },
+    "inflect-nano-v2-finalist": {
+      "count": 200,
+      "reference_cosine_mean": 0.8975338079035282,
+      "reference_cosine_median": 0.9178927540779114,
+      "reference_cosine_p05": 0.7353541612625122,
+      "reference_cosine_p95": 0.946146696805954,
+      "system_centroid_cosine_mean": 0.9236905339360237,
+      "system_centroid_cosine_p05": 0.7671847462654113
+    }
+  },
+  "systems_by_category": {
+    "inflect-micro-v2-finalist::conversational": {
+      "count": 25,
+      "reference_cosine_mean": 0.9139436888694763,
+      "reference_cosine_median": 0.9138887524604797,
+      "reference_cosine_p05": 0.8873096585273743,
+      "reference_cosine_p95": 0.9331717014312744,
+      "system_centroid_cosine_mean": 0.9376359748840332,
+      "system_centroid_cosine_p05": 0.9182322859764099
+    },
+    "inflect-micro-v2-finalist::descriptive": {
+      "count": 16,
+      "reference_cosine_mean": 0.9220070615410805,
+      "reference_cosine_median": 0.9219895601272583,
+      "reference_cosine_p05": 0.8994174748659134,
+      "reference_cosine_p95": 0.9472714811563492,
+      "system_centroid_cosine_mean": 0.9414824359118938,
+      "system_centroid_cosine_p05": 0.9153823703527451
+    },
+    "inflect-micro-v2-finalist::emotional": {
+      "count": 24,
+      "reference_cosine_mean": 0.9218592420220375,
+      "reference_cosine_median": 0.9236226677894592,
+      "reference_cosine_p05": 0.896009761095047,
+      "reference_cosine_p95": 0.9521152794361114,
+      "system_centroid_cosine_mean": 0.9431363741556803,
+      "system_centroid_cosine_p05": 0.9297195106744767
+    },
+    "inflect-micro-v2-finalist::homographs": {
+      "count": 6,
+      "reference_cosine_mean": 0.9016784826914469,
+      "reference_cosine_median": 0.9143826961517334,
+      "reference_cosine_p05": 0.8602327704429626,
+      "reference_cosine_p95": 0.9188010692596436,
+      "system_centroid_cosine_mean": 0.9227530062198639,
+      "system_centroid_cosine_p05": 0.8839313089847565
+    },
+    "inflect-micro-v2-finalist::long": {
+      "count": 15,
+      "reference_cosine_mean": 0.9424067616462708,
+      "reference_cosine_median": 0.9414758682250977,
+      "reference_cosine_p05": 0.9292309999465942,
+      "reference_cosine_p95": 0.9554343402385712,
+      "system_centroid_cosine_mean": 0.9673661112785339,
+      "system_centroid_cosine_p05": 0.9578824639320374
+    },
+    "inflect-micro-v2-finalist::long_range": {
+      "count": 6,
+      "reference_cosine_mean": 0.9468695918718973,
+      "reference_cosine_median": 0.9462248682975769,
+      "reference_cosine_p05": 0.9344874769449234,
+      "reference_cosine_p95": 0.9605438858270645,
+      "system_centroid_cosine_mean": 0.9628004630406698,
+      "system_centroid_cosine_p05": 0.9456458687782288
+    },
+    "inflect-micro-v2-finalist::names_places": {
+      "count": 6,
+      "reference_cosine_mean": 0.9157689611117045,
+      "reference_cosine_median": 0.9212036728858948,
+      "reference_cosine_p05": 0.8714424669742584,
+      "reference_cosine_p95": 0.9455099701881409,
+      "system_centroid_cosine_mean": 0.9366900821526846,
+      "system_centroid_cosine_p05": 0.8990786373615265
+    },
+    "inflect-micro-v2-finalist::narrative": {
+      "count": 24,
+      "reference_cosine_mean": 0.920630931854248,
+      "reference_cosine_median": 0.9232715964317322,
+      "reference_cosine_p05": 0.9030191957950592,
+      "reference_cosine_p95": 0.9469668209552765,
+      "system_centroid_cosine_mean": 0.9427467087904612,
+      "system_centroid_cosine_p05": 0.9199487030506134
+    },
+    "inflect-micro-v2-finalist::numbers": {
+      "count": 6,
+      "reference_cosine_mean": 0.9326508939266205,
+      "reference_cosine_median": 0.9313071966171265,
+      "reference_cosine_p05": 0.9172701835632324,
+      "reference_cosine_p95": 0.9509273022413254,
+      "system_centroid_cosine_mean": 0.951929897069931,
+      "system_centroid_cosine_p05": 0.9386754631996155
+    },
+    "inflect-micro-v2-finalist::punctuation": {
+      "count": 6,
+      "reference_cosine_mean": 0.9413094222545624,
+      "reference_cosine_median": 0.9380442202091217,
+      "reference_cosine_p05": 0.9343572407960892,
+      "reference_cosine_p95": 0.9544158428907394,
+      "system_centroid_cosine_mean": 0.9496991535027822,
+      "system_centroid_cosine_p05": 0.9365403354167938
+    },
+    "inflect-micro-v2-finalist::question": {
+      "count": 15,
+      "reference_cosine_mean": 0.891196087996165,
+      "reference_cosine_median": 0.891516387462616,
+      "reference_cosine_p05": 0.8434692203998566,
+      "reference_cosine_p95": 0.9273699462413788,
+      "system_centroid_cosine_mean": 0.9139200647672018,
+      "system_centroid_cosine_p05": 0.8653586804866791
+    },
+    "inflect-micro-v2-finalist::rare_words": {
+      "count": 6,
+      "reference_cosine_mean": 0.9439626634120941,
+      "reference_cosine_median": 0.944472998380661,
+      "reference_cosine_p05": 0.932221531867981,
+      "reference_cosine_p95": 0.9552849233150482,
+      "system_centroid_cosine_mean": 0.960921456416448,
+      "system_centroid_cosine_p05": 0.9522077441215515
+    },
+    "inflect-micro-v2-finalist::short": {
+      "count": 24,
+      "reference_cosine_mean": 0.7740145114560922,
+      "reference_cosine_median": 0.8164414763450623,
+      "reference_cosine_p05": 0.5830616354942322,
+      "reference_cosine_p95": 0.8809338510036469,
+      "system_centroid_cosine_mean": 0.800063061217467,
+      "system_centroid_cosine_p05": 0.617935448884964
+    },
+    "inflect-micro-v2-finalist::technical": {
+      "count": 21,
+      "reference_cosine_mean": 0.9277475504648118,
+      "reference_cosine_median": 0.9319859743118286,
+      "reference_cosine_p05": 0.9036513566970825,
+      "reference_cosine_p95": 0.9455512166023254,
+      "system_centroid_cosine_mean": 0.9439653442019508,
+      "system_centroid_cosine_p05": 0.9178051948547363
+    },
+    "inflect-nano-v2-finalist::conversational": {
+      "count": 25,
+      "reference_cosine_mean": 0.9092928338050842,
+      "reference_cosine_median": 0.9059870839118958,
+      "reference_cosine_p05": 0.8836398243904113,
+      "reference_cosine_p95": 0.9361859798431397,
+      "system_centroid_cosine_mean": 0.9373506665229797,
+      "system_centroid_cosine_p05": 0.9134238719940185
+    },
+    "inflect-nano-v2-finalist::descriptive": {
+      "count": 16,
+      "reference_cosine_mean": 0.9188673309981823,
+      "reference_cosine_median": 0.9206716418266296,
+      "reference_cosine_p05": 0.894992008805275,
+      "reference_cosine_p95": 0.9390376806259155,
+      "system_centroid_cosine_mean": 0.9431007876992226,
+      "system_centroid_cosine_p05": 0.9160878360271454
+    },
+    "inflect-nano-v2-finalist::emotional": {
+      "count": 24,
+      "reference_cosine_mean": 0.9127361178398132,
+      "reference_cosine_median": 0.9181386530399323,
+      "reference_cosine_p05": 0.8833694070577621,
+      "reference_cosine_p95": 0.9381078273057938,
+      "system_centroid_cosine_mean": 0.9406606604655584,
+      "system_centroid_cosine_p05": 0.9168506801128388
+    },
+    "inflect-nano-v2-finalist::homographs": {
+      "count": 6,
+      "reference_cosine_mean": 0.908358633518219,
+      "reference_cosine_median": 0.914387047290802,
+      "reference_cosine_p05": 0.8755304366350174,
+      "reference_cosine_p95": 0.9242926687002182,
+      "system_centroid_cosine_mean": 0.9307341972986857,
+      "system_centroid_cosine_p05": 0.8999106287956238
+    },
+    "inflect-nano-v2-finalist::long": {
+      "count": 15,
+      "reference_cosine_mean": 0.9382105032602946,
+      "reference_cosine_median": 0.9384850263595581,
+      "reference_cosine_p05": 0.9240471005439759,
+      "reference_cosine_p95": 0.9524985492229462,
+      "system_centroid_cosine_mean": 0.9689757903416951,
+      "system_centroid_cosine_p05": 0.9598417639732361
+    },
+    "inflect-nano-v2-finalist::long_range": {
+      "count": 6,
+      "reference_cosine_mean": 0.9405539035797119,
+      "reference_cosine_median": 0.9379453659057617,
+      "reference_cosine_p05": 0.9266068637371063,
+      "reference_cosine_p95": 0.9597494304180145,
+      "system_centroid_cosine_mean": 0.9636660019556681,
+      "system_centroid_cosine_p05": 0.9552741944789886
+    },
+    "inflect-nano-v2-finalist::names_places": {
+      "count": 6,
+      "reference_cosine_mean": 0.9113012353579203,
+      "reference_cosine_median": 0.9119357764720917,
+      "reference_cosine_p05": 0.8682866543531418,
+      "reference_cosine_p95": 0.9430657923221588,
+      "system_centroid_cosine_mean": 0.9352895220120748,
+      "system_centroid_cosine_p05": 0.903939962387085
+    },
+    "inflect-nano-v2-finalist::narrative": {
+      "count": 24,
+      "reference_cosine_mean": 0.9179906671245893,
+      "reference_cosine_median": 0.9208799004554749,
+      "reference_cosine_p05": 0.87001873254776,
+      "reference_cosine_p95": 0.9442918866872787,
+      "system_centroid_cosine_mean": 0.9444205164909363,
+      "system_centroid_cosine_p05": 0.917294853925705
+    },
+    "inflect-nano-v2-finalist::numbers": {
+      "count": 6,
+      "reference_cosine_mean": 0.9285498062769572,
+      "reference_cosine_median": 0.9288731813430786,
+      "reference_cosine_p05": 0.9102296233177185,
+      "reference_cosine_p95": 0.9489256143569946,
+      "system_centroid_cosine_mean": 0.9480861326058706,
+      "system_centroid_cosine_p05": 0.9378992468118668
+    },
+    "inflect-nano-v2-finalist::punctuation": {
+      "count": 6,
+      "reference_cosine_mean": 0.9254332383473715,
+      "reference_cosine_median": 0.9220907688140869,
+      "reference_cosine_p05": 0.9143942892551422,
+      "reference_cosine_p95": 0.9415996968746185,
+      "system_centroid_cosine_mean": 0.9461007714271545,
+      "system_centroid_cosine_p05": 0.937388151884079
+    },
+    "inflect-nano-v2-finalist::question": {
+      "count": 15,
+      "reference_cosine_mean": 0.8863190054893494,
+      "reference_cosine_median": 0.8815686702728271,
+      "reference_cosine_p05": 0.8477757394313812,
+      "reference_cosine_p95": 0.9280099689960479,
+      "system_centroid_cosine_mean": 0.9105314493179322,
+      "system_centroid_cosine_p05": 0.8822363495826722
+    },
+    "inflect-nano-v2-finalist::rare_words": {
+      "count": 6,
+      "reference_cosine_mean": 0.9354390203952789,
+      "reference_cosine_median": 0.9387620389461517,
+      "reference_cosine_p05": 0.9184393882751465,
+      "reference_cosine_p95": 0.9469597637653351,
+      "system_centroid_cosine_mean": 0.9583907326062521,
+      "system_centroid_cosine_p05": 0.954312339425087
+    },
+    "inflect-nano-v2-finalist::short": {
+      "count": 24,
+      "reference_cosine_mean": 0.7531019461651643,
+      "reference_cosine_median": 0.7750166356563568,
+      "reference_cosine_p05": 0.5188983529806137,
+      "reference_cosine_p95": 0.8795914500951767,
+      "system_centroid_cosine_mean": 0.7857235992948214,
+      "system_centroid_cosine_p05": 0.5674794971942901
+    },
+    "inflect-nano-v2-finalist::technical": {
+      "count": 21,
+      "reference_cosine_mean": 0.9235674398286002,
+      "reference_cosine_median": 0.928881049156189,
+      "reference_cosine_p05": 0.8994393348693848,
+      "reference_cosine_p95": 0.9459622502326965,
+      "system_centroid_cosine_mean": 0.9442476261229742,
+      "system_centroid_cosine_p05": 0.9218538403511047
+    }
+  }
+}

inference.py

@@ -0,0 +1,170 @@
+from __future__ import annotations
+
+import argparse
+import re
+import sys
+from pathlib import Path
+
+import numpy as np
+import soundfile as sf
+import torch
+
+
+PACKAGE_ROOT = Path(__file__).resolve().parent
+RUNTIME_ROOT = PACKAGE_ROOT / "runtime"
+sys.path.insert(0, str(RUNTIME_ROOT))
+sys.path.insert(0, str(PACKAGE_ROOT))
+
+import commons  # noqa: E402
+import utils  # noqa: E402
+from inflect_vits_frontend import run_vits_frontend  # noqa: E402
+from models import SynthesizerTrn  # noqa: E402
+from text import cleaned_text_to_sequence  # noqa: E402
+from text.symbols import symbols  # noqa: E402
+
+
+def split_text(text: str, limit: int = 280) -> list[str]:
+    normalized = " ".join(text.split())
+    sentences = [
+        part.strip()
+        for part in re.split(r"(?<=[.!?;:])\s+", normalized)
+        if part.strip()
+    ]
+    chunks: list[str] = []
+    for sentence in sentences or [normalized]:
+        while len(sentence) > limit:
+            search = sentence[: limit + 1]
+            punctuation = max(search.rfind(mark) for mark in (",", ";", ":"))
+            split_at = (
+                punctuation + 1
+                if punctuation >= limit // 2
+                else sentence.rfind(" ", 0, limit + 1)
+            )
+            if split_at < limit // 2:
+                split_at = limit
+            chunks.append(sentence[:split_at].strip())
+            sentence = sentence[split_at:].strip()
+        if sentence:
+            chunks.append(sentence)
+    return chunks
+
+
+def boundary_pause_seconds(chunk: str) -> float:
+    ending = chunk.rstrip()[-1:] if chunk.strip() else ""
+    return {
+        "?": 0.28,
+        "!": 0.24,
+        ".": 0.22,
+        ";": 0.16,
+        ":": 0.13,
+        ",": 0.09,
+    }.get(ending, 0.08)
+
+
+def edge_fade(waveform: np.ndarray, sample_rate: int, milliseconds: float = 5.0) -> np.ndarray:
+    frames = min(round(sample_rate * milliseconds / 1000.0), waveform.size // 2)
+    if frames <= 0:
+        return waveform
+    output = waveform.copy()
+    ramp = np.linspace(0.0, 1.0, frames, endpoint=True, dtype=np.float32)
+    output[:frames] *= ramp
+    output[-frames:] *= ramp[::-1]
+    return output
+
+
+class InflectTTS:
+    def __init__(self, model_dir: str | Path = PACKAGE_ROOT, device: str = "cpu") -> None:
+        self.root = Path(model_dir).resolve()
+        self.device = torch.device(device)
+        self.hps = utils.get_hparams_from_file(str(self.root / "config.json"))
+        self.model = SynthesizerTrn(
+            len(symbols),
+            self.hps.data.filter_length // 2 + 1,
+            self.hps.train.segment_size // self.hps.data.hop_length,
+            **self.hps.model,
+        ).to(self.device).eval()
+        utils.load_checkpoint(str(self.root / "model.pth"), self.model, None)
+        self.sample_rate = int(self.hps.data.sampling_rate)
+
+    def _tokens(self, text: str) -> tuple[torch.Tensor, torch.Tensor]:
+        phonemes = run_vits_frontend(text).phoneme_text
+        sequence = cleaned_text_to_sequence(phonemes)
+        if self.hps.data.add_blank:
+            sequence = commons.intersperse(sequence, 0)
+        if not sequence:
+            raise ValueError("The text frontend produced no speakable tokens.")
+        tokens = torch.LongTensor(sequence).to(self.device).unsqueeze(0)
+        lengths = torch.LongTensor([tokens.size(1)]).to(self.device)
+        return tokens, lengths
+
+    @torch.inference_mode()
+    def synthesize(
+        self,
+        text: str,
+        *,
+        speed: float = 1.0,
+        variation: float = 0.667,
+        seed: int = 0,
+    ) -> np.ndarray:
+        normalized = " ".join(text.split())
+        if not normalized:
+            raise ValueError("Text must not be empty.")
+        if not 0.5 <= speed <= 2.0:
+            raise ValueError("speed must be between 0.5 and 2.0")
+        if not 0.0 <= variation <= 1.0:
+            raise ValueError("variation must be between 0.0 and 1.0")
+        chunks = split_text(normalized)
+        pieces: list[np.ndarray] = []
+        for index, chunk in enumerate(chunks):
+            if index:
+                pieces.append(
+                    np.zeros(
+                        round(self.sample_rate * boundary_pause_seconds(chunks[index - 1])),
+                        dtype=np.float32,
+                    )
+                )
+            tokens, lengths = self._tokens(chunk)
+            torch.manual_seed(seed + index)
+            if self.device.type == "cuda":
+                torch.cuda.manual_seed_all(seed + index)
+            waveform = self.model.infer(
+                tokens,
+                lengths,
+                noise_scale=variation,
+                noise_scale_w=0.8,
+                length_scale=1.0 / speed,
+                max_len=4000,
+            )[0][0, 0].float().cpu().numpy()
+            pieces.append(edge_fade(waveform, self.sample_rate))
+        return np.clip(np.concatenate(pieces), -1.0, 1.0)
+
+    def save(self, text: str, output: str | Path, **kwargs: object) -> Path:
+        destination = Path(output)
+        destination.parent.mkdir(parents=True, exist_ok=True)
+        sf.write(destination, self.synthesize(text, **kwargs), self.sample_rate)
+        return destination
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description="Run standalone Inflect v2 synthesis.")
+    parser.add_argument("--model-dir", type=Path, default=PACKAGE_ROOT)
+    parser.add_argument("--text", required=True)
+    parser.add_argument("--output", type=Path, required=True)
+    parser.add_argument("--device", default="cpu")
+    parser.add_argument("--speed", type=float, default=1.0)
+    parser.add_argument("--variation", type=float, default=0.667)
+    parser.add_argument("--seed", type=int, default=0)
+    args = parser.parse_args()
+    engine = InflectTTS(args.model_dir, args.device)
+    engine.save(
+        args.text,
+        args.output,
+        speed=args.speed,
+        variation=args.variation,
+        seed=args.seed,
+    )
+    print(f"wrote {args.output} at {engine.sample_rate} Hz")
+
+
+if __name__ == "__main__":
+    main()

