loudr-1: complete v0.1.1 bundle

NOTICE

@@ -75,8 +75,10 @@ Voice encoder (`ve.safetensors`)
     MIT terms.
 
     `docs/PROVENANCE-voice-encoder.md` records the artefact's size and sha256,
-    its tensors, the upstream repository and licence, and the fact that the
-    upstream revision is not recorded in the artefact metadata.
+    its tensors, the upstream repository and licence, and an exact tensor match
+    with ResembleAI/chatterbox revision
+    5bb1f6ee58e50c3b8d408bc82a6d3740c2db6e18. The original export's checkout
+    revision was not recorded.
 
     MIT License text:
 
@@ -183,14 +185,14 @@ Unicode CLDR
     Licensed under the Unicode License v3 — https://www.unicode.org/license.txt
 
     `tests/data/cldr/*.ssv` are CLDR's own RBNF spellout test data, imported
-    verbatim with their copyright header intact, and `numbers_cldr.json` is the
+    verbatim with their copyright header intact, and `tests/data/conformance/numbers_cldr.json` is the
     subset mapped onto rulesets this project implements. Test corpus only: none
     of it is read at runtime and none of it ships in the wheel or the sdist. It
     is a differential test corpus, independent of this implementation.
 
 NST pronunciation lexicons
     The Swedish, Danish and Norwegian NST lexicons that
-    `tools/fetch_nst_lexicons.py` downloads are public domain (CC0), from
+    `research/fetch_nst_lexicons.py` downloads are public domain (CC0), from
     Språkbanken Tal. They are fetched on demand and never vendored, so nothing
     of theirs is redistributed here.
 

README.md

@@ -26,14 +26,19 @@ language:
 
 # loudr-1
 
-**Natural-sounding text-to-speech with twenty voices, ten languages and voice
+**Natural-sounding text-to-speech with 28 voices, ten languages and voice
 cloning.**
 
 loudr-1 runs on your own hardware through
 [loudkit](https://github.com/loudreader/loudkit). Download it once and work offline
 from Python, Swift, Go, Rust or TypeScript with PyTorch, ONNX Runtime or CoreML.
 
-[**Hear all 20 voices**](https://loudreader.github.io/loudkit/demo/) |
+These are the weights behind [LoudReader](https://loudreader.io), a reading app
+that speaks articles, PDFs and books on device. They are published here so the
+engine can be used and checked on its own.
+
+[**Try it in the browser**](https://huggingface.co/spaces/jer3mi/loudkit) |
+[**Hear all 28 voices**](https://loudreader.github.io/loudkit/demo/) |
 [**Open in Colab**](https://colab.research.google.com/github/loudreader/loudkit/blob/main/notebooks/loudkit_quickstart.ipynb) |
 [**GitHub**](https://github.com/loudreader/loudkit) |
 [**Documentation**](https://loudreader.github.io/loudkit/)
@@ -67,7 +72,7 @@ pip install "loudkit[torch,audio,hub]"
 import loudkit as lk
 
 engine = lk.load("loudreader/loudr-1")
-voice = lk.voice("joe", repo="loudreader/loudr-1")
+voice = engine.voice("joe")
 
