kevinqz/MOSS-Transcribe-Diarize-Audio-CoreAI

Canonical: kevinqz/MOSS-Transcribe-Diarize-Audio-CoreAI — source of truth.

MOSS-Transcribe-Diarize Audio Encoder (fabric)

An Apple Core AI conversion of OpenMOSS-Team/MOSS-Transcribe-Diarize — the audio encoder of an automatic-speech-recognition + diarization model, mapping a log-mel spectrogram to encoder hidden states. Produced by coreai-fabric and indexed by coreai-catalog.

Encoder only — not the full ASR pipeline. This asset is ONLY the audio encoder (log-mel → encoder hidden states). The host owns mel-spectrogram extraction (upstream feature config), the autoregressive decoder (published separately), and any speaker-diarization post-processing. It does not, by itself, transcribe audio.

Model facts

Field	Value
Parameters	0.9B
Architecture	transformer
Capabilities	speech-to-text
Input (log-mel)	1×80×3000
Output (encoder states)	1×375×1024
Quantization / precision	none / float32
On-disk size	1.2 GB
Asset kind	single-graph audio encoder (log-mel -> encoder hidden states)
assetVersion	2.0

Use it — this needs host code you supply

The bundle is a single static-size graph: a log-mel spectrogram (1×80×3000) in → encoder hidden states (1×375×1024) out. You supply the mel front-end, the decoder loop, and diarization in your host code (Swift or Python), using the upstream feature/tokenizer config.

pip install coreai-catalog && coreai-catalog install moss-transcribe-diarize-audio

Requirements

• Deployment: macOS 27.0+ / iOS 27.0+, Xcode 27+. The asset serializes with minimum_os v27, so the on-device Swift runtime requires macOS/iOS 27+. A Mac on macOS 26 can convert and inspect it but not run it on-device.
• Apple Silicon.

Verification (output parity)

• Gate A (structure): passed — the bundle's layout + metadata were validated; the graph loads.
• Gate B — graph_output_cosine: 1.000000 min output cosine (median 1.000000) vs the fp32 torch audio encoder over 8 seeded log-mel spectrograms, measured on apple_silicon. Certifies the export computes the SAME output as the source — a conversion-fidelity metric, not task accuracy.
• This certifies the export is numerically faithful to the source encoder — it does NOT certify word-error-rate or diarization accuracy on your audio. Reproduce with coreai-fabric verify.

Provenance

Field	Value
Base model	OpenMOSS-Team/MOSS-Transcribe-Diarize @ d7231bbae2587a4af278735eb765b318c4f64edd
Converted by	models/moss_transcribe/export.py (version not reported)
Recipe	moss-transcribe-diarize-audio (recipe_source: fabric)
Precision / quantization	float32 / none
Conversion date	2026-07-10

Machine-readable, in this repo: parity-report.json · reproduce-manifest.json · LICENSE.

License and attribution

Weights licensed apache-2.0 — see the bundled LICENSE. This artifact is a converted derivative of the base model: its weights were converted to Apple Core AI format. The conversion itself is community work.

Links

• Base model: OpenMOSS-Team/MOSS-Transcribe-Diarize
• Reproduce: recipe moss-transcribe-diarize-audio
• Index: coreai-catalog
• HF Collection

The on-device Core AI ecosystem

• coreai-fabric — the reproducible recipe → .aimodel pipeline that produced this asset.
• coreai-catalog — the index of Core AI models with provenance and integration snippets.
• apple/coreai-models — Apple's official exporters and runtimes.

Not affiliated with Apple

Community conversion. Not produced, hosted, or endorsed by Apple. Apple and Core AI are trademarks of Apple Inc., used here only to describe the target runtime/format.