inflect_nano_v2_frontend.py

@@ -0,0 +1,341 @@
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import re
+import sys
+from dataclasses import asdict, dataclass
+from datetime import date
+from pathlib import Path
+
+from num2words import num2words
+
+
+MONTHS = [
+    "January",
+    "February",
+    "March",
+    "April",
+    "May",
+    "June",
+    "July",
+    "August",
+    "September",
+    "October",
+    "November",
+    "December",
+]
+
+WORD_OVERRIDES = {
+    "Qwen3": "Qwen three",
+    "Qwen": "Qwen",
+    "PyTorch": "pie torch",
+    "SQLite": "ess cue lite",
+    "USB-C": "you ess bee see",
+    "RTX 3060": "ar tee ex thirty sixty",
+    "RTX 3090": "ar tee ex thirty ninety",
+    "RTX 4090": "ar tee ex forty ninety",
+    "RTX 5080": "ar tee ex fifty eighty",
+    "RTX 5090": "ar tee ex fifty ninety",
+}
+
+LETTER_NAMES = {
+    "A": "ay",
+    "B": "bee",
+    "C": "see",
+    "D": "dee",
+    "E": "ee",
+    "F": "eff",
+    "G": "gee",
+    "H": "aitch",
+    "I": "eye",
+    "J": "jay",
+    "K": "kay",
+    "L": "ell",
+    "M": "em",
+    "N": "en",
+    "O": "oh",
+    "P": "pee",
+    "Q": "cue",
+    "R": "ar",
+    "S": "ess",
+    "T": "tee",
+    "U": "you",
+    "V": "vee",
+    "W": "double you",
+    "X": "ex",
+    "Y": "why",
+    "Z": "zee",
+}
+
+ABBREVIATIONS = {
+    "Dr.": "doctor",
+    "Mr.": "mister",
+    "Mrs.": "missus",
+    "Ms.": "miss",
+    "Prof.": "professor",
+    "St.": "saint",
+    "vs.": "versus",
+    "etc.": "et cetera",
+    "e.g.": "for example",
+    "i.e.": "that is",
+}
+
+PUNCT_TRANSLATION = str.maketrans(
+    {
+        "\u2018": "'",
+        "\u2019": "'",
+        "\u201c": '"',
+        "\u201d": '"',
+        "\u2013": "-",
+        "\u2014": ", ",
+        "\u2026": "...",
+    }
+)
+
+_ESPEAK_CONFIGURED = False
+_ESPEAK_BACKEND = None
+
+
+@dataclass
+class FrontendOutput:
+    raw_text: str
+    normalized_text: str
+    phoneme_text: str
+    tokens: list[str]
+    token_count: int
+
+
+def _words(value: int | float, *, ordinal: bool = False) -> str:
+    if ordinal:
+        text = num2words(value, to="ordinal")
+    else:
+        text = num2words(value)
+    return text.replace("-", " ").replace(",", "")
+
+
+def _digit_words(text: str) -> str:
+    return " ".join(_words(int(ch)) for ch in text if ch.isdigit())
+
+
+def _expand_money(match: re.Match[str]) -> str:
+    raw = match.group(1).replace(",", "")
+    dollars, _, cents = raw.partition(".")
+    dollar_count = int(dollars)
+    parts = [_words(dollar_count), "dollar" if dollar_count == 1 else "dollars"]
+    if cents:
+        cents = cents[:2].ljust(2, "0")
+        cent_count = int(cents)
+        if cent_count:
+            parts.extend(["and", _words(cent_count), "cent" if cent_count == 1 else "cents"])
+    return " ".join(parts)
+
+
+def _expand_date_slash(match: re.Match[str]) -> str:
+    month = int(match.group(1))
+    day = int(match.group(2))
+    year = int(match.group(3))
+    try:
+        date(year, month, day)
+    except ValueError:
+        return match.group(0)
+    return f"{MONTHS[month - 1]} {_words(day, ordinal=True)} {_words(year)}"
+
+
+def _expand_time(match: re.Match[str]) -> str:
+    hour = int(match.group(1))
+    minute = int(match.group(2))
+    suffix = match.group(3) or ""
+    pieces = [_words(hour)]
+    if minute == 0:
+        pieces.append("o clock")
+    elif minute < 10:
+        pieces.extend(["oh", _words(minute)])
+    else:
+        pieces.append(_words(minute))
+    if suffix:
+        suffix = suffix.lower().replace(".", "")
+        pieces.extend(list(suffix))
+    return " ".join(pieces)
+
+
+def _expand_bare_hour_time(match: re.Match[str]) -> str:
+    hour = int(match.group(1))
+    suffix = re.sub(r"[^A-Za-z]", "", match.group(2)).lower()
+    return f"{_words(hour)} {' '.join(suffix)}"
+
+
+def _expand_version(match: re.Match[str]) -> str:
+    return " point ".join(_words(int(part)) for part in match.group(0).split("."))
+
+
+def _expand_decimal(match: re.Match[str]) -> str:
+    whole, frac = match.group(1), match.group(2)
+    return f"{_words(int(whole))} point {_digit_words(frac)}"
+
+
+def _expand_ordinal(match: re.Match[str]) -> str:
+    return _words(int(match.group(1)), ordinal=True)
+
+
+def _expand_number(match: re.Match[str]) -> str:
+    value = match.group(0).replace(",", "")
+    if len(value) >= 5 and not value.startswith("20"):
+        return _digit_words(value)
+    return _words(int(value))
+
+
+def _expand_phone(match: re.Match[str]) -> str:
+    left, right = match.group(1), match.group(2)
+    return f"{_digit_words(left)}, {_digit_words(right)}"
+
+
+def _expand_acronym(match: re.Match[str]) -> str:
+    acronym = match.group(0)
+    if len(acronym) <= 1:
+        return acronym
+    return " ".join(LETTER_NAMES.get(ch, ch) for ch in acronym)
+
+
+def normalize_text(text: str) -> str:
+    text = text.translate(PUNCT_TRANSLATION)
+    text = re.sub(r"\s+", " ", text).strip()
+
+    for src, dst in WORD_OVERRIDES.items():
+        text = re.sub(rf"\b{re.escape(src)}\b", dst, text)
+    for src, dst in ABBREVIATIONS.items():
+        text = re.sub(rf"\b{re.escape(src)}", dst, text, flags=re.IGNORECASE)
+
+    text = re.sub(r"\b([A-Z])(?:\.([A-Z]))+\.", lambda m: " ".join(re.findall(r"[A-Z]", m.group(0))), text)
+    text = re.sub(r"\$(\d[\d,]*(?:\.\d{1,2})?)", _expand_money, text)
+    text = re.sub(r"\b(0?[1-9]|1[0-2])/(0?[1-9]|[12]\d|3[01])/(20\d{2}|19\d{2})\b", _expand_date_slash, text)
+    text = re.sub(r"\b(\d{1,2}):(\d{2})\s*([AaPp]\.?\s*[Mm]\.?)?\b", _expand_time, text)
+    text = re.sub(r"\b(\d{1,2})\s*([AaPp]\.?\s*[Mm]\.?)\b", _expand_bare_hour_time, text)
+    text = re.sub(r"\b(\d{3})-(\d{4})\b", _expand_phone, text)
+    text = re.sub(r"\b\d+(?:\.\d+){2,}\b", _expand_version, text)
+    text = re.sub(r"\b(\d+)\.(\d+)\b", _expand_decimal, text)
+    text = re.sub(r"\b(\d+)(st|nd|rd|th)\b", _expand_ordinal, text, flags=re.IGNORECASE)
+    text = re.sub(r"\b\d[\d,]*\b", _expand_number, text)
+    text = re.sub(r"\b[A-Z]{2,}\b", _expand_acronym, text)
+    text = re.sub(r"\s+([,;:.!?])", r"\1", text)
+    text = re.sub(r"([,;:.!?])(?=\S)", r"\1 ", text)
+    return re.sub(r"\s+", " ", text).strip()
+
+
+def _configure_espeak() -> None:
+    global _ESPEAK_CONFIGURED
+    if _ESPEAK_CONFIGURED:
+        return
+
+    # Prefer the persistent distro library for long Linux preprocessing jobs.
+    # espeakng-loader extracts a temporary shared object, which can exhaust mmap
+    # resources when phonemizer repeatedly creates backends over a large corpus.
+    system_libraries = (
+        Path("/usr/lib/x86_64-linux-gnu/libespeak-ng.so.1"),
+        Path("/usr/lib/aarch64-linux-gnu/libespeak-ng.so.1"),
+        Path("/usr/lib64/libespeak-ng.so.1"),
+    )
+    system_library = next((path for path in system_libraries if path.is_file()), None)
+    if system_library is not None:
+        os.environ.setdefault("PHONEMIZER_ESPEAK_LIBRARY", str(system_library))
+    else:
+        import espeakng_loader
+
+        os.environ.setdefault("PHONEMIZER_ESPEAK_LIBRARY", espeakng_loader.get_library_path())
+        os.environ.setdefault("ESPEAK_DATA_PATH", espeakng_loader.get_data_path())
+        espeakng_loader.make_library_available()
+        espeakng_loader.load_library()
+    _ESPEAK_CONFIGURED = True
+
+
+def phonemize_normalized_text(normalized_text: str) -> str:
+    global _ESPEAK_BACKEND
+    _configure_espeak()
+    from phonemizer.backend import EspeakBackend
+    from phonemizer.separator import Separator
+
+    if _ESPEAK_BACKEND is None:
+        _ESPEAK_BACKEND = EspeakBackend(
+            language="en-us",
+            preserve_punctuation=True,
+            with_stress=True,
+            language_switch="remove-flags",
+        )
+    return _ESPEAK_BACKEND.phonemize(
+        [normalized_text],
+        separator=Separator(phone=" ", word=" | ", syllable=""),
+        strip=True,
+        njobs=1,
+    )[0]
+
+
+def tokenize_phoneme_text(phoneme_text: str) -> list[str]:
+    text = phoneme_text.replace("|", " <word> ")
+    text = re.sub(r"([,;:.!?])", r" \1 ", text)
+    tokens = [tok for tok in re.split(r"\s+", text.strip()) if tok]
+    return tokens
+
+
+def run_frontend(text: str) -> FrontendOutput:
+    normalized = normalize_text(text)
+    phoneme_text = phonemize_normalized_text(normalized)
+    tokens = tokenize_phoneme_text(phoneme_text)
+    return FrontendOutput(
+        raw_text=text,
+        normalized_text=normalized,
+        phoneme_text=phoneme_text,
+        tokens=tokens,
+        token_count=len(tokens),
+    )
+
+
+def _iter_input_rows(path: Path):
+    with path.open("r", encoding="utf-8") as f:
+        for line_number, line in enumerate(f, 1):
+            line = line.strip()
+            if not line:
+                continue
+            if line.startswith("{"):
+                row = json.loads(line)
+                text = row.get("target_text") or row.get("text") or row.get("source_text")
+                if not text:
+                    raise ValueError(f"No text field found at {path}:{line_number}")
+                yield row, str(text)
+            else:
+                yield {"line_number": line_number}, line
+
+
+def main() -> None:
+    if hasattr(sys.stdout, "reconfigure"):
+        sys.stdout.reconfigure(encoding="utf-8", errors="replace")
+    if hasattr(sys.stderr, "reconfigure"):
+        sys.stderr.reconfigure(encoding="utf-8", errors="replace")
+
+    ap = argparse.ArgumentParser(description="Inflect-Nano-v2 English text frontend.")
+    ap.add_argument("--text", help="Single text string to normalize and phonemize.")
+    ap.add_argument("--input", type=Path, help="Text file or JSONL to process.")
+    ap.add_argument("--out", type=Path, help="Output JSONL path for --input.")
+    args = ap.parse_args()
+
+    if bool(args.text) == bool(args.input):
+        raise SystemExit("Provide exactly one of --text or --input.")
+
+    if args.text:
+        print(json.dumps(asdict(run_frontend(args.text)), ensure_ascii=False, indent=2))
+        return
+
+    if not args.out:
+        raise SystemExit("--out is required with --input.")
+
+    args.out.parent.mkdir(parents=True, exist_ok=True)
+    with args.out.open("w", encoding="utf-8") as f:
+        for metadata, text in _iter_input_rows(args.input):
+            result = asdict(run_frontend(text))
+            result["metadata"] = metadata
+            f.write(json.dumps(result, ensure_ascii=False) + "\n")
+    print(f"Wrote {args.out}")
+
+
+if __name__ == "__main__":
+    main()

inflect_vits_frontend.py

@@ -0,0 +1,67 @@
+from __future__ import annotations
+
+import re
+from dataclasses import dataclass
+
+from inflect_nano_v2_frontend import _configure_espeak, normalize_text
+
+
+# eSpeak is the general fallback. This table contains verified exceptions only;
+# every entry is covered by a regression test and listening audit.
+PHONEME_OVERRIDES = {
+    "sˈæskɐtʃˌuːən": "sɐskˈætʃəwən",
+    "flʊɹɹˈɛsənt": "flʊˈɹɛsənt",
+}
+
+
+@dataclass(frozen=True)
+class VitsFrontendOutput:
+    raw_text: str
+    normalized_text: str
+    phoneme_text: str
+
+
+def phonemize_normalized(normalized_text: str) -> str:
+    return phonemize_normalized_batch([normalized_text], jobs=1)[0]
+
+
+def _apply_phoneme_overrides(phoneme_text: str) -> str:
+    for source, replacement in PHONEME_OVERRIDES.items():
+        phoneme_text = phoneme_text.replace(source, replacement)
+    return re.sub(r"\s+", " ", phoneme_text).strip()
+
+
+def phonemize_normalized_batch(normalized_texts: list[str], *, jobs: int = 1) -> list[str]:
+    if not normalized_texts:
+        return []
+    _configure_espeak()
+    from phonemizer import phonemize
+
+    phoneme_texts = phonemize(
+        normalized_texts,
+        language="en-us",
+        backend="espeak",
+        strip=True,
+        preserve_punctuation=True,
+        with_stress=True,
+        njobs=jobs,
+    )
+    return [_apply_phoneme_overrides(text) for text in phoneme_texts]
+
+
+def run_vits_frontend_batch(texts: list[str], *, jobs: int = 1) -> list[VitsFrontendOutput]:
+    normalized = [normalize_text(text) for text in texts]
+    phonemes = phonemize_normalized_batch(normalized, jobs=jobs)
+    return [
+        VitsFrontendOutput(raw_text=raw, normalized_text=norm, phoneme_text=phones)
+        for raw, norm, phones in zip(texts, normalized, phonemes, strict=True)
+    ]
+
+
+def run_vits_frontend(text: str) -> VitsFrontendOutput:
+    normalized = normalize_text(text)
+    return VitsFrontendOutput(
+        raw_text=text,
+        normalized_text=normalized,
+        phoneme_text=phonemize_normalized(normalized),
+    )