 engine.synthesize("Hello from loudkit.", voice, seed=7).save("hello.wav")
 ```
@@ -90,49 +95,33 @@ mine.save("voices/my-voice.safetensors")
 The reusable profile is about 150 KB. Install
 `loudkit[torch,audio,enroll,hub]` for enrollment.
 
-## Choose your runtime
-
-The repository contains all supported formats, but the downloader fetches only
-the runtime you select. Add `--with-cloning` when the installation also needs
-enrollment.
-
-| path | command | download |
-|---|---|---:|
-| Python, synthesis | `loudkit download loudreader/loudr-1 --for torch` | 750 MB |
-| Python, with cloning | `--for torch --with-cloning` | 1.28 GB |
-| JS, Go or Rust with ONNX | `--for onnx` | 2.60 GB |
-| ONNX, with cloning | `--for onnx --with-cloning` | 3.13 GB |
-| Swift or Python with CoreML | `--for coreml` | 1.16 GB |
-| CoreML, with cloning | `--for coreml --with-cloning` | 1.69 GB |
-
-Add `--local-dir loudr-1` to create a portable directory instead of using the
-shared cache. The synthesis checkpoint will be at
-`loudr-1/loudr-1.safetensors`.
-
-### Measured speed
-
-| path | hardware | real-time factor |
-|---|---|---:|
-| PyTorch with CUDA graphs | RTX 3090 | 7.47x |
-| PyTorch with CUDA graphs | Jetson Orin Nano | 1.83x |
-| split PyTorch engine\* | Apple M3 Pro | 3.43x |
-| ONNX Runtime, CPU provider | Apple M3 Pro | 1.21x |
-| PyTorch CPU reference | Apple M3 Pro | 0.33x |
-
-\* "Split" describes device placement, not a different model or checkpoint.
-The token generator runs on the CPU while the mel and vocoder renderer runs on
-the Apple GPU through MPS. Adjacent windows can overlap across the two devices.
-
-Higher is faster, and 1.0x means real time. ONNX Runtime on the measured M3 Pro
-CPU is faster than real time. The PyTorch CPU reference path on the same machine
-is not.
-
-For batched workloads, the token generator reaches 20.1x aggregate throughput
-at batch 1 and 153.1x at batch 64 on the RTX 3090. The highest measured result
-is 170.8x on an A100 at batch 64. These are generator-only throughput numbers,
-not single-request latency or end-to-end RTF. See the
-[benchmark report](https://loudreader.github.io/loudkit/benchmarks/) for commands,
-hardware and caveats.
+The Swift, Go, Rust and TypeScript packages load the same repo id and fetch
+what they need themselves; the
+[README](https://github.com/loudreader/loudkit#the-same-in-swift-go-rust-and-typescript)
+shows each one.
+
+## What each runtime downloads
+
+The repository contains every supported format. A download takes one runtime
+and leaves the rest behind. Add `--with-cloning` when the installation also
+needs enrollment.
+
+| path | command |
+|---|---|
+| Python, synthesis | `loudkit download loudreader/loudr-1 --for torch` |
+| Python, with cloning | `--for torch --with-cloning` |
+| TypeScript, Go or Rust with ONNX Runtime | `--for onnx` |
+| ONNX Runtime, with cloning | `--for onnx --with-cloning` |
+| Swift or Python with CoreML | `--for coreml` |
+| CoreML, with cloning | `--for coreml --with-cloning` |
+
+Download size depends on the model, backend and release revision. The
+synthesis checkpoint is 747 MB; graph downloads also include backend-specific
+weights.
+
+Speed depends on the hardware and the backend.
+[Benchmarks](https://loudreader.github.io/loudkit/benchmarks/) has the
+measured figures, the machines and the commands.
 
 ## What ships
 
@@ -141,33 +130,33 @@ hardware and caveats.
 | `loudr-1.safetensors` | 747 MB | synthesis |
 | `loudr-1-enrollment.safetensors` | 523 MB | PyTorch enrollment |
 | `ve.safetensors` | 5.7 MB | PyTorch enrollment |
-| `onnx/` | 2.38 GB | nine graphs: six synthesis, three enrollment |
-| `coreml/` | 941 MB | six packages: three synthesis, three enrollment |
-| `voices/` | 3.1 MB | twenty voice profiles |
+| `onnx/` | varies by model | synthesis and enrollment graphs |
+| `coreml/` | varies by model | native generation, rendering and enrollment packages |
+| `voices/` | 4.3 MB | 28 voice profiles |
 | `samples/` | 108 KB | the two players above |
 | `tokenizer.json` | 70 KB | text processing |
 
 Synthesis and enrollment are separate so users who only need speech generation
 do not download the enrollment weights. ONNX and CoreML use their own enrollment
-graphs. loudkit also verifies that paired model files came from the same source
-checkpoint.
+graphs. Every download is checked against the release's `SHA256SUMS` before
+it is used.
 
 ## Voices and consent
 
-The release includes two profiles for each of these languages: English,
-Spanish, French, German, Italian, Polish, Portuguese, Dutch, Swedish and Danish.
+The release includes ten English profiles and two for each of Spanish, French,
+German, Italian, Polish, Portuguese, Dutch, Swedish and Danish.
 
 The profiles were built from recordings donated for speech technology or from
 CC0 and CC-BY speech corpora. No scraped celebrity voices ship with the model.
 [The full roster](https://github.com/loudreader/loudkit/blob/main/VOICES.md) records
 the source, licence and consent basis for every profile. The
 [voice gallery](https://loudreader.github.io/loudkit/demo/) provides a generated
-sample and enrollment preview for all twenty.
+sample and enrollment preview for all 28.
 
 The source enrollment WAVs are not redistributed in the model repository. Their
-digests, construction notes and the digests of every shipped profile and sample
+hashes, construction notes and the hashes of every shipped profile and sample
 are recorded in
-[provenance.json](https://github.com/loudreader/loudkit/blob/main/docs/voices/roster/provenance.json).
+[the roster record](https://github.com/loudreader/loudkit/blob/main/docs/voices/roster/provenance.json).
 
 ## Model lineage
 
@@ -202,9 +191,9 @@ and [measured parity report](https://loudreader.github.io/loudkit/parity-measure
   prosody.
 - Voice cloning requires consent. A recording being public does not grant
   permission to clone the speaker.
-- Saved WAVs and server responses include an unsigned C2PA Content Credentials
-  manifest by default. It records the model, voice, seed, backend and audio
-  digest in a machine-readable form.
+- Saved WAVs and server responses carry unsigned C2PA Content Credentials by
+  default, recording the model, voice, seed, backend and a hash of the audio
+  in a machine-readable form.
 
 Read [Responsible use](https://huggingface.co/loudreader/loudr-1/blob/main/RESPONSIBLE_USE.md)
 before exposing enrollment to other people.
@@ -227,3 +216,14 @@ the public roster.
 [Apache-2.0](https://huggingface.co/loudreader/loudr-1/blob/main/LICENSE).
 Upstream attributions and component licences are listed in
 [NOTICE](https://huggingface.co/loudreader/loudr-1/blob/main/NOTICE).
+
+## Download sizes for 0.1.1
+
+Approximate decimal sizes for the release files; backend weights are included.
+Cloning adds enrollment assets only when requested. Both models ship separate
+synthesis and enrollment checkpoints.
+
+| Model | Torch | Torch + cloning | ONNX | ONNX + cloning | CoreML | CoreML + cloning |
+|---|---:|---:|---:|---:|---:|---:|
+| loudr-1 | 0.75 GB | 1.28 GB | 2.60 GB | 3.13 GB | 2.46 GB | 2.99 GB |
+| loudr-1-turbo | 0.72 GB | 1.25 GB | 2.44 GB | 2.97 GB | 2.44 GB | 2.97 GB |

RESPONSIBLE_USE.md

@@ -65,7 +65,7 @@ add their own signing policy and disclosure around it.
 
 ## What we ship, and what we will not
 
-We ship twenty voice profiles enrolled from recordings donated for speech
+We ship 28 voice profiles enrolled from recordings donated for speech
 technology or released under terms that permit this use. Every profile has a
 named source, licence and consent basis in the public roster. We do not ship
 profiles made from private recordings or recordings published for an unrelated
@@ -77,3 +77,7 @@ you have the right to use.
 Issues and pull requests asking for help with undisclosed impersonation, voice
 authentication bypass, or stripping provenance from generated audio will be
 closed.
+
+The same consent and attribution requirements apply to graph-based cloning
+without PyTorch and to voice profiles created by Python, Swift, Go, Rust or
+TypeScript. A portable profile can reproduce a voice in either model.

SHA256SUMS

@@ -1,6 +1,6 @@
-688e87cddfc95a37b6705bda05ad95a7ddb219a4c0c232fa4b93ccadc42efe2c  loudr-1.safetensors
-dc825dcdec843cc0fb40ad3a0f051e869db97689a33712d1d11c009ce16af022  loudr-1-enrollment.safetensors
-795d019797197a0f01c722af516ac6521cda59c6a27d8d07828489175659a99f  manifest.json
+73e69a78d58176a3e4b39ca992c7b37a9c51f7251d08e4334c27d9994faac44e  loudr-1.safetensors
+ac7f4a0ce1bf16855f6d1c5fe048bbfb90b6a114d74bc2cd85974d11d02d0d83  loudr-1-enrollment.safetensors
+9fde37e0dd43150725ea09bf15711602f2a9dfffca7a82b22aeeefbbd3c8f3bf  manifest.json
 69632f47220a788a52ce2661d096453c5655e9bf25289d89a8d832c46ee07dbf  tokenizer.json
 f0921cab452fa278bc25cd23ffd59d36f816d7dc5181dd1bef9751a7fb61f63c  ve.safetensors
 78988a6769f90c209240f9f285eb9cdd760d1ba471cfa7d6e67db632599cfa1b  voices/darkman.safetensors
@@ -23,38 +23,57 @@ be75721c99547ab3ecdf1f8629101e9fa3ea1d569fc6e1220b8e2f259fb49f4e  voices/nils.sa
 d3ba9ab18c676db23028c9561ad79a183b35adae377d20d46f1d8f253ce6eeec  voices/selma.safetensors
 30d9fe9ee02928cd65ef7f4cb286c401c1809f07c13c19de79ffa375a027412e  voices/soren.safetensors
 0a3f913b9b47c529393f843850228bd8eee6a883b4c138978fd3afa7758561e8  voices/freja.safetensors
-7de8b5afaa4b352673508768de757195fc7fda340efdb6f56a4a48d663df6a89  onnx/t3_cond.onnx
-fdbb5dfe5e0651adf6961a62b9adea45c90bef5849a05b4968bc00c923368766  onnx/t3_prefill.onnx
-e8c39bc5512dd3fbac441ce5322673535ecbe281333bece33ea88330a348ca03  onnx/t3_step.onnx
+bd2fdb4ff8fbc12b070654dc5dae82b24aee4990bda726185547ec613b9f9cb8  voices/oliver.safetensors
+5476670162065397f01fca60a1944eb75ddef8dc7e6792dde7fb603c76c14d8c  voices/henry.safetensors
+184d5552881d2e6fd644f0546d9b470130664d84faf39a31979001cafe8f4ed9  voices/miles.safetensors
+8b8ff60604338eedddcd7317ed76ad2109fba6fa0ab79c9684a7954b64f963b2  voices/oscar.safetensors
+c2b5db4c210657d6b2b81602fae8e73a733ac4312f33e80536a59a874e7d865f  voices/clara.safetensors
+0f5d8bfaceca353ca92db751c48b6a1653823f421ead3c901efe6366b8fa2dfb  voices/emma.safetensors
+009a5fe1b34a9c65b09254f41c01cfdc641ec671a83449ad8aa77dd933d5f27e  voices/lucy.safetensors
+efb8047f93eda2a8e7cedaaa254b51957a845d55287298868a409662f8f75e1a  voices/sophie.safetensors
+483988e5adac3f826508cb8266a62c9191043ac43941cd5d9aaee6f3a610bf87  onnx/t3_cond.onnx
+cfb2662b33ef142730d3758abd5f9cf205c27abacfbf8c1039719b451ac6a480  onnx/t3_prefill.onnx
+4d4196db5924b5a86f1700fe9725cc790c6700e448179853638c9b6afa55d1c0  onnx/t3_step.onnx
 126fee01af55a2cb03317a54e2ceb169b2a7fb149715aa1ddf035a86cca0340e  onnx/flow_encoder.onnx
 e52c19135f8eef43d00706b32e94a1c6288f1cdb82d4aa1ed8ef4a749fd8d65e  onnx/flow_estimator.onnx
 fd15fbf1b8b086ae341c8f5fc205de2fabf66e409d54fdd42033044a3e130c76  onnx/vocoder.onnx
 71a34a40f7bb1f0a8755b946d8bf99fc7a08185989fe6cf20d2b99dab3c74eb9  onnx/s3_tokenizer.onnx
 5b55a3246ef82c9822908393d7b4fc2f982bca3797a079b29365a8f18f1b13a7  onnx/camp.onnx
 8bc9882c7fb910d3e1c2cdc5857a76e2513ff4a3dceb26484de795deff85a9cc  onnx/voice_encoder.onnx
-9d8ed261bdbb73905630575d54ef07c26b79037d5b88bb3978be7f51d7a8eb20  coreml/flow_encoder.mlpackage/Data/com.apple.CoreML/model.mlmodel
-88fa83fe159807f73163624f0f96c989873c9c5207b47530a1cd621f5d29e431  coreml/flow_encoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin
-f5eac8059836bc8273e3dccb7d549a608cba1465de11523d0eec09262160e75c  coreml/flow_encoder.mlpackage/Manifest.json
-eaef922d6b36456608415da273b8a9340bdfc125a55d83191f0e7344deeecd53  coreml/flow_estimator.mlpackage/Data/com.apple.CoreML/model.mlmodel
+d2b8a658d4c04aae8c137b6800175e8abdee942df208c1c5e1909d81298c63e8  onnx/export.json
+792450b7f6f1584185562028877ca3c0d03f7a3cc2ca0a63b79f235e7f767925  coreml/t3_cond.mlpackage/Data/com.apple.CoreML/model.mlmodel
+879aaf6c265a29e1b65c0faba74be3187dbc8b6988dacf5a734983416b129679  coreml/t3_cond.mlpackage/Data/com.apple.CoreML/weights/weight.bin
+672b53d8542048b96378c307110c5bcec9483183f41209c91ec837c07be4f914  coreml/t3_cond.mlpackage/Manifest.json
+668a313f62756e86ab70f269091b497ef5be32f7775dec50cce55ce0dda24798  coreml/t3_prefill.mlpackage/Data/com.apple.CoreML/model.mlmodel
+4439a51cbee4c8689a079001b736a47c2c309979c3b381ef05a2787cffb89ab4  coreml/t3_prefill.mlpackage/Data/com.apple.CoreML/weights/weight.bin
+1fcc19f4761793bb888aaa0a4571f8f3e95f8cc1eadfcf8d03f5905ee4843896  coreml/t3_prefill.mlpackage/Manifest.json
+888eae725a1b3302ab759e0e4195998b56f4cc0cf7ff030c6839530e37321252  coreml/t3_step.mlpackage/Data/com.apple.CoreML/model.mlmodel
+4439a51cbee4c8689a079001b736a47c2c309979c3b381ef05a2787cffb89ab4  coreml/t3_step.mlpackage/Data/com.apple.CoreML/weights/weight.bin
+00ed8bd0fa1523869670e2a0b3ff228dd51c3f4e00ce929af3f23d8cda212c15  coreml/t3_step.mlpackage/Manifest.json
+2f3ef2ce57c2e08c7c149f67191cdf4dfad5cb93273c5e6915533e5eb56121ad  coreml/flow_encoder.mlpackage/Data/com.apple.CoreML/model.mlmodel
+894101bf0dc091056e6c1f8cd3b202f6419baff023a242dc538fe2283ab77a32  coreml/flow_encoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin
+6704fb3cdc6e03eb010e18c45275c49f5d0137bd0d0af1c076fe1acc91ecc555  coreml/flow_encoder.mlpackage/Manifest.json
+ce06ba5d598a4d3608f527d4ee75d880da3c27d13774653c47dab206aae70641  coreml/flow_estimator.mlpackage/Data/com.apple.CoreML/model.mlmodel
 64323e966820fe770fa80a3aa92dad4643c0f6f6562577a59a2519fc2b58a847  coreml/flow_estimator.mlpackage/Data/com.apple.CoreML/weights/weight.bin
-4e1822f6d9670220307790c2672cc5a9c394cbf7bf36f4a68a521abd80bc48e9  coreml/flow_estimator.mlpackage/Manifest.json
-96c9e25573c766284c21c24b73fbe63483f75035154bc814012474e06c06f5d7  coreml/vocoder.mlpackage/Data/com.apple.CoreML/model.mlmodel
+96a1845725528f70f2ddb0214b0ff6c2ca8e081b73b4e5e1ac179553594ff23e  coreml/flow_estimator.mlpackage/Manifest.json
+eec74247d893c15c2a057fe600c074cadb7180c97d9bde7526f2e97458a412c4  coreml/vocoder.mlpackage/Data/com.apple.CoreML/model.mlmodel
 9bb3248f8940ec2510545a807e07ea023b160c3ee9501b3df42381da9813bb50  coreml/vocoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin
-b692d43de35ee92e3dc58689c493be74b543f9a82c5bb7c9362cadd5af739db7  coreml/vocoder.mlpackage/Manifest.json
-dd09cab62547c1ede5902996a62fefff719a84cdeae94073a56bc175c57db33e  coreml/s3_tokenizer.mlpackage/Data/com.apple.CoreML/model.mlmodel
+ed3737db61e51a07c82e22acf6645451bcef3621cee79bba3529dbb8b3077574  coreml/vocoder.mlpackage/Manifest.json
+070b0c37bcae77f3ca9c79df67a683a96caec8fefb1d7ffefe182f5914da088e  coreml/s3_tokenizer.mlpackage/Data/com.apple.CoreML/model.mlmodel
 7998122d4a9fabf8cd069abf7612001b4f7220dad3356aa989289bd5d02d16a2  coreml/s3_tokenizer.mlpackage/Data/com.apple.CoreML/weights/weight.bin