model.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3eede065c9ccfa88ade0a5a9a5c23de34afcbbb32213e59aad44d5cf100fdee8
+size 37529995

model.pth.json

@@ -0,0 +1,10 @@
+{
+  "format": "inflect_vits_inference_checkpoint_v1",
+  "input": "vits/logs/inflect_micro_v2_release_decoder_polish_v1/G_186000.pth",
+  "output": "release_polish_overnight_20260720/release_candidates/Inflect-Micro-v2-male/model.pth",
+  "deployable_parameters": 9356513,
+  "tensor_count": 410,
+  "removed_tensor_count": 100,
+  "output_bytes": 37529995,
+  "sha256": "3eede065c9ccfa88ade0a5a9a5c23de34afcbbb32213e59aad44d5cf100fdee8"
+}

release_manifest.json

@@ -0,0 +1,149 @@
+{
+  "family": "micro",
+  "source_label": "micro-186k",
+  "source_checkpoint": "vits/logs/inflect_micro_v2_release_decoder_polish_v1/G_186000.pth",
+  "source_config": "vits/configs/runtime_inflect_micro_v2_release_decoder_polish_v1.json",
+  "deployable_parameters": 9356513,
+  "parameter_limit_exclusive": 10000000,
+  "model_bytes": 37529995,
+  "sample_rate": 24000,
+  "smoke_audio_seconds": 3.968,
+  "standalone_cpu_smoke_audio_seconds": 3.616,
+  "files": [
+    {
+      "path": "LICENSE",
+      "bytes": 10455,
+      "sha256": "12c48c640d1d0baf158d45175a4238b9227c71c77f50b0d15791b8efc08cd793"
+    },
+    {
+      "path": "THIRD_PARTY_NOTICES.md",
+      "bytes": 892,
+      "sha256": "ef5343cd39713d313aefcfd09bb55fd6606ae2dacabeee13e0abdaa2e5b62589"
+    },
+    {
+      "path": "config.json",
+      "bytes": 1992,
+      "sha256": "627f66cf339155d4b8524ce78b65230d2e2f33850f6e3f87f2e46c1add7e27d7"
+    },
+    {
+      "path": "inference.py",
+      "bytes": 6019,
+      "sha256": "91943b432ab3fd3568658eff8f51301409f26f147041e46915db0b350e0efd98"
+    },
+    {
+      "path": "inflect_nano_v2_frontend.py",
+      "bytes": 10246,
+      "sha256": "9f800c7d86939b6eadbdec723d2411caaab820035f87de72ca8bece1a0ab3312"
+    },
+    {
+      "path": "inflect_vits_frontend.py",
+      "bytes": 2031,
+      "sha256": "70eb4cdb96b2752015f72be984f5e62e0f26d7dc80e708875601702057b95cc2"
+    },
+    {
+      "path": "model.pth",
+      "bytes": 37529995,
+      "sha256": "3eede065c9ccfa88ade0a5a9a5c23de34afcbbb32213e59aad44d5cf100fdee8"
+    },
+    {
+      "path": "model.pth.json",
+      "bytes": 443,
+      "sha256": "4c3d2d7c361792e3fe398745989dbf166dde40193122afdfbe8d2b1538b307ae"
+    },
+    {
+      "path": "requirements.txt",
+      "bytes": 128,
+      "sha256": "51778f4f55911af92747707c7d1e61191f7af5381efea9b71cf5a1abb2efcb03"
+    },
+    {
+      "path": "runtime/attentions.py",
+      "bytes": 11780,
+      "sha256": "dc65085d0e5d67da9f7603c98efad1e27eaf42c572f632a58080a5cc0f577f92"
+    },
+    {
+      "path": "runtime/commons.py",
+      "bytes": 4778,
+      "sha256": "f223dc1df791d2fb2953cfebd7014a7bc2a6ed614c05de6c17c7b44718a42cdd"
+    },
+    {
+      "path": "runtime/inflect_alias_free.py",
+      "bytes": 4670,
+      "sha256": "d2144153504a8986b737f2043468bf8bc73027ae8158cd1d8ca58ef54bf82ea7"
+    },
+    {
+      "path": "runtime/models.py",
+      "bytes": 21110,
+      "sha256": "2ae0d939079e8072d4bfa438f2e02f81503ab70665b3027af4f57324c6e65206"
+    },
+    {
+      "path": "runtime/modules.py",
+      "bytes": 13166,
+      "sha256": "f6514eaf7fab47d866ee0f9da94ca6ab6e45ebaf246151408da20dd14f242a75"
+    },
+    {
+      "path": "runtime/monotonic_align.py",
+      "bytes": 206,
+      "sha256": "16ca4890f3e0107250b4a522c0127779994a5a2b99fc35d6a27f93496a72153a"
+    },
+    {
+      "path": "runtime/text/LICENSE",
+      "bytes": 1053,
+      "sha256": "f89accd9d66f6a011f19997d7a21cf71a35ffe2e2d0911b5460cbd0cab52a581"
+    },
+    {
+      "path": "runtime/text/__init__.py",
+      "bytes": 1572,
+      "sha256": "7f88b1aae463edcb42e1dce6f5c4227b167b3f5a40853bc644c73db015c1e471"
+    },
+    {
+      "path": "runtime/text/cleaners.py",
+      "bytes": 2880,
+      "sha256": "11dc7325547290e529c597e620f3aaf8d0bd8a2a2987d48eb42d2289534e42d1"
+    },
+    {
+      "path": "runtime/text/symbols.py",
+      "bytes": 642,
+      "sha256": "e8871acbc37bd79f61ed063a2b3aa20e20555ddec15c6b9f29e1ab05e7f8b776"
+    },
+    {
+      "path": "runtime/transforms.py",
+      "bytes": 8490,
+      "sha256": "9d95e518610c35581c4f1c3b4e8eeca91b7c3d6899e7f0ebb86a0387394d3046"
+    },
+    {
+      "path": "runtime/utils.py",
+      "bytes": 7526,
+      "sha256": "db11ba9b5ffd342abffc59229de7575a7d9e23468f91c6398e70eb4e01636716"
+    },
+    {
+      "path": "smoke/manifest.jsonl",
+      "bytes": 275,
+      "sha256": "7f41bbcd0a8c1cb302b7309a51847b92794d3fb1a5d5e055eb68c535b7a66bb3"
+    },
+    {
+      "path": "smoke/release_smoke.wav",
+      "bytes": 190508,
+      "sha256": "31be94d7b74095b22b68dd5750596e23406e9a85d51bb9fbc49969bfa68abdbf"
+    },
+    {
+      "path": "smoke/standalone_cpu.wav",
+      "bytes": 173612,
+      "sha256": "912f69948b48abdc1227131b8fc07e1fce4ab7b6046016c9e9b63c4b729a5152"
+    },
+    {
+      "path": "smoke_prompt.jsonl",
+      "bytes": 127,
+      "sha256": "a5aad12c2a13cca7807e6e0f9df254cb57183954b0b23c64f94d8cc732c06e0e"
+    },
+    {
+      "path": "third_party/BIGVGAN_LICENSE.txt",
+      "bytes": 1097,
+      "sha256": "330cef542a6aa375ae13de52ab691f5cf8265bcfeee5d37fff67c291daef1dcf"
+    },
+    {
+      "path": "third_party/VITS_LICENSE.txt",
+      "bytes": 1090,
+      "sha256": "1cdf858b4d555567398b378b4a0166e97f7908c56607101b8bf9bb6754cdff3f"
+    }
+  ]
+}

requirements.txt

@@ -0,0 +1,8 @@
+torch>=2.6
+numpy>=1.26,<3
+scipy>=1.13
+soundfile>=0.13
+phonemizer>=3.3
+espeakng-loader>=0.2.4
+num2words>=0.5.14
+Unidecode>=1.3.8

runtime/attentions.py

@@ -0,0 +1,303 @@
+import copy
+import math
+import numpy as np
+import torch
+from torch import nn
+from torch.nn import functional as F
+
+import commons
+import modules
+from modules import LayerNorm
+   
+
+class Encoder(nn.Module):
+  def __init__(self, hidden_channels, filter_channels, n_heads, n_layers, kernel_size=1, p_dropout=0., window_size=4, **kwargs):
+    super().__init__()
+    self.hidden_channels = hidden_channels
+    self.filter_channels = filter_channels
+    self.n_heads = n_heads
+    self.n_layers = n_layers
+    self.kernel_size = kernel_size
+    self.p_dropout = p_dropout
+    self.window_size = window_size
+
+    self.drop = nn.Dropout(p_dropout)
+    self.attn_layers = nn.ModuleList()
+    self.norm_layers_1 = nn.ModuleList()
+    self.ffn_layers = nn.ModuleList()
+    self.norm_layers_2 = nn.ModuleList()
+    for i in range(self.n_layers):
+      self.attn_layers.append(MultiHeadAttention(hidden_channels, hidden_channels, n_heads, p_dropout=p_dropout, window_size=window_size))
+      self.norm_layers_1.append(LayerNorm(hidden_channels))
+      self.ffn_layers.append(FFN(hidden_channels, hidden_channels, filter_channels, kernel_size, p_dropout=p_dropout))
+      self.norm_layers_2.append(LayerNorm(hidden_channels))
+
+  def forward(self, x, x_mask):
+    attn_mask = x_mask.unsqueeze(2) * x_mask.unsqueeze(-1)
+    x = x * x_mask
+    for i in range(self.n_layers):
+      y = self.attn_layers[i](x, x, attn_mask)
+      y = self.drop(y)
+      x = self.norm_layers_1[i](x + y)
+
+      y = self.ffn_layers[i](x, x_mask)
+      y = self.drop(y)
+      x = self.norm_layers_2[i](x + y)
+    x = x * x_mask
+    return x
+
+
+class Decoder(nn.Module):
+  def __init__(self, hidden_channels, filter_channels, n_heads, n_layers, kernel_size=1, p_dropout=0., proximal_bias=False, proximal_init=True, **kwargs):
+    super().__init__()
+    self.hidden_channels = hidden_channels
+    self.filter_channels = filter_channels
+    self.n_heads = n_heads
+    self.n_layers = n_layers
+    self.kernel_size = kernel_size
+    self.p_dropout = p_dropout
+    self.proximal_bias = proximal_bias
+    self.proximal_init = proximal_init
+
+    self.drop = nn.Dropout(p_dropout)
+    self.self_attn_layers = nn.ModuleList()
+    self.norm_layers_0 = nn.ModuleList()
+    self.encdec_attn_layers = nn.ModuleList()
+    self.norm_layers_1 = nn.ModuleList()
+    self.ffn_layers = nn.ModuleList()
+    self.norm_layers_2 = nn.ModuleList()
+    for i in range(self.n_layers):
+      self.self_attn_layers.append(MultiHeadAttention(hidden_channels, hidden_channels, n_heads, p_dropout=p_dropout, proximal_bias=proximal_bias, proximal_init=proximal_init))
+      self.norm_layers_0.append(LayerNorm(hidden_channels))
+      self.encdec_attn_layers.append(MultiHeadAttention(hidden_channels, hidden_channels, n_heads, p_dropout=p_dropout))
+      self.norm_layers_1.append(LayerNorm(hidden_channels))
+      self.ffn_layers.append(FFN(hidden_channels, hidden_channels, filter_channels, kernel_size, p_dropout=p_dropout, causal=True))
+      self.norm_layers_2.append(LayerNorm(hidden_channels))
+
+  def forward(self, x, x_mask, h, h_mask):
+    """
+    x: decoder input
+    h: encoder output
+    """
+    self_attn_mask = commons.subsequent_mask(x_mask.size(2)).to(device=x.device, dtype=x.dtype)
+    encdec_attn_mask = h_mask.unsqueeze(2) * x_mask.unsqueeze(-1)
+    x = x * x_mask
+    for i in range(self.n_layers):
+      y = self.self_attn_layers[i](x, x, self_attn_mask)
+      y = self.drop(y)
+      x = self.norm_layers_0[i](x + y)
+
+      y = self.encdec_attn_layers[i](x, h, encdec_attn_mask)
+      y = self.drop(y)
+      x = self.norm_layers_1[i](x + y)
+      
+      y = self.ffn_layers[i](x, x_mask)
+      y = self.drop(y)
+      x = self.norm_layers_2[i](x + y)
+    x = x * x_mask
+    return x
+
+
+class MultiHeadAttention(nn.Module):
+  def __init__(self, channels, out_channels, n_heads, p_dropout=0., window_size=None, heads_share=True, block_length=None, proximal_bias=False, proximal_init=False):
+    super().__init__()
+    assert channels % n_heads == 0
+
+    self.channels = channels
+    self.out_channels = out_channels
+    self.n_heads = n_heads
+    self.p_dropout = p_dropout
+    self.window_size = window_size
+    self.heads_share = heads_share
+    self.block_length = block_length
+    self.proximal_bias = proximal_bias
+    self.proximal_init = proximal_init
+    self.attn = None
+
+    self.k_channels = channels // n_heads
+    self.conv_q = nn.Conv1d(channels, channels, 1)
+    self.conv_k = nn.Conv1d(channels, channels, 1)
+    self.conv_v = nn.Conv1d(channels, channels, 1)
+    self.conv_o = nn.Conv1d(channels, out_channels, 1)
+    self.drop = nn.Dropout(p_dropout)
+
+    if window_size is not None:
+      n_heads_rel = 1 if heads_share else n_heads
+      rel_stddev = self.k_channels**-0.5
+      self.emb_rel_k = nn.Parameter(torch.randn(n_heads_rel, window_size * 2 + 1, self.k_channels) * rel_stddev)
+      self.emb_rel_v = nn.Parameter(torch.randn(n_heads_rel, window_size * 2 + 1, self.k_channels) * rel_stddev)
+
+    nn.init.xavier_uniform_(self.conv_q.weight)
+    nn.init.xavier_uniform_(self.conv_k.weight)
+    nn.init.xavier_uniform_(self.conv_v.weight)
+    if proximal_init:
+      with torch.no_grad():
+        self.conv_k.weight.copy_(self.conv_q.weight)
+        self.conv_k.bias.copy_(self.conv_q.bias)
+      
+  def forward(self, x, c, attn_mask=None):
+    q = self.conv_q(x)
+    k = self.conv_k(c)
+    v = self.conv_v(c)
+    
+    x, self.attn = self.attention(q, k, v, mask=attn_mask)
+
+    x = self.conv_o(x)
+    return x
+
+  def attention(self, query, key, value, mask=None):
+    # reshape [b, d, t] -> [b, n_h, t, d_k]
+    b, d, t_s, t_t = (*key.size(), query.size(2))
+    query = query.view(b, self.n_heads, self.k_channels, t_t).transpose(2, 3)
+    key = key.view(b, self.n_heads, self.k_channels, t_s).transpose(2, 3)
+    value = value.view(b, self.n_heads, self.k_channels, t_s).transpose(2, 3)
+
+    scores = torch.matmul(query / math.sqrt(self.k_channels), key.transpose(-2, -1))
+    if self.window_size is not None:
+      assert t_s == t_t, "Relative attention is only available for self-attention."
+      key_relative_embeddings = self._get_relative_embeddings(self.emb_rel_k, t_s)
+      rel_logits = self._matmul_with_relative_keys(query /math.sqrt(self.k_channels), key_relative_embeddings)
+      scores_local = self._relative_position_to_absolute_position(rel_logits)
+      scores = scores + scores_local
+    if self.proximal_bias:
+      assert t_s == t_t, "Proximal bias is only available for self-attention."
+      scores = scores + self._attention_bias_proximal(t_s).to(device=scores.device, dtype=scores.dtype)
+    if mask is not None:
+      scores = scores.masked_fill(mask == 0, -1e4)
+      if self.block_length is not None:
+        assert t_s == t_t, "Local attention is only available for self-attention."
+        block_mask = torch.ones_like(scores).triu(-self.block_length).tril(self.block_length)
+        scores = scores.masked_fill(block_mask == 0, -1e4)
+    p_attn = F.softmax(scores, dim=-1) # [b, n_h, t_t, t_s]
+    p_attn = self.drop(p_attn)
+    output = torch.matmul(p_attn, value)
+    if self.window_size is not None:
+      relative_weights = self._absolute_position_to_relative_position(p_attn)
+      value_relative_embeddings = self._get_relative_embeddings(self.emb_rel_v, t_s)
+      output = output + self._matmul_with_relative_values(relative_weights, value_relative_embeddings)
+    output = output.transpose(2, 3).contiguous().view(b, d, t_t) # [b, n_h, t_t, d_k] -> [b, d, t_t]
+    return output, p_attn
+
+  def _matmul_with_relative_values(self, x, y):
+    """
+    x: [b, h, l, m]
+    y: [h or 1, m, d]
+    ret: [b, h, l, d]
+    """
+    ret = torch.matmul(x, y.unsqueeze(0))
+    return ret
+
+  def _matmul_with_relative_keys(self, x, y):
+    """
+    x: [b, h, l, d]
+    y: [h or 1, m, d]
+    ret: [b, h, l, m]
+    """
+    ret = torch.matmul(x, y.unsqueeze(0).transpose(-2, -1))
+    return ret
+
+  def _get_relative_embeddings(self, relative_embeddings, length):
+    max_relative_position = 2 * self.window_size + 1
+    # Pad first before slice to avoid using cond ops.
+    pad_length = max(length - (self.window_size + 1), 0)
+    slice_start_position = max((self.window_size + 1) - length, 0)
+    slice_end_position = slice_start_position + 2 * length - 1
+    if pad_length > 0:
+      padded_relative_embeddings = F.pad(
+          relative_embeddings,
+          commons.convert_pad_shape([[0, 0], [pad_length, pad_length], [0, 0]]))
+    else:
+      padded_relative_embeddings = relative_embeddings
+    used_relative_embeddings = padded_relative_embeddings[:,slice_start_position:slice_end_position]
+    return used_relative_embeddings
+
+  def _relative_position_to_absolute_position(self, x):
+    """
+    x: [b, h, l, 2*l-1]
+    ret: [b, h, l, l]
+    """
+    batch, heads, length, _ = x.size()
+    # Concat columns of pad to shift from relative to absolute indexing.
+    x = F.pad(x, commons.convert_pad_shape([[0,0],[0,0],[0,0],[0,1]]))
+
+    # Concat extra elements so to add up to shape (len+1, 2*len-1).
+    x_flat = x.view([batch, heads, length * 2 * length])
+    x_flat = F.pad(x_flat, commons.convert_pad_shape([[0,0],[0,0],[0,length-1]]))
+
+    # Reshape and slice out the padded elements.
+    x_final = x_flat.view([batch, heads, length+1, 2*length-1])[:, :, :length, length-1:]
+    return x_final
+
+  def _absolute_position_to_relative_position(self, x):
+    """
+    x: [b, h, l, l]
+    ret: [b, h, l, 2*l-1]
+    """
+    batch, heads, length, _ = x.size()
+    # padd along column
+    x = F.pad(x, commons.convert_pad_shape([[0, 0], [0, 0], [0, 0], [0, length-1]]))
+    x_flat = x.view([batch, heads, length**2 + length*(length -1)])
+    # add 0's in the beginning that will skew the elements after reshape
+    x_flat = F.pad(x_flat, commons.convert_pad_shape([[0, 0], [0, 0], [length, 0]]))
+    x_final = x_flat.view([batch, heads, length, 2*length])[:,:,:,1:]
+    return x_final
+
+  def _attention_bias_proximal(self, length):
+    """Bias for self-attention to encourage attention to close positions.
+    Args:
+      length: an integer scalar.
+    Returns:
+      a Tensor with shape [1, 1, length, length]
+    """
+    r = torch.arange(length, dtype=torch.float32)
+    diff = torch.unsqueeze(r, 0) - torch.unsqueeze(r, 1)
+    return torch.unsqueeze(torch.unsqueeze(-torch.log1p(torch.abs(diff)), 0), 0)
+
+
+class FFN(nn.Module):
+  def __init__(self, in_channels, out_channels, filter_channels, kernel_size, p_dropout=0., activation=None, causal=False):
+    super().__init__()
+    self.in_channels = in_channels
+    self.out_channels = out_channels
+    self.filter_channels = filter_channels
+    self.kernel_size = kernel_size
+    self.p_dropout = p_dropout
+    self.activation = activation
+    self.causal = causal
+
+    if causal:
+      self.padding = self._causal_padding
+    else:
+      self.padding = self._same_padding
+
+    self.conv_1 = nn.Conv1d(in_channels, filter_channels, kernel_size)
+    self.conv_2 = nn.Conv1d(filter_channels, out_channels, kernel_size)
+    self.drop = nn.Dropout(p_dropout)
+
+  def forward(self, x, x_mask):
+    x = self.conv_1(self.padding(x * x_mask))
+    if self.activation == "gelu":
+      x = x * torch.sigmoid(1.702 * x)
+    else:
+      x = torch.relu(x)
+    x = self.drop(x)
+    x = self.conv_2(self.padding(x * x_mask))
+    return x * x_mask
+  
+  def _causal_padding(self, x):
+    if self.kernel_size == 1:
+      return x
+    pad_l = self.kernel_size - 1
+    pad_r = 0
+    padding = [[0, 0], [0, 0], [pad_l, pad_r]]
+    x = F.pad(x, commons.convert_pad_shape(padding))
+    return x
+
+  def _same_padding(self, x):
+    if self.kernel_size == 1:
+      return x
+    pad_l = (self.kernel_size - 1) // 2
+    pad_r = self.kernel_size // 2
+    padding = [[0, 0], [0, 0], [pad_l, pad_r]]
+    x = F.pad(x, commons.convert_pad_shape(padding))
+    return x