-5614dd99eca2488b007f2e5607a530a889e2b376ffb53b81b4f7a09f454369ad  coreml/s3_tokenizer.mlpackage/Manifest.json
-1afdb0652ccf239182b765700b58b6b29c8e8383c2e1dbf040149d0c43a385a9  coreml/camp.mlpackage/Data/com.apple.CoreML/model.mlmodel
+355528c9265320d5e21c58f98864089a51cf41b8af74a9faec0d43f8fa23fefe  coreml/s3_tokenizer.mlpackage/Manifest.json
+0b90452c9aa15806a87295c2ab677fadbf84ed0a65da5e514f27904781e55df3  coreml/camp.mlpackage/Data/com.apple.CoreML/model.mlmodel
 3e5bafbc3d3faa94aff8c72afd22ffae97ce31d87ce70c000585e055cd776c18  coreml/camp.mlpackage/Data/com.apple.CoreML/weights/weight.bin
-e7f2f756511fe047d0cd1997cdf2d42ba5c0c9fc8273c4450d25587e9e7aa77e  coreml/camp.mlpackage/Manifest.json
-759103649d9bf8d388b523b915e1eb9b285fb81e58fdef6d420d4c7e21053765  coreml/voice_encoder.mlpackage/Data/com.apple.CoreML/model.mlmodel
+3c90135609a909f5262d0660c3b4b7e541eb002dd14af91c4f50949c90b56ed4  coreml/camp.mlpackage/Manifest.json
+e3f9d7f3b91d4d733a3993750bd70d3c2eeb83560b99a1e7892335c6d785b572  coreml/voice_encoder.mlpackage/Data/com.apple.CoreML/model.mlmodel
 95b1a8b8c6b558fafee4a1c784076c8e099070c45958526c9a187c15de55d2d3  coreml/voice_encoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin
-fbd810d98add4c3349f01891e1b7c6dc975070bfa83e78f8f8a115f5dcea3c55  coreml/voice_encoder.mlpackage/Manifest.json
-8a75580837b18451c432b17e5f4769507a222cbe770b0591b1707b94c0e57622  README.md
+b75e665d6f2d08dbcb32bf8e9c2d6b1801dc69c7a9df72661ec103653327e65b  coreml/voice_encoder.mlpackage/Manifest.json
+87f5602b6a0fe370a2a5634bbce4f2ebf4ad92d1c4e116a720fb88cb1c165073  coreml/export.json
+aed4054edc684304874253720144bcde3dd2c37ea984548e69f87e7e8836398d  README.md
 5d5671b95e9797856175dd08d5d2f20d21d72e8785aba077ee38eb378d432528  LICENSE
-fda3657fa4d457f9dc379c1ca8550d665eb972ac1c7a260fbe94821ef418ff28  NOTICE
-87c35454a03b026a5a13093566ace245c339dc6af68c7f794f6ed322a731f753  RESPONSIBLE_USE.md
+95aba903ecbd8bda3c170c6b4b8979ec5762f9a18f5c3a3eab240f4a8a5f9417  NOTICE
+9dfbc1fedb3013764f2f0b214d8cadc0330411f1e67837ae2e0cd7df7b934d7b  RESPONSIBLE_USE.md
 f90d7077925cb35277579c93e57c701e3198df09b21830ccfb14f56c4c4507ea  logo.png
-83e0bf0a1f6b03e047de66871536c3ff77291041d0a3f2907b6990e580519fea  samples/joe.opus
-48615517c1d1d2d55dbb7f74f37982918236e33f079947459d56c116cd70b7a3  samples/kathleen.opus
-fb7ed0196f3175e3a4880085ca876c7e76d0f82a020a0e30c3966d292c68adf7  release.json
+29f05b9b182b500b701735c966276c40a383c34e2714f469393ca54b58036d62  samples/joe.opus
+d5f4c35c1b1c30aed341b88634fb88bf6c76f200e457e40a27934af5583202ae  samples/kathleen.opus
+c92dd567b6835f8e138221fcfef40d4e112ffedcd06b3c8e217e4bac2b8bb033  release.json

coreml/camp.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1afdb0652ccf239182b765700b58b6b29c8e8383c2e1dbf040149d0c43a385a9
-size 547281
+oid sha256:0b90452c9aa15806a87295c2ab677fadbf84ed0a65da5e514f27904781e55df3
+size 547283

coreml/camp.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "19A040E3-A54C-4972-B14B-A44D4B44473D": {
-            "author": "com.apple.CoreML",
-            "description": "CoreML Model Weights",
-            "name": "weights",
-            "path": "com.apple.CoreML/weights"
-        },
-        "9A6B03D4-A8CF-47B0-BD8F-80C7C5FA3D9B": {
+        "8DDA9D32-F77E-40D0-BB82-4A55A86660DB": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Specification",
             "name": "model.mlmodel",
             "path": "com.apple.CoreML/model.mlmodel"
+        },
+        "EBDA90D2-6EA2-4644-97A1-3E036CB92318": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Weights",
+            "name": "weights",
+            "path": "com.apple.CoreML/weights"
         }
     },
-    "rootModelIdentifier": "9A6B03D4-A8CF-47B0-BD8F-80C7C5FA3D9B"
+    "rootModelIdentifier": "8DDA9D32-F77E-40D0-BB82-4A55A86660DB"
 }

coreml/export.json

@@ -0,0 +1,155 @@
+{
+  "format": "loudkit-coreml-export",
+  "packages": {
+    "flow_encoder.mlpackage": {
+      "algorithm_fingerprint": "e352f5d090776b54",
+      "checkpoint": "loudr-1.safetensors",
+      "checkpoint_sha256": "73e69a78d58176a3e4b39ca992c7b37a9c51f7251d08e4334c27d9994faac44e",
+      "decode": "single",
+      "euler_steps": 2,
+      "graph_set": [
+        "t3_cond.mlpackage",
+        "t3_prefill.mlpackage",
+        "t3_step.mlpackage",
+        "flow_encoder.mlpackage",
+        "flow_estimator.mlpackage",
+        "vocoder.mlpackage"
+      ],
+      "precision": "fp32",
+      "toolchain": {
+        "coremltools": "9.0",
+        "numpy": "2.3.5",
+        "onnx": "1.22.0",
+        "onnxruntime": "1.29.0",
+        "platform": "macOS-26.1-arm64-arm-64bit",
+        "python": "3.12.12",
+        "torch": "2.13.0"
+      }
+    },
+    "flow_estimator.mlpackage": {
+      "algorithm_fingerprint": "e352f5d090776b54",
+      "checkpoint": "loudr-1.safetensors",
+      "checkpoint_sha256": "73e69a78d58176a3e4b39ca992c7b37a9c51f7251d08e4334c27d9994faac44e",
+      "decode": "single",
+      "euler_steps": 2,
+      "graph_set": [
+        "t3_cond.mlpackage",
+        "t3_prefill.mlpackage",
+        "t3_step.mlpackage",
+        "flow_encoder.mlpackage",
+        "flow_estimator.mlpackage",
+        "vocoder.mlpackage"
+      ],
+      "precision": "fp16",
+      "toolchain": {
+        "coremltools": "9.0",
+        "numpy": "2.3.5",
+        "onnx": "1.22.0",
+        "onnxruntime": "1.29.0",
+        "platform": "macOS-26.1-arm64-arm-64bit",
+        "python": "3.12.12",
+        "torch": "2.13.0"
+      }
+    },
+    "t3_cond.mlpackage": {
+      "algorithm_fingerprint": "e352f5d090776b54",
+      "checkpoint": "loudr-1.safetensors",
+      "checkpoint_sha256": "73e69a78d58176a3e4b39ca992c7b37a9c51f7251d08e4334c27d9994faac44e",
+      "decode": "single",
+      "euler_steps": 2,
+      "graph_set": [
+        "t3_cond.mlpackage",
+        "t3_prefill.mlpackage",
+        "t3_step.mlpackage",
+        "flow_encoder.mlpackage",
+        "flow_estimator.mlpackage",
+        "vocoder.mlpackage"
+      ],
+      "precision": "fp32",
+      "toolchain": {
+        "coremltools": "9.0",
+        "numpy": "2.3.5",
+        "onnx": "1.22.0",
+        "onnxruntime": "1.29.0",
+        "platform": "macOS-26.1-arm64-arm-64bit",
+        "python": "3.12.12",
+        "torch": "2.13.0"
+      }
+    },
+    "t3_prefill.mlpackage": {
+      "algorithm_fingerprint": "e352f5d090776b54",
+      "checkpoint": "loudr-1.safetensors",
+      "checkpoint_sha256": "73e69a78d58176a3e4b39ca992c7b37a9c51f7251d08e4334c27d9994faac44e",
+      "decode": "single",
+      "euler_steps": 2,
+      "graph_set": [
+        "t3_cond.mlpackage",
+        "t3_prefill.mlpackage",
+        "t3_step.mlpackage",
+        "flow_encoder.mlpackage",
+        "flow_estimator.mlpackage",
+        "vocoder.mlpackage"
+      ],
+      "precision": "fp32",
+      "toolchain": {
+        "coremltools": "9.0",
+        "numpy": "2.3.5",
+        "onnx": "1.22.0",
+        "onnxruntime": "1.29.0",
+        "platform": "macOS-26.1-arm64-arm-64bit",
+        "python": "3.12.12",
+        "torch": "2.13.0"
+      }
+    },
+    "t3_step.mlpackage": {
+      "algorithm_fingerprint": "e352f5d090776b54",
+      "checkpoint": "loudr-1.safetensors",
+      "checkpoint_sha256": "73e69a78d58176a3e4b39ca992c7b37a9c51f7251d08e4334c27d9994faac44e",
+      "decode": "single",
+      "euler_steps": 2,
+      "graph_set": [
+        "t3_cond.mlpackage",
+        "t3_prefill.mlpackage",