runtime/commons.py

@@ -0,0 +1,161 @@
+import math
+import numpy as np
+import torch
+from torch import nn
+from torch.nn import functional as F
+
+
+def init_weights(m, mean=0.0, std=0.01):
+  classname = m.__class__.__name__
+  if classname.find("Conv") != -1:
+    m.weight.data.normal_(mean, std)
+
+
+def get_padding(kernel_size, dilation=1):
+  return int((kernel_size*dilation - dilation)/2)
+
+
+def convert_pad_shape(pad_shape):
+  l = pad_shape[::-1]
+  pad_shape = [item for sublist in l for item in sublist]
+  return pad_shape
+
+
+def intersperse(lst, item):
+  result = [item] * (len(lst) * 2 + 1)
+  result[1::2] = lst
+  return result
+
+
+def kl_divergence(m_p, logs_p, m_q, logs_q):
+  """KL(P||Q)"""
+  kl = (logs_q - logs_p) - 0.5
+  kl += 0.5 * (torch.exp(2. * logs_p) + ((m_p - m_q)**2)) * torch.exp(-2. * logs_q)
+  return kl
+
+
+def rand_gumbel(shape):
+  """Sample from the Gumbel distribution, protect from overflows."""
+  uniform_samples = torch.rand(shape) * 0.99998 + 0.00001
+  return -torch.log(-torch.log(uniform_samples))
+
+
+def rand_gumbel_like(x):
+  g = rand_gumbel(x.size()).to(dtype=x.dtype, device=x.device)
+  return g
+
+
+def slice_segments(x, ids_str, segment_size=4):
+  ret = torch.zeros_like(x[:, :, :segment_size])
+  for i in range(x.size(0)):
+    idx_str = ids_str[i]
+    idx_end = idx_str + segment_size
+    ret[i] = x[i, :, idx_str:idx_end]
+  return ret
+
+
+def rand_slice_segments(x, x_lengths=None, segment_size=4):
+  b, d, t = x.size()
+  if x_lengths is None:
+    x_lengths = t
+  ids_str_max = x_lengths - segment_size + 1
+  ids_str = (torch.rand([b]).to(device=x.device) * ids_str_max).to(dtype=torch.long)
+  ret = slice_segments(x, ids_str, segment_size)
+  return ret, ids_str
+
+
+def get_timing_signal_1d(
+    length, channels, min_timescale=1.0, max_timescale=1.0e4):
+  position = torch.arange(length, dtype=torch.float)
+  num_timescales = channels // 2
+  log_timescale_increment = (
+      math.log(float(max_timescale) / float(min_timescale)) /
+      (num_timescales - 1))
+  inv_timescales = min_timescale * torch.exp(
+      torch.arange(num_timescales, dtype=torch.float) * -log_timescale_increment)
+  scaled_time = position.unsqueeze(0) * inv_timescales.unsqueeze(1)
+  signal = torch.cat([torch.sin(scaled_time), torch.cos(scaled_time)], 0)
+  signal = F.pad(signal, [0, 0, 0, channels % 2])
+  signal = signal.view(1, channels, length)
+  return signal
+
+
+def add_timing_signal_1d(x, min_timescale=1.0, max_timescale=1.0e4):
+  b, channels, length = x.size()
+  signal = get_timing_signal_1d(length, channels, min_timescale, max_timescale)
+  return x + signal.to(dtype=x.dtype, device=x.device)
+
+
+def cat_timing_signal_1d(x, min_timescale=1.0, max_timescale=1.0e4, axis=1):
+  b, channels, length = x.size()
+  signal = get_timing_signal_1d(length, channels, min_timescale, max_timescale)
+  return torch.cat([x, signal.to(dtype=x.dtype, device=x.device)], axis)
+
+
+def subsequent_mask(length):
+  mask = torch.tril(torch.ones(length, length)).unsqueeze(0).unsqueeze(0)
+  return mask
+
+
+@torch.jit.script
+def fused_add_tanh_sigmoid_multiply(input_a, input_b, n_channels):
+  n_channels_int = n_channels[0]
+  in_act = input_a + input_b
+  t_act = torch.tanh(in_act[:, :n_channels_int, :])
+  s_act = torch.sigmoid(in_act[:, n_channels_int:, :])
+  acts = t_act * s_act
+  return acts
+
+
+def convert_pad_shape(pad_shape):
+  l = pad_shape[::-1]
+  pad_shape = [item for sublist in l for item in sublist]
+  return pad_shape
+
+
+def shift_1d(x):
+  x = F.pad(x, convert_pad_shape([[0, 0], [0, 0], [1, 0]]))[:, :, :-1]
+  return x
+
+
+def sequence_mask(length, max_length=None):
+  if max_length is None:
+    max_length = length.max()
+  x = torch.arange(max_length, dtype=length.dtype, device=length.device)
+  return x.unsqueeze(0) < length.unsqueeze(1)
+
+
+def generate_path(duration, mask):
+  """
+  duration: [b, 1, t_x]
+  mask: [b, 1, t_y, t_x]
+  """
+  device = duration.device
+  
+  b, _, t_y, t_x = mask.shape
+  cum_duration = torch.cumsum(duration, -1)
+  
+  cum_duration_flat = cum_duration.view(b * t_x)
+  path = sequence_mask(cum_duration_flat, t_y).to(mask.dtype)
+  path = path.view(b, t_x, t_y)
+  path = path - F.pad(path, convert_pad_shape([[0, 0], [1, 0], [0, 0]]))[:, :-1]
+  path = path.unsqueeze(1).transpose(2,3) * mask
+  return path
+
+
+def clip_grad_value_(parameters, clip_value, norm_type=2):
+  if isinstance(parameters, torch.Tensor):
+    parameters = [parameters]
+  parameters = list(filter(lambda p: p.grad is not None, parameters))
+  norm_type = float(norm_type)
+  if clip_value is not None:
+    clip_value = float(clip_value)
+
+  total_norm = 0
+  for p in parameters:
+    param_norm = p.grad.data.norm(norm_type)
+    total_norm += param_norm.item() ** norm_type
+    if clip_value is not None:
+      p.grad.data.clamp_(min=-clip_value, max=clip_value)
+  total_norm = total_norm ** (1. / norm_type)
+  return total_norm

runtime/inflect_alias_free.py

@@ -0,0 +1,143 @@
+"""Lightweight alias-free waveform blocks derived from NVIDIA BigVGAN.
+
+BigVGAN and alias-free-torch are MIT/Apache-2.0 licensed. The implementation
+is kept local so Inflect can train without BigVGAN's optional CUDA extension.
+"""
+
+import math
+
+import torch
+from torch import nn
+from torch.nn import functional as F
+from torch.nn.utils import remove_weight_norm, weight_norm
+
+from commons import get_padding, init_weights
+
+
+def kaiser_sinc_filter1d(cutoff: float, half_width: float, kernel_size: int):
+  even = kernel_size % 2 == 0
+  half_size = kernel_size // 2
+  delta_f = 4 * half_width
+  attenuation = 2.285 * (half_size - 1) * math.pi * delta_f + 7.95
+  if attenuation > 50.0:
+    beta = 0.1102 * (attenuation - 8.7)
+  elif attenuation >= 21.0:
+    beta = 0.5842 * (attenuation - 21) ** 0.4 + 0.07886 * (attenuation - 21.0)
+  else:
+    beta = 0.0
+  window = torch.kaiser_window(kernel_size, beta=beta, periodic=False)
+  if even:
+    time = torch.arange(-half_size, half_size) + 0.5
+  else:
+    time = torch.arange(kernel_size) - half_size
+  values = 2 * cutoff * window * torch.sinc(2 * cutoff * time)
+  values /= values.sum()
+  return values.view(1, 1, kernel_size)
+
+
+class UpSample1d(nn.Module):
+  def __init__(self, ratio=2, kernel_size=12):
+    super().__init__()
+    self.ratio = ratio
+    self.stride = ratio
+    self.kernel_size = kernel_size
+    self.pad = kernel_size // ratio - 1
+    self.pad_left = self.pad * ratio + (kernel_size - ratio) // 2
+    self.pad_right = self.pad * ratio + (kernel_size - ratio + 1) // 2
+    self.register_buffer(
+        "filter",
+        kaiser_sinc_filter1d(0.5 / ratio, 0.6 / ratio, kernel_size))
+
+  def forward(self, x):
+    channels = x.shape[1]
+    x = F.pad(x, (self.pad, self.pad), mode="replicate")
+    x = self.ratio * F.conv_transpose1d(
+        x, self.filter.expand(channels, -1, -1),
+        stride=self.stride, groups=channels)
+    return x[..., self.pad_left:-self.pad_right]
+
+
+class DownSample1d(nn.Module):
+  def __init__(self, ratio=2, kernel_size=12):
+    super().__init__()
+    self.ratio = ratio
+    self.kernel_size = kernel_size
+    self.pad_left = kernel_size // 2 - int(kernel_size % 2 == 0)
+    self.pad_right = kernel_size // 2
+    self.register_buffer(
+        "filter",
+        kaiser_sinc_filter1d(0.5 / ratio, 0.6 / ratio, kernel_size))
+
+  def forward(self, x):
+    channels = x.shape[1]
+    x = F.pad(x, (self.pad_left, self.pad_right), mode="replicate")
+    return F.conv1d(
+        x, self.filter.expand(channels, -1, -1),
+        stride=self.ratio, groups=channels)
+
+
+class SnakeBeta(nn.Module):
+  def __init__(self, channels: int, logscale: bool = True):
+    super().__init__()
+    initial = torch.zeros(channels) if logscale else torch.ones(channels)
+    self.alpha = nn.Parameter(initial.clone())
+    self.beta = nn.Parameter(initial.clone())
+    self.logscale = logscale
+
+  def forward(self, x):
+    alpha = self.alpha.view(1, -1, 1)
+    beta = self.beta.view(1, -1, 1)
+    if self.logscale:
+      alpha = alpha.exp()
+      beta = beta.exp()
+    return x + torch.sin(x * alpha).square() / (beta + 1e-9)
+
+
+class AliasFreeActivation1d(nn.Module):
+  def __init__(self, activation: nn.Module):
+    super().__init__()
+    self.upsample = UpSample1d()
+    self.act = activation
+    self.downsample = DownSample1d()
+
+  def forward(self, x):
+    return self.downsample(self.act(self.upsample(x)))
+
+
+class AliasFreeResBlock1(nn.Module):
+  """Shape-compatible VITS ResBlock1 with filtered SnakeBeta activations."""
+
+  def __init__(self, channels, kernel_size=3, dilation=(1, 3, 5), logscale=True):
+    super().__init__()
+    self.convs1 = nn.ModuleList([
+        weight_norm(nn.Conv1d(
+            channels, channels, kernel_size, 1,
+            dilation=d, padding=get_padding(kernel_size, d)))
+        for d in dilation
+    ])
+    self.convs2 = nn.ModuleList([
+        weight_norm(nn.Conv1d(
+            channels, channels, kernel_size, 1,
+            dilation=1, padding=get_padding(kernel_size, 1)))
+        for _ in dilation
+    ])
+    self.convs1.apply(init_weights)
+    self.convs2.apply(init_weights)
+    self.activations = nn.ModuleList([
+        AliasFreeActivation1d(SnakeBeta(channels, logscale=logscale))
+        for _ in range(2 * len(dilation))
+    ])
+
+  def forward(self, x, x_mask=None):
+    first = self.activations[::2]
+    second = self.activations[1::2]
+    for conv1, conv2, act1, act2 in zip(self.convs1, self.convs2, first, second):
+      residual = conv2(act2(conv1(act1(x))))
+      x = x + residual
+    return x
+
+  def remove_weight_norm(self):
+    for layer in self.convs1:
+      remove_weight_norm(layer)
+    for layer in self.convs2:
+      remove_weight_norm(layer)

runtime/models.py

@@ -0,0 +1,564 @@
+import copy
+import math
+import torch
+from torch import nn
+from torch.nn import functional as F
+
+import commons
+import modules
+from inflect_alias_free import AliasFreeActivation1d, AliasFreeResBlock1, SnakeBeta
+import attentions
+import monotonic_align
+
+from torch.nn import Conv1d, ConvTranspose1d, AvgPool1d, Conv2d
+from torch.nn.utils import weight_norm, remove_weight_norm, spectral_norm
+from commons import init_weights, get_padding
+
+
+class StochasticDurationPredictor(nn.Module):
+  def __init__(self, in_channels, filter_channels, kernel_size, p_dropout, n_flows=4, gin_channels=0):
+    super().__init__()
+    filter_channels = in_channels # it needs to be removed from future version.
+    self.in_channels = in_channels
+    self.filter_channels = filter_channels
+    self.kernel_size = kernel_size
+    self.p_dropout = p_dropout
+    self.n_flows = n_flows
+    self.gin_channels = gin_channels
+
+    self.log_flow = modules.Log()
+    self.flows = nn.ModuleList()
+    self.flows.append(modules.ElementwiseAffine(2))
+    for i in range(n_flows):
+      self.flows.append(modules.ConvFlow(2, filter_channels, kernel_size, n_layers=3))
+      self.flows.append(modules.Flip())
+
+    self.post_pre = nn.Conv1d(1, filter_channels, 1)
+    self.post_proj = nn.Conv1d(filter_channels, filter_channels, 1)
+    self.post_convs = modules.DDSConv(filter_channels, kernel_size, n_layers=3, p_dropout=p_dropout)
+    self.post_flows = nn.ModuleList()
+    self.post_flows.append(modules.ElementwiseAffine(2))
+    for i in range(4):
+      self.post_flows.append(modules.ConvFlow(2, filter_channels, kernel_size, n_layers=3))
+      self.post_flows.append(modules.Flip())
+
+    self.pre = nn.Conv1d(in_channels, filter_channels, 1)
+    self.proj = nn.Conv1d(filter_channels, filter_channels, 1)
+    self.convs = modules.DDSConv(filter_channels, kernel_size, n_layers=3, p_dropout=p_dropout)
+    if gin_channels != 0:
+      self.cond = nn.Conv1d(gin_channels, filter_channels, 1)
+
+  def forward(self, x, x_mask, w=None, g=None, reverse=False, noise_scale=1.0):
+    x = torch.detach(x)
+    x = self.pre(x)
+    if g is not None:
+      g = torch.detach(g)
+      x = x + self.cond(g)
+    x = self.convs(x, x_mask)
+    x = self.proj(x) * x_mask
+
+    if not reverse:
+      flows = self.flows
+      assert w is not None
+
+      logdet_tot_q = 0 
+      h_w = self.post_pre(w)
+      h_w = self.post_convs(h_w, x_mask)
+      h_w = self.post_proj(h_w) * x_mask
+      e_q = torch.randn(w.size(0), 2, w.size(2)).to(device=x.device, dtype=x.dtype) * x_mask
+      z_q = e_q
+      for flow in self.post_flows:
+        z_q, logdet_q = flow(z_q, x_mask, g=(x + h_w))
+        logdet_tot_q += logdet_q
+      z_u, z1 = torch.split(z_q, [1, 1], 1) 
+      u = torch.sigmoid(z_u) * x_mask
+      z0 = (w - u) * x_mask
+      logdet_tot_q += torch.sum((F.logsigmoid(z_u) + F.logsigmoid(-z_u)) * x_mask, [1,2])
+      logq = torch.sum(-0.5 * (math.log(2*math.pi) + (e_q**2)) * x_mask, [1,2]) - logdet_tot_q
+
+      logdet_tot = 0
+      z0, logdet = self.log_flow(z0, x_mask)
+      logdet_tot += logdet
+      z = torch.cat([z0, z1], 1)
+      for flow in flows:
+        z, logdet = flow(z, x_mask, g=x, reverse=reverse)
+        logdet_tot = logdet_tot + logdet
+      nll = torch.sum(0.5 * (math.log(2*math.pi) + (z**2)) * x_mask, [1,2]) - logdet_tot
+      return nll + logq # [b]
+    else:
+      flows = list(reversed(self.flows))
+      flows = flows[:-2] + [flows[-1]] # remove a useless vflow
+      z = torch.randn(x.size(0), 2, x.size(2)).to(device=x.device, dtype=x.dtype) * noise_scale
+      for flow in flows:
+        z = flow(z, x_mask, g=x, reverse=reverse)
+      z0, z1 = torch.split(z, [1, 1], 1)
+      logw = z0
+      return logw
+
+
+class DurationPredictor(nn.Module):
+  def __init__(self, in_channels, filter_channels, kernel_size, p_dropout, gin_channels=0):
+    super().__init__()
+
+    self.in_channels = in_channels
+    self.filter_channels = filter_channels
+    self.kernel_size = kernel_size
+    self.p_dropout = p_dropout
+    self.gin_channels = gin_channels
+
+    self.drop = nn.Dropout(p_dropout)
+    self.conv_1 = nn.Conv1d(in_channels, filter_channels, kernel_size, padding=kernel_size//2)
+    self.norm_1 = modules.LayerNorm(filter_channels)
+    self.conv_2 = nn.Conv1d(filter_channels, filter_channels, kernel_size, padding=kernel_size//2)
+    self.norm_2 = modules.LayerNorm(filter_channels)
+    self.proj = nn.Conv1d(filter_channels, 1, 1)
+
+    if gin_channels != 0:
+      self.cond = nn.Conv1d(gin_channels, in_channels, 1)
+
+  def forward(self, x, x_mask, g=None):
+    x = torch.detach(x)
+    if g is not None:
+      g = torch.detach(g)
+      x = x + self.cond(g)
+    x = self.conv_1(x * x_mask)
+    x = torch.relu(x)
+    x = self.norm_1(x)
+    x = self.drop(x)
+    x = self.conv_2(x * x_mask)
+    x = torch.relu(x)
+    x = self.norm_2(x)
+    x = self.drop(x)
+    x = self.proj(x * x_mask)
+    return x * x_mask
+
+
+class TextEncoder(nn.Module):
+  def __init__(self,
+      n_vocab,
+      out_channels,
+      hidden_channels,
+      filter_channels,
+      n_heads,
+      n_layers,
+      kernel_size,
+      p_dropout):
+    super().__init__()
+    self.n_vocab = n_vocab
+    self.out_channels = out_channels
+    self.hidden_channels = hidden_channels
+    self.filter_channels = filter_channels
+    self.n_heads = n_heads
+    self.n_layers = n_layers
+    self.kernel_size = kernel_size
+    self.p_dropout = p_dropout
+
+    self.emb = nn.Embedding(n_vocab, hidden_channels)
+    nn.init.normal_(self.emb.weight, 0.0, hidden_channels**-0.5)
+
+    self.encoder = attentions.Encoder(
+      hidden_channels,
+      filter_channels,
+      n_heads,
+      n_layers,
+      kernel_size,
+      p_dropout)
+    self.proj= nn.Conv1d(hidden_channels, out_channels * 2, 1)
+
+  def forward(self, x, x_lengths):
+    x = self.emb(x) * math.sqrt(self.hidden_channels) # [b, t, h]
+    x = torch.transpose(x, 1, -1) # [b, h, t]
+    x_mask = torch.unsqueeze(commons.sequence_mask(x_lengths, x.size(2)), 1).to(x.dtype)
+
+    x = self.encoder(x * x_mask, x_mask)
+    stats = self.proj(x) * x_mask
+
+    m, logs = torch.split(stats, self.out_channels, dim=1)
+    return x, m, logs, x_mask
+
+
+class ResidualCouplingBlock(nn.Module):
+  def __init__(self,
+      channels,
+      hidden_channels,
+      kernel_size,
+      dilation_rate,
+      n_layers,
+      n_flows=4,
+      gin_channels=0):
+    super().__init__()
+    self.channels = channels
+    self.hidden_channels = hidden_channels
+    self.kernel_size = kernel_size
+    self.dilation_rate = dilation_rate
+    self.n_layers = n_layers
+    self.n_flows = n_flows
+    self.gin_channels = gin_channels
+
+    self.flows = nn.ModuleList()
+    for i in range(n_flows):
+      self.flows.append(modules.ResidualCouplingLayer(channels, hidden_channels, kernel_size, dilation_rate, n_layers, gin_channels=gin_channels, mean_only=True))
+      self.flows.append(modules.Flip())
+
+  def forward(self, x, x_mask, g=None, reverse=False):
+    if not reverse:
+      for flow in self.flows:
+        x, _ = flow(x, x_mask, g=g, reverse=reverse)
+    else:
+      for flow in reversed(self.flows):
+        x = flow(x, x_mask, g=g, reverse=reverse)
+    return x
+
+
+class PosteriorEncoder(nn.Module):
+  def __init__(self,
+      in_channels,
+      out_channels,
+      hidden_channels,
+      kernel_size,
+      dilation_rate,
+      n_layers,
+      gin_channels=0):
+    super().__init__()
+    self.in_channels = in_channels
+    self.out_channels = out_channels
+    self.hidden_channels = hidden_channels
+    self.kernel_size = kernel_size
+    self.dilation_rate = dilation_rate
+    self.n_layers = n_layers
+    self.gin_channels = gin_channels
+
+    self.pre = nn.Conv1d(in_channels, hidden_channels, 1)
+    self.enc = modules.WN(hidden_channels, kernel_size, dilation_rate, n_layers, gin_channels=gin_channels)
+    self.proj = nn.Conv1d(hidden_channels, out_channels * 2, 1)
+
+  def forward(self, x, x_lengths, g=None):
+    x_mask = torch.unsqueeze(commons.sequence_mask(x_lengths, x.size(2)), 1).to(x.dtype)
+    x = self.pre(x) * x_mask
+    x = self.enc(x, x_mask, g=g)
+    stats = self.proj(x) * x_mask
+    m, logs = torch.split(stats, self.out_channels, dim=1)
+    z = (m + torch.randn_like(m) * torch.exp(logs)) * x_mask
+    return z, m, logs, x_mask
+
+
+class Generator(torch.nn.Module):
+    def __init__(self, initial_channel, resblock, resblock_kernel_sizes, resblock_dilation_sizes, upsample_rates, upsample_initial_channel, upsample_kernel_sizes, gin_channels=0, decoder_alias_free=False, decoder_alias_free_start_stage=2, decoder_snake_logscale=True):
+        super(Generator, self).__init__()
+        self.num_kernels = len(resblock_kernel_sizes)
+        self.num_upsamples = len(upsample_rates)
+        self.conv_pre = Conv1d(initial_channel, upsample_initial_channel, 7, 1, padding=3)
+        resblock_class = modules.ResBlock1 if resblock == '1' else modules.ResBlock2
+        self.decoder_alias_free = bool(decoder_alias_free)
+        self.decoder_alias_free_start_stage = int(decoder_alias_free_start_stage)
+
+        self.ups = nn.ModuleList()
+        for i, (u, k) in enumerate(zip(upsample_rates, upsample_kernel_sizes)):
+            self.ups.append(weight_norm(
+                ConvTranspose1d(upsample_initial_channel//(2**i), upsample_initial_channel//(2**(i+1)),
+                                k, u, padding=(k-u)//2)))
+
+        self.resblocks = nn.ModuleList()
+        for i in range(len(self.ups)):
+            ch = upsample_initial_channel//(2**(i+1))
+            for j, (k, d) in enumerate(zip(resblock_kernel_sizes, resblock_dilation_sizes)):
+                if self.decoder_alias_free and i >= self.decoder_alias_free_start_stage:
+                    self.resblocks.append(AliasFreeResBlock1(
+                        ch, k, d, logscale=decoder_snake_logscale))
+                else:
+                    self.resblocks.append(resblock_class(ch, k, d))
+
+        self.alias_free_pre_activations = nn.ModuleList()
+        if self.decoder_alias_free:
+            for i in range(self.num_upsamples):
+                channels = upsample_initial_channel // (2 ** i)
+                if i >= self.decoder_alias_free_start_stage:
+                    self.alias_free_pre_activations.append(
+                        AliasFreeActivation1d(nn.LeakyReLU(modules.LRELU_SLOPE)))
+                else:
+                    self.alias_free_pre_activations.append(nn.Identity())
+            self.alias_free_post_activation = AliasFreeActivation1d(
+                SnakeBeta(ch, logscale=decoder_snake_logscale))
+
+        self.conv_post = Conv1d(ch, 1, 7, 1, padding=3, bias=False)