+        "t3_step.mlpackage",
+        "flow_encoder.mlpackage",
+        "flow_estimator.mlpackage",
+        "vocoder.mlpackage"
+      ],
+      "precision": "fp32",
+      "toolchain": {
+        "coremltools": "9.0",
+        "numpy": "2.3.5",
+        "onnx": "1.22.0",
+        "onnxruntime": "1.29.0",
+        "platform": "macOS-26.1-arm64-arm-64bit",
+        "python": "3.12.12",
+        "torch": "2.13.0"
+      }
+    },
+    "vocoder.mlpackage": {
+      "algorithm_fingerprint": "e352f5d090776b54",
+      "checkpoint": "loudr-1.safetensors",
+      "checkpoint_sha256": "73e69a78d58176a3e4b39ca992c7b37a9c51f7251d08e4334c27d9994faac44e",
+      "decode": "single",
+      "euler_steps": 2,
+      "graph_set": [
+        "t3_cond.mlpackage",
+        "t3_prefill.mlpackage",
+        "t3_step.mlpackage",
+        "flow_encoder.mlpackage",
+        "flow_estimator.mlpackage",
+        "vocoder.mlpackage"
+      ],
+      "precision": "fp32",
+      "toolchain": {
+        "coremltools": "9.0",
+        "numpy": "2.3.5",
+        "onnx": "1.22.0",
+        "onnxruntime": "1.29.0",
+        "platform": "macOS-26.1-arm64-arm-64bit",
+        "python": "3.12.12",
+        "torch": "2.13.0"
+      }
+    }
+  }
+}

coreml/flow_encoder.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9d8ed261bdbb73905630575d54ef07c26b79037d5b88bb3978be7f51d7a8eb20
-size 139670
+oid sha256:2f3ef2ce57c2e08c7c149f67191cdf4dfad5cb93273c5e6915533e5eb56121ad
+size 139672

coreml/flow_encoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:88fa83fe159807f73163624f0f96c989873c9c5207b47530a1cd621f5d29e431
+oid sha256:894101bf0dc091056e6c1f8cd3b202f6419baff023a242dc538fe2283ab77a32
 size 182751040

coreml/flow_encoder.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "4C4CF55C-5F4D-4BDA-A898-1DCC9883BF54": {
-            "author": "com.apple.CoreML",
-            "description": "CoreML Model Specification",
-            "name": "model.mlmodel",
-            "path": "com.apple.CoreML/model.mlmodel"
-        },
-        "5EC0CD40-1483-4B58-BB3E-C53418D95E6E": {
+        "D2077815-8937-470C-B62A-AF4B31695141": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Weights",
             "name": "weights",
             "path": "com.apple.CoreML/weights"
+        },
+        "D6D89F7B-2F17-47C0-9079-138169DCBC2A": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Specification",
+            "name": "model.mlmodel",
+            "path": "com.apple.CoreML/model.mlmodel"
         }
     },
-    "rootModelIdentifier": "4C4CF55C-5F4D-4BDA-A898-1DCC9883BF54"
+    "rootModelIdentifier": "D6D89F7B-2F17-47C0-9079-138169DCBC2A"
 }

coreml/flow_estimator.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:eaef922d6b36456608415da273b8a9340bdfc125a55d83191f0e7344deeecd53
-size 686448
+oid sha256:ce06ba5d598a4d3608f527d4ee75d880da3c27d13774653c47dab206aae70641
+size 686450

coreml/flow_estimator.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "C000A808-BC9B-4914-9ABB-B7070B8D9FBC": {
-            "author": "com.apple.CoreML",
-            "description": "CoreML Model Weights",
-            "name": "weights",
-            "path": "com.apple.CoreML/weights"
-        },
-        "D50C51D0-4869-4187-9591-F692AF38B90F": {
+        "29F4AB05-B5B9-47B8-81D8-283A29241E17": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Specification",
             "name": "model.mlmodel",
             "path": "com.apple.CoreML/model.mlmodel"
+        },
+        "B8ABD9C8-242A-47E1-BEF7-88E09B02A5A7": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Weights",
+            "name": "weights",
+            "path": "com.apple.CoreML/weights"
         }
     },
-    "rootModelIdentifier": "D50C51D0-4869-4187-9591-F692AF38B90F"
+    "rootModelIdentifier": "29F4AB05-B5B9-47B8-81D8-283A29241E17"
 }

coreml/s3_tokenizer.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dd09cab62547c1ede5902996a62fefff719a84cdeae94073a56bc175c57db33e
-size 109304
+oid sha256:070b0c37bcae77f3ca9c79df67a683a96caec8fefb1d7ffefe182f5914da088e
+size 109306

coreml/s3_tokenizer.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "95089144-3266-40CD-AD87-FAF5995D15B1": {
-            "author": "com.apple.CoreML",
-            "description": "CoreML Model Weights",
-            "name": "weights",
-            "path": "com.apple.CoreML/weights"
-        },
-        "C0C80A3E-2EDA-4ACD-A752-A8CC0EDE1CFE": {
+        "3584EE4D-1925-43FD-923F-85983E77BD60": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Specification",
             "name": "model.mlmodel",
             "path": "com.apple.CoreML/model.mlmodel"
+        },
+        "3A6CAA22-779E-47BF-A62A-0CCC3A285970": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Weights",
+            "name": "weights",
+            "path": "com.apple.CoreML/weights"
         }
     },
-    "rootModelIdentifier": "C0C80A3E-2EDA-4ACD-A752-A8CC0EDE1CFE"
+    "rootModelIdentifier": "3584EE4D-1925-43FD-923F-85983E77BD60"
 }

coreml/t3_cond.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:792450b7f6f1584185562028877ca3c0d03f7a3cc2ca0a63b79f235e7f767925
+size 17922

coreml/t3_cond.mlpackage/Data/com.apple.CoreML/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:879aaf6c265a29e1b65c0faba74be3187dbc8b6988dacf5a734983416b129679
+size 68482240

coreml/t3_cond.mlpackage/Manifest.json

@@ -0,0 +1,18 @@
+{
+    "fileFormatVersion": "1.0.0",
+    "itemInfoEntries": {
+        "93C25745-31B7-473F-A1BB-34FBF5369970": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Weights",
+            "name": "weights",
+            "path": "com.apple.CoreML/weights"
+        },
+        "D688ADC9-D218-464C-A1D0-2349BF6C387D": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Specification",
+            "name": "model.mlmodel",
+            "path": "com.apple.CoreML/model.mlmodel"
+        }
+    },
+    "rootModelIdentifier": "D688ADC9-D218-464C-A1D0-2349BF6C387D"
+}

coreml/t3_prefill.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:668a313f62756e86ab70f269091b497ef5be32f7775dec50cce55ce0dda24798
+size 361088

coreml/t3_prefill.mlpackage/Data/com.apple.CoreML/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4439a51cbee4c8689a079001b736a47c2c309979c3b381ef05a2787cffb89ab4
+size 614402952

coreml/t3_prefill.mlpackage/Manifest.json

@@ -0,0 +1,18 @@
+{
+    "fileFormatVersion": "1.0.0",
+    "itemInfoEntries": {
+        "6106A0B3-88C1-42A1-ADE9-3BC5509864DF": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Specification",
+            "name": "model.mlmodel",
+            "path": "com.apple.CoreML/model.mlmodel"
+        },
+        "63EEAFFC-82F3-43CF-BA3D-0033B379A746": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Weights",
+            "name": "weights",
+            "path": "com.apple.CoreML/weights"
+        }
+    },
+    "rootModelIdentifier": "6106A0B3-88C1-42A1-ADE9-3BC5509864DF"
+}

coreml/t3_step.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:888eae725a1b3302ab759e0e4195998b56f4cc0cf7ff030c6839530e37321252
+size 303478

coreml/t3_step.mlpackage/Data/com.apple.CoreML/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4439a51cbee4c8689a079001b736a47c2c309979c3b381ef05a2787cffb89ab4
+size 614402952

coreml/t3_step.mlpackage/Manifest.json

@@ -0,0 +1,18 @@
+{
+    "fileFormatVersion": "1.0.0",
+    "itemInfoEntries": {
+        "9A4E8CA9-5C8F-4615-9F62-2203438EDBA9": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Specification",