+        self.ups.apply(init_weights)
+
+        if gin_channels != 0:
+            self.cond = nn.Conv1d(gin_channels, upsample_initial_channel, 1)
+
+    def forward(self, x, g=None):
+        x = self.conv_pre(x)
+        if g is not None:
+          x = x + self.cond(g)
+
+        for i in range(self.num_upsamples):
+            if self.decoder_alias_free and i >= self.decoder_alias_free_start_stage:
+                x = self.alias_free_pre_activations[i](x)
+            else:
+                x = F.leaky_relu(x, modules.LRELU_SLOPE)
+            x = self.ups[i](x)
+            xs = None
+            for j in range(self.num_kernels):
+                if xs is None:
+                    xs = self.resblocks[i*self.num_kernels+j](x)
+                else:
+                    xs += self.resblocks[i*self.num_kernels+j](x)
+            x = xs / self.num_kernels
+        if self.decoder_alias_free:
+            x = self.alias_free_post_activation(x)
+        else:
+            x = F.leaky_relu(x)
+        x = self.conv_post(x)
+        x = torch.tanh(x)
+
+        return x
+
+    def remove_weight_norm(self):
+        print('Removing weight norm...')
+        for l in self.ups:
+            remove_weight_norm(l)
+        for l in self.resblocks:
+            l.remove_weight_norm()
+
+
+class DiscriminatorP(torch.nn.Module):
+    def __init__(self, period, kernel_size=5, stride=3, use_spectral_norm=False):
+        super(DiscriminatorP, self).__init__()
+        self.period = period
+        self.use_spectral_norm = use_spectral_norm
+        norm_f = weight_norm if use_spectral_norm == False else spectral_norm
+        self.convs = nn.ModuleList([
+            norm_f(Conv2d(1, 32, (kernel_size, 1), (stride, 1), padding=(get_padding(kernel_size, 1), 0))),
+            norm_f(Conv2d(32, 128, (kernel_size, 1), (stride, 1), padding=(get_padding(kernel_size, 1), 0))),
+            norm_f(Conv2d(128, 512, (kernel_size, 1), (stride, 1), padding=(get_padding(kernel_size, 1), 0))),
+            norm_f(Conv2d(512, 1024, (kernel_size, 1), (stride, 1), padding=(get_padding(kernel_size, 1), 0))),
+            norm_f(Conv2d(1024, 1024, (kernel_size, 1), 1, padding=(get_padding(kernel_size, 1), 0))),
+        ])
+        self.conv_post = norm_f(Conv2d(1024, 1, (3, 1), 1, padding=(1, 0)))
+
+    def forward(self, x):
+        fmap = []
+
+        # 1d to 2d
+        b, c, t = x.shape
+        if t % self.period != 0: # pad first
+            n_pad = self.period - (t % self.period)
+            x = F.pad(x, (0, n_pad), "reflect")
+            t = t + n_pad
+        x = x.view(b, c, t // self.period, self.period)
+
+        for l in self.convs:
+            x = l(x)
+            x = F.leaky_relu(x, modules.LRELU_SLOPE)
+            fmap.append(x)
+        x = self.conv_post(x)
+        fmap.append(x)
+        x = torch.flatten(x, 1, -1)
+
+        return x, fmap
+
+
+class DiscriminatorS(torch.nn.Module):
+    def __init__(self, use_spectral_norm=False):
+        super(DiscriminatorS, self).__init__()
+        norm_f = weight_norm if use_spectral_norm == False else spectral_norm
+        self.convs = nn.ModuleList([
+            norm_f(Conv1d(1, 16, 15, 1, padding=7)),
+            norm_f(Conv1d(16, 64, 41, 4, groups=4, padding=20)),
+            norm_f(Conv1d(64, 256, 41, 4, groups=16, padding=20)),
+            norm_f(Conv1d(256, 1024, 41, 4, groups=64, padding=20)),
+            norm_f(Conv1d(1024, 1024, 41, 4, groups=256, padding=20)),
+            norm_f(Conv1d(1024, 1024, 5, 1, padding=2)),
+        ])
+        self.conv_post = norm_f(Conv1d(1024, 1, 3, 1, padding=1))
+
+    def forward(self, x):
+        fmap = []
+
+        for l in self.convs:
+            x = l(x)
+            x = F.leaky_relu(x, modules.LRELU_SLOPE)
+            fmap.append(x)
+        x = self.conv_post(x)
+        fmap.append(x)
+        x = torch.flatten(x, 1, -1)
+
+        return x, fmap
+
+
+class MultiPeriodDiscriminator(torch.nn.Module):
+    def __init__(self, use_spectral_norm=False):
+        super(MultiPeriodDiscriminator, self).__init__()
+        periods = [2,3,5,7,11]
+
+        discs = [DiscriminatorS(use_spectral_norm=use_spectral_norm)]
+        discs = discs + [DiscriminatorP(i, use_spectral_norm=use_spectral_norm) for i in periods]
+        self.discriminators = nn.ModuleList(discs)
+
+    def forward(self, y, y_hat):
+        y_d_rs = []
+        y_d_gs = []
+        fmap_rs = []
+        fmap_gs = []
+        for i, d in enumerate(self.discriminators):
+            y_d_r, fmap_r = d(y)
+            y_d_g, fmap_g = d(y_hat)
+            y_d_rs.append(y_d_r)
+            y_d_gs.append(y_d_g)
+            fmap_rs.append(fmap_r)
+            fmap_gs.append(fmap_g)
+
+        return y_d_rs, y_d_gs, fmap_rs, fmap_gs
+
+
+
+class SynthesizerTrn(nn.Module):
+  """
+  Synthesizer for Training
+  """
+
+  def __init__(self, 
+    n_vocab,
+    spec_channels,
+    segment_size,
+    inter_channels,
+    hidden_channels,
+    filter_channels,
+    n_heads,
+    n_layers,
+    kernel_size,
+    p_dropout,
+    resblock, 
+    resblock_kernel_sizes, 
+    resblock_dilation_sizes, 
+    upsample_rates, 
+    upsample_initial_channel, 
+    upsample_kernel_sizes,
+    n_speakers=0,
+    gin_channels=0,
+    use_sdp=True,
+    **kwargs):
+
+    super().__init__()
+    self.n_vocab = n_vocab
+    self.spec_channels = spec_channels
+    self.inter_channels = inter_channels
+    self.hidden_channels = hidden_channels
+    self.filter_channels = filter_channels
+    self.n_heads = n_heads
+    self.n_layers = n_layers
+    self.kernel_size = kernel_size
+    self.p_dropout = p_dropout
+    self.resblock = resblock
+    self.resblock_kernel_sizes = resblock_kernel_sizes
+    self.resblock_dilation_sizes = resblock_dilation_sizes
+    self.upsample_rates = upsample_rates
+    self.upsample_initial_channel = upsample_initial_channel
+    self.upsample_kernel_sizes = upsample_kernel_sizes
+    self.segment_size = segment_size
+    self.n_speakers = n_speakers
+    self.gin_channels = gin_channels
+
+    self.use_sdp = use_sdp
+
+    self.enc_p = TextEncoder(n_vocab,
+        inter_channels,
+        hidden_channels,
+        filter_channels,
+        n_heads,
+        n_layers,
+        kernel_size,
+        p_dropout)
+    self.dec = Generator(
+        inter_channels, resblock, resblock_kernel_sizes,
+        resblock_dilation_sizes, upsample_rates, upsample_initial_channel,
+        upsample_kernel_sizes, gin_channels=gin_channels,
+        decoder_alias_free=kwargs.get("decoder_alias_free", False),
+        decoder_alias_free_start_stage=kwargs.get("decoder_alias_free_start_stage", 2),
+        decoder_snake_logscale=kwargs.get("decoder_snake_logscale", True))
+    self.enc_q = PosteriorEncoder(spec_channels, inter_channels, hidden_channels, 5, 1, 16, gin_channels=gin_channels)
+    self.flow = ResidualCouplingBlock(inter_channels, hidden_channels, 5, 1, 4, gin_channels=gin_channels)
+
+    if use_sdp:
+      self.dp = StochasticDurationPredictor(hidden_channels, 192, 3, 0.5, 4, gin_channels=gin_channels)
+    else:
+      self.dp = DurationPredictor(hidden_channels, 256, 3, 0.5, gin_channels=gin_channels)
+
+    if n_speakers > 1:
+      self.emb_g = nn.Embedding(n_speakers, gin_channels)
+
+  def forward(self, x, x_lengths, y, y_lengths, sid=None):
+
+    x, m_p, logs_p, x_mask = self.enc_p(x, x_lengths)
+    if self.n_speakers > 0:
+      g = self.emb_g(sid).unsqueeze(-1) # [b, h, 1]
+    else:
+      g = None
+
+    z, m_q, logs_q, y_mask = self.enc_q(y, y_lengths, g=g)
+    z_p = self.flow(z, y_mask, g=g)
+
+    with torch.no_grad():
+      # negative cross-entropy
+      s_p_sq_r = torch.exp(-2 * logs_p) # [b, d, t]
+      neg_cent1 = torch.sum(-0.5 * math.log(2 * math.pi) - logs_p, [1], keepdim=True) # [b, 1, t_s]
+      neg_cent2 = torch.matmul(-0.5 * (z_p ** 2).transpose(1, 2), s_p_sq_r) # [b, t_t, d] x [b, d, t_s] = [b, t_t, t_s]
+      neg_cent3 = torch.matmul(z_p.transpose(1, 2), (m_p * s_p_sq_r)) # [b, t_t, d] x [b, d, t_s] = [b, t_t, t_s]
+      neg_cent4 = torch.sum(-0.5 * (m_p ** 2) * s_p_sq_r, [1], keepdim=True) # [b, 1, t_s]
+      neg_cent = neg_cent1 + neg_cent2 + neg_cent3 + neg_cent4
+
+      attn_mask = torch.unsqueeze(x_mask, 2) * torch.unsqueeze(y_mask, -1)
+      attn = monotonic_align.maximum_path(neg_cent, attn_mask.squeeze(1)).unsqueeze(1).detach()
+
+    w = attn.sum(2)
+    if self.use_sdp:
+      l_length = self.dp(x, x_mask, w, g=g)
+      l_length = l_length / torch.sum(x_mask)
+    else:
+      logw_ = torch.log(w + 1e-6) * x_mask
+      logw = self.dp(x, x_mask, g=g)
+      l_length = torch.sum((logw - logw_)**2, [1,2]) / torch.sum(x_mask) # for averaging 
+
+    # expand prior
+    m_p = torch.matmul(attn.squeeze(1), m_p.transpose(1, 2)).transpose(1, 2)
+    logs_p = torch.matmul(attn.squeeze(1), logs_p.transpose(1, 2)).transpose(1, 2)
+
+    z_slice, ids_slice = commons.rand_slice_segments(z, y_lengths, self.segment_size)
+    o = self.dec(z_slice, g=g)
+    return o, l_length, attn, ids_slice, x_mask, y_mask, (z, z_p, m_p, logs_p, m_q, logs_q)
+
+  def infer(self, x, x_lengths, sid=None, noise_scale=1, length_scale=1, noise_scale_w=1., max_len=None):
+    x, m_p, logs_p, x_mask = self.enc_p(x, x_lengths)
+    if self.n_speakers > 0:
+      g = self.emb_g(sid).unsqueeze(-1) # [b, h, 1]
+    else:
+      g = None
+
+    if self.use_sdp:
+      logw = self.dp(x, x_mask, g=g, reverse=True, noise_scale=noise_scale_w)
+    else:
+      logw = self.dp(x, x_mask, g=g)
+    w = torch.exp(logw) * x_mask * length_scale
+    w_ceil = torch.ceil(w)
+    y_lengths = torch.clamp_min(torch.sum(w_ceil, [1, 2]), 1).long()
+    y_mask = torch.unsqueeze(commons.sequence_mask(y_lengths, None), 1).to(x_mask.dtype)
+    attn_mask = torch.unsqueeze(x_mask, 2) * torch.unsqueeze(y_mask, -1)
+    attn = commons.generate_path(w_ceil, attn_mask)
+
+    m_p = torch.matmul(attn.squeeze(1), m_p.transpose(1, 2)).transpose(1, 2) # [b, t', t], [b, t, d] -> [b, d, t']
+    logs_p = torch.matmul(attn.squeeze(1), logs_p.transpose(1, 2)).transpose(1, 2) # [b, t', t], [b, t, d] -> [b, d, t']
+
+    z_p = m_p + torch.randn_like(m_p) * torch.exp(logs_p) * noise_scale
+    z = self.flow(z_p, y_mask, g=g, reverse=True)
+    o = self.dec((z * y_mask)[:,:,:max_len], g=g)
+    return o, attn, y_mask, (z, z_p, m_p, logs_p)
+
+  def voice_conversion(self, y, y_lengths, sid_src, sid_tgt):
+    assert self.n_speakers > 0, "n_speakers have to be larger than 0."
+    g_src = self.emb_g(sid_src).unsqueeze(-1)
+    g_tgt = self.emb_g(sid_tgt).unsqueeze(-1)
+    z, m_q, logs_q, y_mask = self.enc_q(y, y_lengths, g=g_src)
+    z_p = self.flow(z, y_mask, g=g_src)
+    z_hat = self.flow(z_p, y_mask, g=g_tgt, reverse=True)
+    o_hat = self.dec(z_hat * y_mask, g=g_tgt)
+    return o_hat, y_mask, (z, z_p, z_hat)