+            "name": "model.mlmodel",
+            "path": "com.apple.CoreML/model.mlmodel"
+        },
+        "FA4EFBD7-B70F-47AF-A9C9-CA0A1D707BB4": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Weights",
+            "name": "weights",
+            "path": "com.apple.CoreML/weights"
+        }
+    },
+    "rootModelIdentifier": "9A4E8CA9-5C8F-4615-9F62-2203438EDBA9"
+}

coreml/vocoder.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:96c9e25573c766284c21c24b73fbe63483f75035154bc814012474e06c06f5d7
-size 205318
+oid sha256:eec74247d893c15c2a057fe600c074cadb7180c97d9bde7526f2e97458a412c4
+size 205320

coreml/vocoder.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "C6DE14E5-E5DB-4A68-A090-30575FD94414": {
+        "0B49AFA8-B13D-4BB2-BFA6-B278ECA43D91": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Weights",
             "name": "weights",
             "path": "com.apple.CoreML/weights"
         },
-        "D99CB598-3271-48C4-A589-09994F1FF116": {
+        "A284EC56-D0FD-44CF-9AFB-30AEF9AA98AE": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Specification",
             "name": "model.mlmodel",
             "path": "com.apple.CoreML/model.mlmodel"
         }
     },
-    "rootModelIdentifier": "D99CB598-3271-48C4-A589-09994F1FF116"
+    "rootModelIdentifier": "A284EC56-D0FD-44CF-9AFB-30AEF9AA98AE"
 }

coreml/voice_encoder.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:759103649d9bf8d388b523b915e1eb9b285fb81e58fdef6d420d4c7e21053765
-size 12973
+oid sha256:e3f9d7f3b91d4d733a3993750bd70d3c2eeb83560b99a1e7892335c6d785b572
+size 12975

coreml/voice_encoder.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "5FA30C01-DF2C-4B3F-BE13-2B3A2A7035CD": {
-            "author": "com.apple.CoreML",
-            "description": "CoreML Model Specification",
-            "name": "model.mlmodel",
-            "path": "com.apple.CoreML/model.mlmodel"
-        },
-        "F68BBC46-DC35-4A6C-BA2A-003FE8484988": {
+        "83CFAFD1-C769-4577-98D7-44449412FE41": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Weights",
             "name": "weights",
             "path": "com.apple.CoreML/weights"
+        },
+        "E96506D2-636B-43DE-81E5-9E6A754B268A": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Specification",
+            "name": "model.mlmodel",
+            "path": "com.apple.CoreML/model.mlmodel"
         }
     },
-    "rootModelIdentifier": "5FA30C01-DF2C-4B3F-BE13-2B3A2A7035CD"
+    "rootModelIdentifier": "E96506D2-636B-43DE-81E5-9E6A754B268A"
 }

loudr-1-enrollment.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dc825dcdec843cc0fb40ad3a0f051e869db97689a33712d1d11c009ce16af022
-size 522832312
+oid sha256:ac7f4a0ce1bf16855f6d1c5fe048bbfb90b6a114d74bc2cd85974d11d02d0d83
+size 522833520

loudr-1.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:688e87cddfc95a37b6705bda05ad95a7ddb219a4c0c232fa4b93ccadc42efe2c
-size 746630956
+oid sha256:73e69a78d58176a3e4b39ca992c7b37a9c51f7251d08e4334c27d9994faac44e
+size 746632188

manifest.json

@@ -24,6 +24,7 @@
   "s3gen.mel2wav": "EXP-003: fp16 produces audible ~12 kHz tone (NSF cumulative phase accumulator). Non-negotiable fp32.",
   "t3": "EXP-010: fp16 teacher-forced median KL 1.29e-06 vs fp32, top-1 99.9%; bf16 rejected (no ANE datapath)"
  },
+ "edge_fade_seconds": 0.02,
  "eos_floor": {
   "min_tokens_floor": 10,
   "min_tokens_text_ratio": 1.2
@@ -86,6 +87,43 @@
   "trailing_filler_threshold": 0.7,
   "trailing_silence_run_tokens": 12
  },
+ "quiet_render_ids": [
+  1458,
+  1461,
+  1488,
+  1701,
+  1704,
+  1707,
+  1716,
+  1731,
+  1785,
+  1788,
+  1869,
+  1947,
+  1950,
+  1951,
+  1959,
+  1978,
+  2028,
+  2031,
+  2040,
+  2058,
+  2076,
+  2112,
+  2139,
+  3645,
+  3648,
+  3651,
+  3704,
+  3888,
+  3894,
+  4188,
+  5838,
+  6081,
+  6183,
+  6537
+ ],
+ "quiet_render_ids_source": "per-instance RMS attribution (960 samples/token, threshold 0.012, >= 90% of instances quiet, >= 5 sightings) on nils+dave+joe anomalous chunks plus a healthy sample, minus the silence census: contextually silent breath/decay ids. A dead-air run continues across these but they never count toward its gate",
  "recipe_version": "loudkit-1",
  "runtime_requirements": {
   "allow_tf32_cudnn": false,
@@ -102,6 +140,17 @@
   "repetition_penalty": 1.2,
   "temperature": 0.8
  },
+ "silence_render_ids": [
+  4137,
+  4215,
+  4218,
+  4299,
+  6162,
+  6324,
+  6405,
+  6486
+ ],
+ "silence_render_ids_source": "union of two independent render censuses: (1) coordinator census, per-id median rendered energy < -80 dBFS, seen >= 8, 11479 tokens, 4 voices (joe kathleen thorsten colette), exact 0.04 s/token alignment; (2) interior-stall census over trap-step traces. Both agreed on 4137/4218/4299/6324/6405/6486; 4215 and 6162 each appeared in one census only and are included. Read by the postprocess stall rule; NOT a sampler exemption list (widening the min_p exemption to exactly these ids was measured harmful: pause-time share doubles)",
  "silence_token_ids": [
   1731,
   1821,

onnx/export.json

@@ -0,0 +1,149 @@
+{
+  "format": "loudkit-onnx-export",
+  "graphs": {
+    "flow_encoder.onnx": {
+      "algorithm_fingerprint": "e352f5d090776b54",
+      "checkpoint": "loudr-1.safetensors",
+      "checkpoint_sha256": "73e69a78d58176a3e4b39ca992c7b37a9c51f7251d08e4334c27d9994faac44e",
+      "decode": "single",
+      "euler_steps": 2,
+      "graph_set": [
+        "t3_cond.onnx",
+        "t3_prefill.onnx",
+        "t3_step.onnx",
+        "flow_encoder.onnx",
+        "flow_estimator.onnx",
+        "vocoder.onnx"
+      ],
+      "toolchain": {
+        "coremltools": "9.0",
+        "numpy": "2.3.5",
+        "onnx": "1.22.0",
+        "onnxruntime": "1.29.0",
+        "platform": "macOS-26.1-arm64-arm-64bit",
+        "python": "3.12.12",
+        "torch": "2.13.0"
+      }
+    },
+    "flow_estimator.onnx": {
+      "algorithm_fingerprint": "e352f5d090776b54",
+      "checkpoint": "loudr-1.safetensors",
+      "checkpoint_sha256": "73e69a78d58176a3e4b39ca992c7b37a9c51f7251d08e4334c27d9994faac44e",
+      "decode": "single",
+      "euler_steps": 2,
+      "graph_set": [
+        "t3_cond.onnx",
+        "t3_prefill.onnx",
+        "t3_step.onnx",
+        "flow_encoder.onnx",
+        "flow_estimator.onnx",
+        "vocoder.onnx"
+      ],
+      "toolchain": {
+        "coremltools": "9.0",
+        "numpy": "2.3.5",
+        "onnx": "1.22.0",
+        "onnxruntime": "1.29.0",
+        "platform": "macOS-26.1-arm64-arm-64bit",
+        "python": "3.12.12",
+        "torch": "2.13.0"
+      }
+    },
+    "t3_cond.onnx": {
+      "algorithm_fingerprint": "e352f5d090776b54",
+      "checkpoint": "loudr-1.safetensors",
+      "checkpoint_sha256": "73e69a78d58176a3e4b39ca992c7b37a9c51f7251d08e4334c27d9994faac44e",
+      "decode": "single",
+      "euler_steps": 2,
+      "graph_set": [
+        "t3_cond.onnx",
+        "t3_prefill.onnx",
+        "t3_step.onnx",
+        "flow_encoder.onnx",
+        "flow_estimator.onnx",
+        "vocoder.onnx"
+      ],
+      "toolchain": {
+        "coremltools": "9.0",
+        "numpy": "2.3.5",
+        "onnx": "1.22.0",
+        "onnxruntime": "1.29.0",
+        "platform": "macOS-26.1-arm64-arm-64bit",
+        "python": "3.12.12",
+        "torch": "2.13.0"
+      }
+    },
+    "t3_prefill.onnx": {
+      "algorithm_fingerprint": "e352f5d090776b54",
+      "checkpoint": "loudr-1.safetensors",
+      "checkpoint_sha256": "73e69a78d58176a3e4b39ca992c7b37a9c51f7251d08e4334c27d9994faac44e",
+      "decode": "single",
+      "euler_steps": 2,
+      "graph_set": [
+        "t3_cond.onnx",
+        "t3_prefill.onnx",
+        "t3_step.onnx",
+        "flow_encoder.onnx",
+        "flow_estimator.onnx",
+        "vocoder.onnx"
+      ],
+      "toolchain": {
+        "coremltools": "9.0",
+        "numpy": "2.3.5",
+        "onnx": "1.22.0",
+        "onnxruntime": "1.29.0",
+        "platform": "macOS-26.1-arm64-arm-64bit",
+        "python": "3.12.12",
+        "torch": "2.13.0"
+      }
+    },
+    "t3_step.onnx": {
+      "algorithm_fingerprint": "e352f5d090776b54",
+      "checkpoint": "loudr-1.safetensors",
+      "checkpoint_sha256": "73e69a78d58176a3e4b39ca992c7b37a9c51f7251d08e4334c27d9994faac44e",
+      "decode": "single",
+      "euler_steps": 2,
+      "graph_set": [
+        "t3_cond.onnx",
+        "t3_prefill.onnx",
+        "t3_step.onnx",
+        "flow_encoder.onnx",
+        "flow_estimator.onnx",
+        "vocoder.onnx"
+      ],
+      "toolchain": {
+        "coremltools": "9.0",
+        "numpy": "2.3.5",
+        "onnx": "1.22.0",
+        "onnxruntime": "1.29.0",
+        "platform": "macOS-26.1-arm64-arm-64bit",
+        "python": "3.12.12",
+        "torch": "2.13.0"
+      }
+    },
+    "vocoder.onnx": {
+      "algorithm_fingerprint": "e352f5d090776b54",
+      "checkpoint": "loudr-1.safetensors",
+      "checkpoint_sha256": "73e69a78d58176a3e4b39ca992c7b37a9c51f7251d08e4334c27d9994faac44e",
+      "decode": "single",
+      "euler_steps": 2,
+      "graph_set": [
+        "t3_cond.onnx",
+        "t3_prefill.onnx",
+        "t3_step.onnx",
+        "flow_encoder.onnx",
+        "flow_estimator.onnx",
+        "vocoder.onnx"
+      ],
+      "toolchain": {
+        "coremltools": "9.0",
+        "numpy": "2.3.5",
+        "onnx": "1.22.0",
+        "onnxruntime": "1.29.0",
+        "platform": "macOS-26.1-arm64-arm-64bit",
+        "python": "3.12.12",
+        "torch": "2.13.0"
+      }
+    }
+  }
+}

onnx/t3_cond.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7de8b5afaa4b352673508768de757195fc7fda340efdb6f56a4a48d663df6a89
-size 68371968
+oid sha256:483988e5adac3f826508cb8266a62c9191043ac43941cd5d9aaee6f3a610bf87
+size 68368775

onnx/t3_prefill.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fdbb5dfe5e0651adf6961a62b9adea45c90bef5849a05b4968bc00c923368766
-size 614814483
+oid sha256:cfb2662b33ef142730d3758abd5f9cf205c27abacfbf8c1039719b451ac6a480
+size 614814404

onnx/t3_step.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e8c39bc5512dd3fbac441ce5322673535ecbe281333bece33ea88330a348ca03
-size 614786516
+oid sha256:4d4196db5924b5a86f1700fe9725cc790c6700e448179853638c9b6afa55d1c0
+size 614779546

release.json

@@ -2,18 +2,18 @@
  "profile": "full-0.1",
  "checkpoint": {
   "path": "loudr-1.safetensors",
-  "sha256": "688e87cddfc95a37b6705bda05ad95a7ddb219a4c0c232fa4b93ccadc42efe2c",
-  "bytes": 746630956
+  "sha256": "73e69a78d58176a3e4b39ca992c7b37a9c51f7251d08e4334c27d9994faac44e",
+  "bytes": 746632188
  },
  "enrollment_checkpoint": {
   "path": "loudr-1-enrollment.safetensors",
-  "sha256": "dc825dcdec843cc0fb40ad3a0f051e869db97689a33712d1d11c009ce16af022",
-  "bytes": 522832312
+  "sha256": "ac7f4a0ce1bf16855f6d1c5fe048bbfb90b6a114d74bc2cd85974d11d02d0d83",
+  "bytes": 522833520
  },
  "manifest": {
   "path": "manifest.json",
-  "sha256": "795d019797197a0f01c722af516ac6521cda59c6a27d8d07828489175659a99f",
-  "bytes": 5635
+  "sha256": "9fde37e0dd43150725ea09bf15711602f2a9dfffca7a82b22aeeefbbd3c8f3bf",
+  "bytes": 6951
  },
  "tokenizer": {
   "path": "tokenizer.json",
@@ -125,23 +125,63 @@
    "path": "voices/freja.safetensors",
    "sha256": "0a3f913b9b47c529393f843850228bd8eee6a883b4c138978fd3afa7758561e8",
    "bytes": 145448
+  },
+  {
+   "path": "voices/oliver.safetensors",
+   "sha256": "bd2fdb4ff8fbc12b070654dc5dae82b24aee4990bda726185547ec613b9f9cb8",
+   "bytes": 163600
+  },
+  {
+   "path": "voices/henry.safetensors",
+   "sha256": "5476670162065397f01fca60a1944eb75ddef8dc7e6792dde7fb603c76c14d8c",
+   "bytes": 159712
+  },
+  {
+   "path": "voices/miles.safetensors",
+   "sha256": "184d5552881d2e6fd644f0546d9b470130664d84faf39a31979001cafe8f4ed9",
+   "bytes": 159064
+  },
+  {
+   "path": "voices/oscar.safetensors",
+   "sha256": "8b8ff60604338eedddcd7317ed76ad2109fba6fa0ab79c9684a7954b64f963b2",
+   "bytes": 165544
+  },
+  {
+   "path": "voices/clara.safetensors",
+   "sha256": "c2b5db4c210657d6b2b81602fae8e73a733ac4312f33e80536a59a874e7d865f",
+   "bytes": 159064
+  },
+  {
+   "path": "voices/emma.safetensors",
+   "sha256": "0f5d8bfaceca353ca92db751c48b6a1653823f421ead3c901efe6366b8fa2dfb",
+   "bytes": 154528
+  },
+  {
+   "path": "voices/lucy.safetensors",
+   "sha256": "009a5fe1b34a9c65b09254f41c01cfdc641ec671a83449ad8aa77dd933d5f27e",
+   "bytes": 163600
+  },
+  {
+   "path": "voices/sophie.safetensors",
+   "sha256": "efb8047f93eda2a8e7cedaaa254b51957a845d55287298868a409662f8f75e1a",
+   "bytes": 153232
   }
  ],
  "onnx": [
   {
    "path": "onnx/t3_cond.onnx",
-   "sha256": "7de8b5afaa4b352673508768de757195fc7fda340efdb6f56a4a48d663df6a89",
-   "bytes": 68371968
+   "sha256": "483988e5adac3f826508cb8266a62c9191043ac43941cd5d9aaee6f3a610bf87",
+   "bytes": 68368775
   },
   {
    "path": "onnx/t3_prefill.onnx",
-   "sha256": "fdbb5dfe5e0651adf6961a62b9adea45c90bef5849a05b4968bc00c923368766",
-   "bytes": 614814483
+   "sha256": "cfb2662b33ef142730d3758abd5f9cf205c27abacfbf8c1039719b451ac6a480",
+   "bytes": 614814404
   },
   {
    "path": "onnx/t3_step.onnx",
-   "sha256": "e8c39bc5512dd3fbac441ce5322673535ecbe281333bece33ea88330a348ca03",
-   "bytes": 614786516
+   "sha256": "4d4196db5924b5a86f1700fe9725cc790c6700e448179853638c9b6afa55d1c0",
+   "bytes": 614779546
   },
   {
    "path": "onnx/flow_encoder.onnx",
@@ -172,28 +212,78 @@
    "path": "onnx/voice_encoder.onnx",
    "sha256": "8bc9882c7fb910d3e1c2cdc5857a76e2513ff4a3dceb26484de795deff85a9cc",
    "bytes": 5700082
+  },
+  {
+   "path": "onnx/export.json",
+   "sha256": "d2b8a658d4c04aae8c137b6800175e8abdee942df208c1c5e1909d81298c63e8",
+   "bytes": 4301
   }
  ],
  "coreml": [
+  {
+   "path": "coreml/t3_cond.mlpackage/Data/com.apple.CoreML/model.mlmodel",
+   "sha256": "792450b7f6f1584185562028877ca3c0d03f7a3cc2ca0a63b79f235e7f767925",
+   "bytes": 17922
+  },
+  {
+   "path": "coreml/t3_cond.mlpackage/Data/com.apple.CoreML/weights/weight.bin",
+   "sha256": "879aaf6c265a29e1b65c0faba74be3187dbc8b6988dacf5a734983416b129679",
+   "bytes": 68482240
+  },
+  {
+   "path": "coreml/t3_cond.mlpackage/Manifest.json",
+   "sha256": "672b53d8542048b96378c307110c5bcec9483183f41209c91ec837c07be4f914",
+   "bytes": 617
+  },
+  {
+   "path": "coreml/t3_prefill.mlpackage/Data/com.apple.CoreML/model.mlmodel",
+   "sha256": "668a313f62756e86ab70f269091b497ef5be32f7775dec50cce55ce0dda24798",
+   "bytes": 361088
+  },
+  {
+   "path": "coreml/t3_prefill.mlpackage/Data/com.apple.CoreML/weights/weight.bin",
+   "sha256": "4439a51cbee4c8689a079001b736a47c2c309979c3b381ef05a2787cffb89ab4",
+   "bytes": 614402952
+  },
+  {
+   "path": "coreml/t3_prefill.mlpackage/Manifest.json",
+   "sha256": "1fcc19f4761793bb888aaa0a4571f8f3e95f8cc1eadfcf8d03f5905ee4843896",