runtime/modules.py

@@ -0,0 +1,390 @@
+import copy
+import math
+import numpy as np
+import scipy
+import torch
+from torch import nn
+from torch.nn import functional as F
+
+from torch.nn import Conv1d, ConvTranspose1d, AvgPool1d, Conv2d
+from torch.nn.utils import weight_norm, remove_weight_norm
+
+import commons
+from commons import init_weights, get_padding
+from transforms import piecewise_rational_quadratic_transform
+
+
+LRELU_SLOPE = 0.1
+
+
+class LayerNorm(nn.Module):
+  def __init__(self, channels, eps=1e-5):
+    super().__init__()
+    self.channels = channels
+    self.eps = eps
+
+    self.gamma = nn.Parameter(torch.ones(channels))
+    self.beta = nn.Parameter(torch.zeros(channels))
+
+  def forward(self, x):
+    x = x.transpose(1, -1)
+    x = F.layer_norm(x, (self.channels,), self.gamma, self.beta, self.eps)
+    return x.transpose(1, -1)
+
+ 
+class ConvReluNorm(nn.Module):
+  def __init__(self, in_channels, hidden_channels, out_channels, kernel_size, n_layers, p_dropout):
+    super().__init__()
+    self.in_channels = in_channels
+    self.hidden_channels = hidden_channels
+    self.out_channels = out_channels
+    self.kernel_size = kernel_size
+    self.n_layers = n_layers
+    self.p_dropout = p_dropout
+    assert n_layers > 1, "Number of layers should be larger than 0."
+
+    self.conv_layers = nn.ModuleList()
+    self.norm_layers = nn.ModuleList()
+    self.conv_layers.append(nn.Conv1d(in_channels, hidden_channels, kernel_size, padding=kernel_size//2))
+    self.norm_layers.append(LayerNorm(hidden_channels))
+    self.relu_drop = nn.Sequential(
+        nn.ReLU(),
+        nn.Dropout(p_dropout))
+    for _ in range(n_layers-1):
+      self.conv_layers.append(nn.Conv1d(hidden_channels, hidden_channels, kernel_size, padding=kernel_size//2))
+      self.norm_layers.append(LayerNorm(hidden_channels))
+    self.proj = nn.Conv1d(hidden_channels, out_channels, 1)
+    self.proj.weight.data.zero_()
+    self.proj.bias.data.zero_()
+
+  def forward(self, x, x_mask):
+    x_org = x
+    for i in range(self.n_layers):
+      x = self.conv_layers[i](x * x_mask)
+      x = self.norm_layers[i](x)
+      x = self.relu_drop(x)
+    x = x_org + self.proj(x)
+    return x * x_mask
+
+
+class DDSConv(nn.Module):
+  """
+  Dialted and Depth-Separable Convolution
+  """
+  def __init__(self, channels, kernel_size, n_layers, p_dropout=0.):
+    super().__init__()
+    self.channels = channels
+    self.kernel_size = kernel_size
+    self.n_layers = n_layers
+    self.p_dropout = p_dropout
+
+    self.drop = nn.Dropout(p_dropout)
+    self.convs_sep = nn.ModuleList()
+    self.convs_1x1 = nn.ModuleList()
+    self.norms_1 = nn.ModuleList()
+    self.norms_2 = nn.ModuleList()
+    for i in range(n_layers):
+      dilation = kernel_size ** i
+      padding = (kernel_size * dilation - dilation) // 2
+      self.convs_sep.append(nn.Conv1d(channels, channels, kernel_size, 
+          groups=channels, dilation=dilation, padding=padding
+      ))
+      self.convs_1x1.append(nn.Conv1d(channels, channels, 1))
+      self.norms_1.append(LayerNorm(channels))
+      self.norms_2.append(LayerNorm(channels))
+
+  def forward(self, x, x_mask, g=None):
+    if g is not None:
+      x = x + g
+    for i in range(self.n_layers):
+      y = self.convs_sep[i](x * x_mask)
+      y = self.norms_1[i](y)
+      y = F.gelu(y)
+      y = self.convs_1x1[i](y)
+      y = self.norms_2[i](y)
+      y = F.gelu(y)
+      y = self.drop(y)
+      x = x + y
+    return x * x_mask
+
+
+class WN(torch.nn.Module):
+  def __init__(self, hidden_channels, kernel_size, dilation_rate, n_layers, gin_channels=0, p_dropout=0):
+    super(WN, self).__init__()
+    assert(kernel_size % 2 == 1)
+    self.hidden_channels =hidden_channels
+    self.kernel_size = kernel_size,
+    self.dilation_rate = dilation_rate
+    self.n_layers = n_layers
+    self.gin_channels = gin_channels
+    self.p_dropout = p_dropout
+
+    self.in_layers = torch.nn.ModuleList()
+    self.res_skip_layers = torch.nn.ModuleList()
+    self.drop = nn.Dropout(p_dropout)
+
+    if gin_channels != 0:
+      cond_layer = torch.nn.Conv1d(gin_channels, 2*hidden_channels*n_layers, 1)
+      self.cond_layer = torch.nn.utils.weight_norm(cond_layer, name='weight')
+
+    for i in range(n_layers):
+      dilation = dilation_rate ** i
+      padding = int((kernel_size * dilation - dilation) / 2)
+      in_layer = torch.nn.Conv1d(hidden_channels, 2*hidden_channels, kernel_size,
+                                 dilation=dilation, padding=padding)
+      in_layer = torch.nn.utils.weight_norm(in_layer, name='weight')
+      self.in_layers.append(in_layer)
+
+      # last one is not necessary
+      if i < n_layers - 1:
+        res_skip_channels = 2 * hidden_channels
+      else:
+        res_skip_channels = hidden_channels
+
+      res_skip_layer = torch.nn.Conv1d(hidden_channels, res_skip_channels, 1)
+      res_skip_layer = torch.nn.utils.weight_norm(res_skip_layer, name='weight')
+      self.res_skip_layers.append(res_skip_layer)
+
+  def forward(self, x, x_mask, g=None, **kwargs):
+    output = torch.zeros_like(x)
+    n_channels_tensor = torch.IntTensor([self.hidden_channels])
+
+    if g is not None:
+      g = self.cond_layer(g)
+
+    for i in range(self.n_layers):
+      x_in = self.in_layers[i](x)
+      if g is not None:
+        cond_offset = i * 2 * self.hidden_channels
+        g_l = g[:,cond_offset:cond_offset+2*self.hidden_channels,:]
+      else:
+        g_l = torch.zeros_like(x_in)
+
+      acts = commons.fused_add_tanh_sigmoid_multiply(
+          x_in,
+          g_l,
+          n_channels_tensor)
+      acts = self.drop(acts)
+
+      res_skip_acts = self.res_skip_layers[i](acts)
+      if i < self.n_layers - 1:
+        res_acts = res_skip_acts[:,:self.hidden_channels,:]
+        x = (x + res_acts) * x_mask
+        output = output + res_skip_acts[:,self.hidden_channels:,:]
+      else:
+        output = output + res_skip_acts
+    return output * x_mask
+
+  def remove_weight_norm(self):
+    if self.gin_channels != 0:
+      torch.nn.utils.remove_weight_norm(self.cond_layer)
+    for l in self.in_layers:
+      torch.nn.utils.remove_weight_norm(l)
+    for l in self.res_skip_layers:
+     torch.nn.utils.remove_weight_norm(l)
+
+
+class ResBlock1(torch.nn.Module):
+    def __init__(self, channels, kernel_size=3, dilation=(1, 3, 5)):
+        super(ResBlock1, self).__init__()
+        self.convs1 = nn.ModuleList([
+            weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=dilation[0],
+                               padding=get_padding(kernel_size, dilation[0]))),
+            weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=dilation[1],
+                               padding=get_padding(kernel_size, dilation[1]))),
+            weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=dilation[2],
+                               padding=get_padding(kernel_size, dilation[2])))
+        ])
+        self.convs1.apply(init_weights)
+
+        self.convs2 = nn.ModuleList([
+            weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=1,
+                               padding=get_padding(kernel_size, 1))),
+            weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=1,
+                               padding=get_padding(kernel_size, 1))),
+            weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=1,
+                               padding=get_padding(kernel_size, 1)))
+        ])
+        self.convs2.apply(init_weights)
+
+    def forward(self, x, x_mask=None):
+        for c1, c2 in zip(self.convs1, self.convs2):
+            xt = F.leaky_relu(x, LRELU_SLOPE)
+            if x_mask is not None:
+                xt = xt * x_mask
+            xt = c1(xt)
+            xt = F.leaky_relu(xt, LRELU_SLOPE)
+            if x_mask is not None:
+                xt = xt * x_mask
+            xt = c2(xt)
+            x = xt + x
+        if x_mask is not None:
+            x = x * x_mask
+        return x
+
+    def remove_weight_norm(self):
+        for l in self.convs1:
+            remove_weight_norm(l)
+        for l in self.convs2:
+            remove_weight_norm(l)
+
+
+class ResBlock2(torch.nn.Module):
+    def __init__(self, channels, kernel_size=3, dilation=(1, 3)):
+        super(ResBlock2, self).__init__()
+        self.convs = nn.ModuleList([
+            weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=dilation[0],
+                               padding=get_padding(kernel_size, dilation[0]))),
+            weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=dilation[1],
+                               padding=get_padding(kernel_size, dilation[1])))
+        ])
+        self.convs.apply(init_weights)
+
+    def forward(self, x, x_mask=None):
+        for c in self.convs:
+            xt = F.leaky_relu(x, LRELU_SLOPE)
+            if x_mask is not None:
+                xt = xt * x_mask
+            xt = c(xt)
+            x = xt + x
+        if x_mask is not None:
+            x = x * x_mask
+        return x
+
+    def remove_weight_norm(self):
+        for l in self.convs:
+            remove_weight_norm(l)
+
+
+class Log(nn.Module):
+  def forward(self, x, x_mask, reverse=False, **kwargs):
+    if not reverse:
+      y = torch.log(torch.clamp_min(x, 1e-5)) * x_mask
+      logdet = torch.sum(-y, [1, 2])
+      return y, logdet
+    else:
+      x = torch.exp(x) * x_mask
+      return x
+    
+
+class Flip(nn.Module):
+  def forward(self, x, *args, reverse=False, **kwargs):
+    x = torch.flip(x, [1])
+    if not reverse:
+      logdet = torch.zeros(x.size(0)).to(dtype=x.dtype, device=x.device)
+      return x, logdet
+    else:
+      return x
+
+
+class ElementwiseAffine(nn.Module):
+  def __init__(self, channels):
+    super().__init__()
+    self.channels = channels
+    self.m = nn.Parameter(torch.zeros(channels,1))
+    self.logs = nn.Parameter(torch.zeros(channels,1))
+
+  def forward(self, x, x_mask, reverse=False, **kwargs):
+    if not reverse:
+      y = self.m + torch.exp(self.logs) * x
+      y = y * x_mask
+      logdet = torch.sum(self.logs * x_mask, [1,2])
+      return y, logdet
+    else:
+      x = (x - self.m) * torch.exp(-self.logs) * x_mask
+      return x
+
+
+class ResidualCouplingLayer(nn.Module):
+  def __init__(self,
+      channels,
+      hidden_channels,
+      kernel_size,
+      dilation_rate,
+      n_layers,
+      p_dropout=0,
+      gin_channels=0,
+      mean_only=False):
+    assert channels % 2 == 0, "channels should be divisible by 2"
+    super().__init__()
+    self.channels = channels
+    self.hidden_channels = hidden_channels
+    self.kernel_size = kernel_size
+    self.dilation_rate = dilation_rate
+    self.n_layers = n_layers
+    self.half_channels = channels // 2
+    self.mean_only = mean_only
+
+    self.pre = nn.Conv1d(self.half_channels, hidden_channels, 1)
+    self.enc = WN(hidden_channels, kernel_size, dilation_rate, n_layers, p_dropout=p_dropout, gin_channels=gin_channels)
+    self.post = nn.Conv1d(hidden_channels, self.half_channels * (2 - mean_only), 1)
+    self.post.weight.data.zero_()
+    self.post.bias.data.zero_()
+
+  def forward(self, x, x_mask, g=None, reverse=False):
+    x0, x1 = torch.split(x, [self.half_channels]*2, 1)
+    h = self.pre(x0) * x_mask
+    h = self.enc(h, x_mask, g=g)
+    stats = self.post(h) * x_mask
+    if not self.mean_only:
+      m, logs = torch.split(stats, [self.half_channels]*2, 1)
+    else:
+      m = stats
+      logs = torch.zeros_like(m)
+
+    if not reverse:
+      x1 = m + x1 * torch.exp(logs) * x_mask
+      x = torch.cat([x0, x1], 1)
+      logdet = torch.sum(logs, [1,2])
+      return x, logdet
+    else:
+      x1 = (x1 - m) * torch.exp(-logs) * x_mask
+      x = torch.cat([x0, x1], 1)
+      return x
+
+
+class ConvFlow(nn.Module):
+  def __init__(self, in_channels, filter_channels, kernel_size, n_layers, num_bins=10, tail_bound=5.0):
+    super().__init__()
+    self.in_channels = in_channels
+    self.filter_channels = filter_channels
+    self.kernel_size = kernel_size
+    self.n_layers = n_layers
+    self.num_bins = num_bins
+    self.tail_bound = tail_bound
+    self.half_channels = in_channels // 2
+
+    self.pre = nn.Conv1d(self.half_channels, filter_channels, 1)
+    self.convs = DDSConv(filter_channels, kernel_size, n_layers, p_dropout=0.)
+    self.proj = nn.Conv1d(filter_channels, self.half_channels * (num_bins * 3 - 1), 1)
+    self.proj.weight.data.zero_()
+    self.proj.bias.data.zero_()
+
+  def forward(self, x, x_mask, g=None, reverse=False):
+    x0, x1 = torch.split(x, [self.half_channels]*2, 1)
+    h = self.pre(x0)
+    h = self.convs(h, x_mask, g=g)
+    h = self.proj(h) * x_mask
+
+    b, c, t = x0.shape
+    h = h.reshape(b, c, -1, t).permute(0, 1, 3, 2) # [b, cx?, t] -> [b, c, t, ?]
+
+    unnormalized_widths = h[..., :self.num_bins] / math.sqrt(self.filter_channels)
+    unnormalized_heights = h[..., self.num_bins:2*self.num_bins] / math.sqrt(self.filter_channels)
+    unnormalized_derivatives = h[..., 2 * self.num_bins:]
+
+    x1, logabsdet = piecewise_rational_quadratic_transform(x1,
+        unnormalized_widths,
+        unnormalized_heights,
+        unnormalized_derivatives,
+        inverse=reverse,
+        tails='linear',
+        tail_bound=self.tail_bound
+    )
+
+    x = torch.cat([x0, x1], 1) * x_mask
+    logdet = torch.sum(logabsdet * x_mask, [1,2])
+    if not reverse:
+        return x, logdet
+    else:
+        return x

runtime/monotonic_align.py

@@ -0,0 +1,4 @@
+"""Training-only alignment stub; deployable inference never calls maximum_path."""
+
+def maximum_path(*args, **kwargs):
+    raise RuntimeError("Monotonic alignment is unavailable in the inference package.")

runtime/text/LICENSE

@@ -0,0 +1,19 @@
+Copyright (c) 2017 Keith Ito
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.

runtime/text/__init__.py

@@ -0,0 +1,54 @@
+""" from https://github.com/keithito/tacotron """
+from text import cleaners
+from text.symbols import symbols
+
+
+# Mappings from symbol to numeric ID and vice versa:
+_symbol_to_id = {s: i for i, s in enumerate(symbols)}
+_id_to_symbol = {i: s for i, s in enumerate(symbols)}
+
+
+def text_to_sequence(text, cleaner_names):
+  '''Converts a string of text to a sequence of IDs corresponding to the symbols in the text.
+    Args:
+      text: string to convert to a sequence
+      cleaner_names: names of the cleaner functions to run the text through
+    Returns:
+      List of integers corresponding to the symbols in the text
+  '''
+  sequence = []
+
+  clean_text = _clean_text(text, cleaner_names)
+  for symbol in clean_text:
+    symbol_id = _symbol_to_id[symbol]
+    sequence += [symbol_id]
+  return sequence
+
+
+def cleaned_text_to_sequence(cleaned_text):
+  '''Converts a string of text to a sequence of IDs corresponding to the symbols in the text.
+    Args:
+      text: string to convert to a sequence
+    Returns:
+      List of integers corresponding to the symbols in the text
+  '''
+  sequence = [_symbol_to_id[symbol] for symbol in cleaned_text]
+  return sequence
+
+
+def sequence_to_text(sequence):
+  '''Converts a sequence of IDs back to a string'''
+  result = ''
+  for symbol_id in sequence:
+    s = _id_to_symbol[symbol_id]
+    result += s
+  return result
+
+
+def _clean_text(text, cleaner_names):
+  for name in cleaner_names:
+    cleaner = getattr(cleaners, name)
+    if not cleaner:
+      raise Exception('Unknown cleaner: %s' % name)
+    text = cleaner(text)
+  return text

runtime/text/cleaners.py

@@ -0,0 +1,100 @@
+""" from https://github.com/keithito/tacotron """
+
+'''
+Cleaners are transformations that run over the input text at both training and eval time.
+
+Cleaners can be selected by passing a comma-delimited list of cleaner names as the "cleaners"
+hyperparameter. Some cleaners are English-specific. You'll typically want to use:
+  1. "english_cleaners" for English text
+  2. "transliteration_cleaners" for non-English text that can be transliterated to ASCII using
+     the Unidecode library (https://pypi.python.org/pypi/Unidecode)
+  3. "basic_cleaners" if you do not want to transliterate (in this case, you should also update
+     the symbols in symbols.py to match your data).
+'''
+
+import re
+from unidecode import unidecode
+from phonemizer import phonemize
+
+
+# Regular expression matching whitespace:
+_whitespace_re = re.compile(r'\s+')
+
+# List of (regular expression, replacement) pairs for abbreviations:
+_abbreviations = [(re.compile('\\b%s\\.' % x[0], re.IGNORECASE), x[1]) for x in [
+  ('mrs', 'misess'),
+  ('mr', 'mister'),
+  ('dr', 'doctor'),
+  ('st', 'saint'),
+  ('co', 'company'),
+  ('jr', 'junior'),
+  ('maj', 'major'),
+  ('gen', 'general'),
+  ('drs', 'doctors'),
+  ('rev', 'reverend'),
+  ('lt', 'lieutenant'),
+  ('hon', 'honorable'),
+  ('sgt', 'sergeant'),
+  ('capt', 'captain'),
+  ('esq', 'esquire'),
+  ('ltd', 'limited'),
+  ('col', 'colonel'),
+  ('ft', 'fort'),
+]]
+
+
+def expand_abbreviations(text):
+  for regex, replacement in _abbreviations:
+    text = re.sub(regex, replacement, text)
+  return text
+
+
+def expand_numbers(text):
+  return normalize_numbers(text)
+
+
+def lowercase(text):
+  return text.lower()
+
+
+def collapse_whitespace(text):
+  return re.sub(_whitespace_re, ' ', text)
+
+
+def convert_to_ascii(text):
+  return unidecode(text)
+
+
+def basic_cleaners(text):
+  '''Basic pipeline that lowercases and collapses whitespace without transliteration.'''
+  text = lowercase(text)
+  text = collapse_whitespace(text)
+  return text
+
+
+def transliteration_cleaners(text):
+  '''Pipeline for non-English text that transliterates to ASCII.'''
+  text = convert_to_ascii(text)
+  text = lowercase(text)
+  text = collapse_whitespace(text)
+  return text
+
+
+def english_cleaners(text):
+  '''Pipeline for English text, including abbreviation expansion.'''
+  text = convert_to_ascii(text)
+  text = lowercase(text)
+  text = expand_abbreviations(text)
+  phonemes = phonemize(text, language='en-us', backend='espeak', strip=True)
+  phonemes = collapse_whitespace(phonemes)
+  return phonemes
+
+
+def english_cleaners2(text):
+  '''Pipeline for English text, including abbreviation expansion. + punctuation + stress'''
+  text = convert_to_ascii(text)
+  text = lowercase(text)
+  text = expand_abbreviations(text)
+  phonemes = phonemize(text, language='en-us', backend='espeak', strip=True, preserve_punctuation=True, with_stress=True)
+  phonemes = collapse_whitespace(phonemes)
+  return phonemes

runtime/text/symbols.py

@@ -0,0 +1,16 @@
+""" from https://github.com/keithito/tacotron """
+
+'''
+Defines the set of symbols used in text input to the model.
+'''
+_pad        = '_'
+_punctuation = ';:,.!?¡¿—…"«»“” '
+_letters = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz'
+_letters_ipa = "ɑɐɒæɓʙβɔɕçɗɖðʤəɘɚɛɜɝɞɟʄɡɠɢʛɦɧħɥʜɨɪʝɭɬɫɮʟɱɯɰŋɳɲɴøɵɸθœɶʘɹɺɾɻʀʁɽʂʃʈʧʉʊʋⱱʌɣɤʍχʎʏʑʐʒʔʡʕʢǀǁǂǃˈˌːˑʼʴʰʱʲʷˠˤ˞↓↑→↗↘'̩'ᵻ"
+
+
+# Export all symbols:
+symbols = [_pad] + list(_punctuation) + list(_letters) + list(_letters_ipa)
+
+# Special symbol ids
+SPACE_ID = symbols.index(" ")