+   "bytes": 617
+  },
+  {
+   "path": "coreml/t3_step.mlpackage/Data/com.apple.CoreML/model.mlmodel",
+   "sha256": "888eae725a1b3302ab759e0e4195998b56f4cc0cf7ff030c6839530e37321252",
+   "bytes": 303478
+  },
+  {
+   "path": "coreml/t3_step.mlpackage/Data/com.apple.CoreML/weights/weight.bin",
+   "sha256": "4439a51cbee4c8689a079001b736a47c2c309979c3b381ef05a2787cffb89ab4",
+   "bytes": 614402952
+  },
+  {
+   "path": "coreml/t3_step.mlpackage/Manifest.json",
+   "sha256": "00ed8bd0fa1523869670e2a0b3ff228dd51c3f4e00ce929af3f23d8cda212c15",
+   "bytes": 617
+  },
   {
    "path": "coreml/flow_encoder.mlpackage/Data/com.apple.CoreML/model.mlmodel",
-   "sha256": "9d8ed261bdbb73905630575d54ef07c26b79037d5b88bb3978be7f51d7a8eb20",
-   "bytes": 139670
+   "sha256": "2f3ef2ce57c2e08c7c149f67191cdf4dfad5cb93273c5e6915533e5eb56121ad",
+   "bytes": 139672
   },
   {
    "path": "coreml/flow_encoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin",
-   "sha256": "88fa83fe159807f73163624f0f96c989873c9c5207b47530a1cd621f5d29e431",
+   "sha256": "894101bf0dc091056e6c1f8cd3b202f6419baff023a242dc538fe2283ab77a32",
    "bytes": 182751040
   },
   {
    "path": "coreml/flow_encoder.mlpackage/Manifest.json",
-   "sha256": "f5eac8059836bc8273e3dccb7d549a608cba1465de11523d0eec09262160e75c",
+   "sha256": "6704fb3cdc6e03eb010e18c45275c49f5d0137bd0d0af1c076fe1acc91ecc555",
    "bytes": 617
   },
   {
    "path": "coreml/flow_estimator.mlpackage/Data/com.apple.CoreML/model.mlmodel",
-   "sha256": "eaef922d6b36456608415da273b8a9340bdfc125a55d83191f0e7344deeecd53",
-   "bytes": 686448
+   "sha256": "ce06ba5d598a4d3608f527d4ee75d880da3c27d13774653c47dab206aae70641",
+   "bytes": 686450
   },
   {
    "path": "coreml/flow_estimator.mlpackage/Data/com.apple.CoreML/weights/weight.bin",
@@ -202,13 +292,13 @@
   },
   {
    "path": "coreml/flow_estimator.mlpackage/Manifest.json",
-   "sha256": "4e1822f6d9670220307790c2672cc5a9c394cbf7bf36f4a68a521abd80bc48e9",
+   "sha256": "96a1845725528f70f2ddb0214b0ff6c2ca8e081b73b4e5e1ac179553594ff23e",
    "bytes": 617
   },
   {
    "path": "coreml/vocoder.mlpackage/Data/com.apple.CoreML/model.mlmodel",
-   "sha256": "96c9e25573c766284c21c24b73fbe63483f75035154bc814012474e06c06f5d7",
-   "bytes": 205318
+   "sha256": "eec74247d893c15c2a057fe600c074cadb7180c97d9bde7526f2e97458a412c4",
+   "bytes": 205320
   },
   {
    "path": "coreml/vocoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin",
@@ -217,13 +307,13 @@
   },
   {
    "path": "coreml/vocoder.mlpackage/Manifest.json",
-   "sha256": "b692d43de35ee92e3dc58689c493be74b543f9a82c5bb7c9362cadd5af739db7",
+   "sha256": "ed3737db61e51a07c82e22acf6645451bcef3621cee79bba3529dbb8b3077574",
    "bytes": 617
   },
   {
    "path": "coreml/s3_tokenizer.mlpackage/Data/com.apple.CoreML/model.mlmodel",
-   "sha256": "dd09cab62547c1ede5902996a62fefff719a84cdeae94073a56bc175c57db33e",
-   "bytes": 109304
+   "sha256": "070b0c37bcae77f3ca9c79df67a683a96caec8fefb1d7ffefe182f5914da088e",
+   "bytes": 109306
   },
   {
    "path": "coreml/s3_tokenizer.mlpackage/Data/com.apple.CoreML/weights/weight.bin",
@@ -232,13 +322,13 @@
   },
   {
    "path": "coreml/s3_tokenizer.mlpackage/Manifest.json",
-   "sha256": "5614dd99eca2488b007f2e5607a530a889e2b376ffb53b81b4f7a09f454369ad",
+   "sha256": "355528c9265320d5e21c58f98864089a51cf41b8af74a9faec0d43f8fa23fefe",
    "bytes": 617
   },
   {
    "path": "coreml/camp.mlpackage/Data/com.apple.CoreML/model.mlmodel",
-   "sha256": "1afdb0652ccf239182b765700b58b6b29c8e8383c2e1dbf040149d0c43a385a9",
-   "bytes": 547281
+   "sha256": "0b90452c9aa15806a87295c2ab677fadbf84ed0a65da5e514f27904781e55df3",
+   "bytes": 547283
   },
   {
    "path": "coreml/camp.mlpackage/Data/com.apple.CoreML/weights/weight.bin",
@@ -247,13 +337,13 @@
   },
   {
    "path": "coreml/camp.mlpackage/Manifest.json",
-   "sha256": "e7f2f756511fe047d0cd1997cdf2d42ba5c0c9fc8273c4450d25587e9e7aa77e",
+   "sha256": "3c90135609a909f5262d0660c3b4b7e541eb002dd14af91c4f50949c90b56ed4",
    "bytes": 617
   },
   {
    "path": "coreml/voice_encoder.mlpackage/Data/com.apple.CoreML/model.mlmodel",
-   "sha256": "759103649d9bf8d388b523b915e1eb9b285fb81e58fdef6d420d4c7e21053765",
-   "bytes": 12973
+   "sha256": "e3f9d7f3b91d4d733a3993750bd70d3c2eeb83560b99a1e7892335c6d785b572",
+   "bytes": 12975
   },
   {
    "path": "coreml/voice_encoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin",
@@ -262,14 +352,19 @@
   },
   {
    "path": "coreml/voice_encoder.mlpackage/Manifest.json",
-   "sha256": "fbd810d98add4c3349f01891e1b7c6dc975070bfa83e78f8f8a115f5dcea3c55",
+   "sha256": "b75e665d6f2d08dbcb32bf8e9c2d6b1801dc69c7a9df72661ec103653327e65b",
    "bytes": 617
+  },
+  {
+   "path": "coreml/export.json",
+   "sha256": "87f5602b6a0fe370a2a5634bbce4f2ebf4ad92d1c4e116a720fb88cb1c165073",
+   "bytes": 4677
   }
  ],
  "readme": {
   "path": "README.md",
-  "sha256": "8a75580837b18451c432b17e5f4769507a222cbe770b0591b1707b94c0e57622",
-  "bytes": 8689
+  "sha256": "aed4054edc684304874253720144bcde3dd2c37ea984548e69f87e7e8836398d",
+  "bytes": 8624
  },
  "license": {
   "path": "LICENSE",
@@ -278,13 +373,13 @@
  },
  "notice": {
   "path": "NOTICE",
-  "sha256": "fda3657fa4d457f9dc379c1ca8550d665eb972ac1c7a260fbe94821ef418ff28",
-  "bytes": 11732
+  "sha256": "95aba903ecbd8bda3c170c6b4b8979ec5762f9a18f5c3a3eab240f4a8a5f9417",
+  "bytes": 11846
  },
  "responsible_use": {
   "path": "RESPONSIBLE_USE.md",
-  "sha256": "87c35454a03b026a5a13093566ace245c339dc6af68c7f794f6ed322a731f753",
-  "bytes": 4022
+  "sha256": "9dfbc1fedb3013764f2f0b214d8cadc0330411f1e67837ae2e0cd7df7b934d7b",
+  "bytes": 4240
  },
  "logo": {
   "path": "logo.png",
@@ -294,13 +389,13 @@
  "samples": [
   {
    "path": "samples/joe.opus",
-   "sha256": "83e0bf0a1f6b03e047de66871536c3ff77291041d0a3f2907b6990e580519fea",
-   "bytes": 52187
+   "sha256": "29f05b9b182b500b701735c966276c40a383c34e2714f469393ca54b58036d62",
+   "bytes": 87475
   },
   {
    "path": "samples/kathleen.opus",
-   "sha256": "48615517c1d1d2d55dbb7f74f37982918236e33f079947459d56c116cd70b7a3",
-   "bytes": 55798
+   "sha256": "d5f4c35c1b1c30aed341b88634fb88bf6c76f200e457e40a27934af5583202ae",
+   "bytes": 80881
   }
  ],
  "verified": true

voices/clara.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2b5db4c210657d6b2b81602fae8e73a733ac4312f33e80536a59a874e7d865f
+size 159064

voices/emma.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f5d8bfaceca353ca92db751c48b6a1653823f421ead3c901efe6366b8fa2dfb
+size 154528

voices/henry.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5476670162065397f01fca60a1944eb75ddef8dc7e6792dde7fb603c76c14d8c
+size 159712

voices/lucy.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:009a5fe1b34a9c65b09254f41c01cfdc641ec671a83449ad8aa77dd933d5f27e
+size 163600

voices/miles.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:184d5552881d2e6fd644f0546d9b470130664d84faf39a31979001cafe8f4ed9
+size 159064

voices/oliver.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd2fdb4ff8fbc12b070654dc5dae82b24aee4990bda726185547ec613b9f9cb8
+size 163600

voices/oscar.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b8ff60604338eedddcd7317ed76ad2109fba6fa0ab79c9684a7954b64f963b2
+size 165544

voices/sophie.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:efb8047f93eda2a8e7cedaaa254b51957a845d55287298868a409662f8f75e1a
+size 153232