runtime/transforms.py

@@ -0,0 +1,193 @@
+import torch
+from torch.nn import functional as F
+
+import numpy as np
+
+
+DEFAULT_MIN_BIN_WIDTH = 1e-3
+DEFAULT_MIN_BIN_HEIGHT = 1e-3
+DEFAULT_MIN_DERIVATIVE = 1e-3
+
+
+def piecewise_rational_quadratic_transform(inputs, 
+                                           unnormalized_widths,
+                                           unnormalized_heights,
+                                           unnormalized_derivatives,
+                                           inverse=False,
+                                           tails=None, 
+                                           tail_bound=1.,
+                                           min_bin_width=DEFAULT_MIN_BIN_WIDTH,
+                                           min_bin_height=DEFAULT_MIN_BIN_HEIGHT,
+                                           min_derivative=DEFAULT_MIN_DERIVATIVE):
+
+    if tails is None:
+        spline_fn = rational_quadratic_spline
+        spline_kwargs = {}
+    else:
+        spline_fn = unconstrained_rational_quadratic_spline
+        spline_kwargs = {
+            'tails': tails,
+            'tail_bound': tail_bound
+        }
+
+    outputs, logabsdet = spline_fn(
+            inputs=inputs,
+            unnormalized_widths=unnormalized_widths,
+            unnormalized_heights=unnormalized_heights,
+            unnormalized_derivatives=unnormalized_derivatives,
+            inverse=inverse,
+            min_bin_width=min_bin_width,
+            min_bin_height=min_bin_height,
+            min_derivative=min_derivative,
+            **spline_kwargs
+    )
+    return outputs, logabsdet
+
+
+def searchsorted(bin_locations, inputs, eps=1e-6):
+    bin_locations[..., -1] += eps
+    return torch.sum(
+        inputs[..., None] >= bin_locations,
+        dim=-1
+    ) - 1
+
+
+def unconstrained_rational_quadratic_spline(inputs,
+                                            unnormalized_widths,
+                                            unnormalized_heights,
+                                            unnormalized_derivatives,
+                                            inverse=False,
+                                            tails='linear',
+                                            tail_bound=1.,
+                                            min_bin_width=DEFAULT_MIN_BIN_WIDTH,
+                                            min_bin_height=DEFAULT_MIN_BIN_HEIGHT,
+                                            min_derivative=DEFAULT_MIN_DERIVATIVE):
+    inside_interval_mask = (inputs >= -tail_bound) & (inputs <= tail_bound)
+    outside_interval_mask = ~inside_interval_mask
+
+    outputs = torch.zeros_like(inputs)
+    logabsdet = torch.zeros_like(inputs)
+
+    if tails == 'linear':
+        unnormalized_derivatives = F.pad(unnormalized_derivatives, pad=(1, 1))
+        constant = np.log(np.exp(1 - min_derivative) - 1)
+        unnormalized_derivatives[..., 0] = constant
+        unnormalized_derivatives[..., -1] = constant
+
+        outputs[outside_interval_mask] = inputs[outside_interval_mask]
+        logabsdet[outside_interval_mask] = 0
+    else:
+        raise RuntimeError('{} tails are not implemented.'.format(tails))
+
+    outputs[inside_interval_mask], logabsdet[inside_interval_mask] = rational_quadratic_spline(
+        inputs=inputs[inside_interval_mask],
+        unnormalized_widths=unnormalized_widths[inside_interval_mask, :],
+        unnormalized_heights=unnormalized_heights[inside_interval_mask, :],
+        unnormalized_derivatives=unnormalized_derivatives[inside_interval_mask, :],
+        inverse=inverse,
+        left=-tail_bound, right=tail_bound, bottom=-tail_bound, top=tail_bound,
+        min_bin_width=min_bin_width,
+        min_bin_height=min_bin_height,
+        min_derivative=min_derivative
+    )
+
+    return outputs, logabsdet
+
+def rational_quadratic_spline(inputs,
+                              unnormalized_widths,
+                              unnormalized_heights,
+                              unnormalized_derivatives,
+                              inverse=False,
+                              left=0., right=1., bottom=0., top=1.,
+                              min_bin_width=DEFAULT_MIN_BIN_WIDTH,
+                              min_bin_height=DEFAULT_MIN_BIN_HEIGHT,
+                              min_derivative=DEFAULT_MIN_DERIVATIVE):
+    if torch.min(inputs) < left or torch.max(inputs) > right:
+        raise ValueError('Input to a transform is not within its domain')
+
+    num_bins = unnormalized_widths.shape[-1]
+
+    if min_bin_width * num_bins > 1.0:
+        raise ValueError('Minimal bin width too large for the number of bins')
+    if min_bin_height * num_bins > 1.0:
+        raise ValueError('Minimal bin height too large for the number of bins')
+
+    widths = F.softmax(unnormalized_widths, dim=-1)
+    widths = min_bin_width + (1 - min_bin_width * num_bins) * widths
+    cumwidths = torch.cumsum(widths, dim=-1)
+    cumwidths = F.pad(cumwidths, pad=(1, 0), mode='constant', value=0.0)
+    cumwidths = (right - left) * cumwidths + left
+    cumwidths[..., 0] = left
+    cumwidths[..., -1] = right
+    widths = cumwidths[..., 1:] - cumwidths[..., :-1]
+
+    derivatives = min_derivative + F.softplus(unnormalized_derivatives)
+
+    heights = F.softmax(unnormalized_heights, dim=-1)
+    heights = min_bin_height + (1 - min_bin_height * num_bins) * heights
+    cumheights = torch.cumsum(heights, dim=-1)
+    cumheights = F.pad(cumheights, pad=(1, 0), mode='constant', value=0.0)
+    cumheights = (top - bottom) * cumheights + bottom
+    cumheights[..., 0] = bottom
+    cumheights[..., -1] = top
+    heights = cumheights[..., 1:] - cumheights[..., :-1]
+
+    if inverse:
+        bin_idx = searchsorted(cumheights, inputs)[..., None]
+    else:
+        bin_idx = searchsorted(cumwidths, inputs)[..., None]
+
+    input_cumwidths = cumwidths.gather(-1, bin_idx)[..., 0]
+    input_bin_widths = widths.gather(-1, bin_idx)[..., 0]
+
+    input_cumheights = cumheights.gather(-1, bin_idx)[..., 0]
+    delta = heights / widths
+    input_delta = delta.gather(-1, bin_idx)[..., 0]
+
+    input_derivatives = derivatives.gather(-1, bin_idx)[..., 0]
+    input_derivatives_plus_one = derivatives[..., 1:].gather(-1, bin_idx)[..., 0]
+
+    input_heights = heights.gather(-1, bin_idx)[..., 0]
+
+    if inverse:
+        a = (((inputs - input_cumheights) * (input_derivatives
+                                             + input_derivatives_plus_one
+                                             - 2 * input_delta)
+              + input_heights * (input_delta - input_derivatives)))
+        b = (input_heights * input_derivatives
+             - (inputs - input_cumheights) * (input_derivatives
+                                              + input_derivatives_plus_one
+                                              - 2 * input_delta))
+        c = - input_delta * (inputs - input_cumheights)
+
+        discriminant = b.pow(2) - 4 * a * c
+        assert (discriminant >= 0).all()
+
+        root = (2 * c) / (-b - torch.sqrt(discriminant))
+        outputs = root * input_bin_widths + input_cumwidths
+
+        theta_one_minus_theta = root * (1 - root)
+        denominator = input_delta + ((input_derivatives + input_derivatives_plus_one - 2 * input_delta)
+                                     * theta_one_minus_theta)
+        derivative_numerator = input_delta.pow(2) * (input_derivatives_plus_one * root.pow(2)
+                                                     + 2 * input_delta * theta_one_minus_theta
+                                                     + input_derivatives * (1 - root).pow(2))
+        logabsdet = torch.log(derivative_numerator) - 2 * torch.log(denominator)
+
+        return outputs, -logabsdet
+    else:
+        theta = (inputs - input_cumwidths) / input_bin_widths
+        theta_one_minus_theta = theta * (1 - theta)
+
+        numerator = input_heights * (input_delta * theta.pow(2)
+                                     + input_derivatives * theta_one_minus_theta)
+        denominator = input_delta + ((input_derivatives + input_derivatives_plus_one - 2 * input_delta)
+                                     * theta_one_minus_theta)
+        outputs = input_cumheights + numerator / denominator
+
+        derivative_numerator = input_delta.pow(2) * (input_derivatives_plus_one * theta.pow(2)
+                                                     + 2 * input_delta * theta_one_minus_theta
+                                                     + input_derivatives * (1 - theta).pow(2))
+        logabsdet = torch.log(derivative_numerator) - 2 * torch.log(denominator)
+
+        return outputs, logabsdet

runtime/utils.py

@@ -0,0 +1,256 @@
+import os
+import glob
+import sys
+import argparse
+import logging
+import json
+import subprocess
+import numpy as np
+from scipy.io.wavfile import read
+import torch
+
+MATPLOTLIB_FLAG = False
+
+logging.basicConfig(stream=sys.stdout, level=logging.DEBUG)
+logger = logging
+
+
+def load_checkpoint(checkpoint_path, model, optimizer=None):
+  assert os.path.isfile(checkpoint_path)
+  checkpoint_dict = torch.load(checkpoint_path, map_location='cpu')
+  iteration = checkpoint_dict['iteration']
+  learning_rate = checkpoint_dict['learning_rate']
+  if optimizer is not None:
+    optimizer.load_state_dict(checkpoint_dict['optimizer'])
+  saved_state_dict = checkpoint_dict['model']
+  if hasattr(model, 'module'):
+    state_dict = model.module.state_dict()
+  else:
+    state_dict = model.state_dict()
+  new_state_dict= {}
+  for k, v in state_dict.items():
+    try:
+      new_state_dict[k] = saved_state_dict[k]
+    except:
+      logger.info("%s is not in the checkpoint" % k)
+      new_state_dict[k] = v
+  if hasattr(model, 'module'):
+    model.module.load_state_dict(new_state_dict)
+  else:
+    model.load_state_dict(new_state_dict)
+  logger.info("Loaded checkpoint '{}' (iteration {})" .format(
+    checkpoint_path, iteration))
+  return model, optimizer, learning_rate, iteration
+
+
+def save_checkpoint(model, optimizer, learning_rate, iteration, checkpoint_path):
+  logger.info("Saving model and optimizer state at iteration {} to {}".format(
+    iteration, checkpoint_path))
+  if hasattr(model, 'module'):
+    state_dict = model.module.state_dict()
+  else:
+    state_dict = model.state_dict()
+  torch.save({'model': state_dict,
+              'iteration': iteration,
+              'optimizer': optimizer.state_dict(),
+              'learning_rate': learning_rate}, checkpoint_path)
+
+
+def summarize(writer, global_step, scalars={}, histograms={}, images={}, audios={}, audio_sampling_rate=22050):
+  for k, v in scalars.items():
+    writer.add_scalar(k, v, global_step)
+  for k, v in histograms.items():
+    writer.add_histogram(k, v, global_step)
+  for k, v in images.items():
+    writer.add_image(k, v, global_step, dataformats='HWC')
+  for k, v in audios.items():
+    writer.add_audio(k, v, global_step, audio_sampling_rate)
+
+
+def latest_checkpoint_path(dir_path, regex="G_*.pth"):
+  f_list = glob.glob(os.path.join(dir_path, regex))
+  f_list.sort(key=lambda f: int("".join(filter(str.isdigit, f))))
+  x = f_list[-1]
+  print(x)
+  return x
+
+
+def plot_spectrogram_to_numpy(spectrogram):
+  global MATPLOTLIB_FLAG
+  if not MATPLOTLIB_FLAG:
+    import matplotlib
+    matplotlib.use("Agg")
+    MATPLOTLIB_FLAG = True
+    mpl_logger = logging.getLogger('matplotlib')
+    mpl_logger.setLevel(logging.WARNING)
+  import matplotlib.pylab as plt
+  import numpy as np
+  
+  fig, ax = plt.subplots(figsize=(10,2))
+  im = ax.imshow(spectrogram, aspect="auto", origin="lower",
+                  interpolation='none')
+  plt.colorbar(im, ax=ax)
+  plt.xlabel("Frames")
+  plt.ylabel("Channels")
+  plt.tight_layout()
+
+  fig.canvas.draw()
+  data = np.asarray(fig.canvas.buffer_rgba(), dtype=np.uint8)[..., :3].copy()
+  plt.close()
+  return data
+
+
+def plot_alignment_to_numpy(alignment, info=None):
+  global MATPLOTLIB_FLAG
+  if not MATPLOTLIB_FLAG:
+    import matplotlib
+    matplotlib.use("Agg")
+    MATPLOTLIB_FLAG = True
+    mpl_logger = logging.getLogger('matplotlib')
+    mpl_logger.setLevel(logging.WARNING)
+  import matplotlib.pylab as plt
+  import numpy as np
+
+  fig, ax = plt.subplots(figsize=(6, 4))
+  im = ax.imshow(alignment.transpose(), aspect='auto', origin='lower',
+                  interpolation='none')
+  fig.colorbar(im, ax=ax)
+  xlabel = 'Decoder timestep'
+  if info is not None:
+      xlabel += '\n\n' + info
+  plt.xlabel(xlabel)
+  plt.ylabel('Encoder timestep')
+  plt.tight_layout()
+
+  fig.canvas.draw()
+  data = np.asarray(fig.canvas.buffer_rgba(), dtype=np.uint8)[..., :3].copy()
+  plt.close()
+  return data
+
+
+def load_wav_to_torch(full_path):
+  sampling_rate, data = read(full_path)
+  return torch.FloatTensor(data.astype(np.float32)), sampling_rate
+
+
+def load_filepaths_and_text(filename, split="|"):
+  with open(filename, encoding='utf-8') as f:
+    filepaths_and_text = [line.strip().split(split) for line in f]
+  return filepaths_and_text
+
+
+def get_hparams(init=True):
+  parser = argparse.ArgumentParser()
+  parser.add_argument('-c', '--config', type=str, default="./configs/base.json",
+                      help='JSON file for configuration')
+  parser.add_argument('-m', '--model', type=str, required=True,
+                      help='Model name')
+  
+  args = parser.parse_args()
+  model_dir = os.path.join("./logs", args.model)
+
+  if not os.path.exists(model_dir):
+    os.makedirs(model_dir)
+
+  config_path = args.config
+  config_save_path = os.path.join(model_dir, "config.json")
+  if init:
+    with open(config_path, "r") as f:
+      data = f.read()
+    with open(config_save_path, "w") as f:
+      f.write(data)
+  else:
+    with open(config_save_path, "r") as f:
+      data = f.read()
+  config = json.loads(data)
+  
+  hparams = HParams(**config)
+  hparams.model_dir = model_dir
+  return hparams
+
+
+def get_hparams_from_dir(model_dir):
+  config_save_path = os.path.join(model_dir, "config.json")
+  with open(config_save_path, "r") as f:
+    data = f.read()
+  config = json.loads(data)
+
+  hparams =HParams(**config)
+  hparams.model_dir = model_dir
+  return hparams
+
+
+def get_hparams_from_file(config_path):
+  with open(config_path, "r") as f:
+    data = f.read()
+  config = json.loads(data)
+
+  hparams =HParams(**config)
+  return hparams
+
+
+def check_git_hash(model_dir):
+  source_dir = os.path.dirname(os.path.realpath(__file__))
+  if not os.path.exists(os.path.join(source_dir, ".git")):
+    logger.warn("{} is not a git repository, therefore hash value comparison will be ignored.".format(
+      source_dir
+    ))
+    return
+
+  cur_hash = subprocess.getoutput("git rev-parse HEAD")
+
+  path = os.path.join(model_dir, "githash")
+  if os.path.exists(path):
+    saved_hash = open(path).read()
+    if saved_hash != cur_hash:
+      logger.warn("git hash values are different. {}(saved) != {}(current)".format(
+        saved_hash[:8], cur_hash[:8]))
+  else:
+    open(path, "w").write(cur_hash)
+
+
+def get_logger(model_dir, filename="train.log"):
+  global logger
+  logger = logging.getLogger(os.path.basename(model_dir))
+  logger.setLevel(logging.DEBUG)
+  
+  formatter = logging.Formatter("%(asctime)s\t%(name)s\t%(levelname)s\t%(message)s")
+  if not os.path.exists(model_dir):
+    os.makedirs(model_dir)
+  h = logging.FileHandler(os.path.join(model_dir, filename))
+  h.setLevel(logging.DEBUG)
+  h.setFormatter(formatter)
+  logger.addHandler(h)
+  return logger
+
+
+class HParams():
+  def __init__(self, **kwargs):
+    for k, v in kwargs.items():
+      if type(v) == dict:
+        v = HParams(**v)
+      self[k] = v
+    
+  def keys(self):
+    return self.__dict__.keys()
+
+  def items(self):
+    return self.__dict__.items()
+
+  def values(self):
+    return self.__dict__.values()
+
+  def __len__(self):
+    return len(self.__dict__)
+
+  def __getitem__(self, key):
+    return getattr(self, key)
+
+  def __setitem__(self, key, value):
+    return setattr(self, key, value)
+
+  def __contains__(self, key):
+    return key in self.__dict__
+
+  def __repr__(self):
+    return self.__dict__.__repr__()

samples/male/conversational.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f9e0f1d0ff2c5ced4ac05d8bc44b8a14a8b0c79ea36ed5aaa936df1840e899ed
+size 180780

samples/male/names_places.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fdeea745130c5d873fcb6ea8a5674d529c325739b2ad307fa422e7e2c9fd4997
+size 182828

samples/male/numbers.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a73cc26c31c11b1a78d26f8ebd382de4efb37b3514d7f2c91b9528ca79e7daf1
+size 229932

samples/male/punctuation.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca433246649b7588982afa6dd67db2f40edf238b8ef87f865d8bb3eb885f386a
+size 255020

samples/male/technical.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b59ad3d449c22b851d4a5f882f7a5e9d3413d0bc1e1037aa3374a3470c91114
+size 192556

smoke/manifest.jsonl

@@ -0,0 +1 @@
+{"id": "release_smoke", "category": "release", "text": "The final release package passed its deterministic synthesis check.", "model": "inflect-micro-v2-release", "audio": "release_polish_overnight_20260720/release_candidates/Inflect-Micro-v2-male/smoke/release_smoke.wav"}

smoke/release_smoke.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31be94d7b74095b22b68dd5750596e23406e9a85d51bb9fbc49969bfa68abdbf
+size 190508

smoke/standalone_cpu.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:912f69948b48abdc1227131b8fc07e1fce4ab7b6046016c9e9b63c4b729a5152
+size 173612

smoke_prompt.jsonl

@@ -0,0 +1 @@
+{"id": "release_smoke", "category": "release", "text": "The final release package passed its deterministic synthesis check."}

third_party/BIGVGAN_LICENSE.txt

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2024 NVIDIA CORPORATION.
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

third_party/VITS_LICENSE.txt

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2021 Jaehyeon Kim
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.