--- license: cc-by-4.0 pipeline_tag: automatic-speech-recognition base_model: nvidia/parakeet-tdt-0.6b-v3 tags: - coreai - apple-silicon - parakeet - speech-to-text - fp16 --- # Parakeet v3 — Apple Core AI streaming conversion Conversion of [nvidia/parakeet-tdt-0.6b-v3 at 541d1f99c6b0c3cd0b11a95167540bb8edefd82b](https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3/tree/541d1f99c6b0c3cd0b11a95167540bb8edefd82b) for Apple's CoreAISpeech runtime on **macOS 27**. Source training is by NVIDIA (v3) and Moondream (Ultra/Redux). Conversion by maxffarrell; no training or endorsement. Weights are **CC-BY-4.0**. Preserve attribution to NVIDIA and, for Ultra/Redux, Moondream when redistributing. [License terms](https://creativecommons.org/licenses/by/4.0/). The export recipe is Apple BSD-3-Clause; see APPLE_LICENSE. ## Format and use Download `model.zip`, verify against `artifact.json`, and extract. Pass the extracted `parakeet-tdt-0.6b-v3_float16_streaming150` directory to `SpeechRecognitionModel(resourcesAt:)` from [apple/coreai-models at 52c84ba](https://github.com/apple/coreai-models/tree/52c84ba874b2c57adcede08a671ce96ed1b3f433). Three `.aimodel` graphs: encoder, recurrent decoder step, joint. Use `startStream`, `append(pcm:)` and `finishStream`; collect **all** finalized update segments, since finishStream returns only its last segment. FP16 static window: 128 mel bins × 1201 frames; left/chunk/right encoder frames **113/12/25**. Audio: mono Float32 16 kHz. Buffered streaming has 2.96 seconds of theoretical audio latency before inference or display stabilization. Streaming is buffered full-window inference, not a cache-aware causal model. No guaranteed exclusive Neural Engine placement or speedup. ## Changes and caveats Graph format, FP16 precision and fixed streaming geometry differ from the source. Redux's ternary base-3 packing is expanded exactly before FP16 conversion: **the original 178 MB packed size and Photon specialized performance are not retained**. All conversions are about 1.2 GB unpacked. Auxiliary Photon VAD tensors are omitted; CoreAISpeech endpointing is used separately. NVIDIA v3 tokenizer vocabulary was checked against each source. ## Validation `conversion-validation.json` compares all three graphs against the pinned PyTorch source, including masked encoder frames, nonzero recurrent state and joint logits, using the macOS Core AI GPU preference. Acceptance: cosine > 0.995 and relative RMSE < 0.05. Swift runtime transcription and stream boundary regression checks are recorded in the [Pladder fork](https://github.com/maxffarrell/pladder/tree/codex/coreai-dictation). These are conversion checks on synthetic speech, not multilingual quality claims or calibrated confidence scores. ## Reproduce See `tools/coreai/export_models.py`, `verify_models.py` and the dependency pins in [Pladder](https://github.com/maxffarrell/pladder/tree/codex/coreai-dictation/tools/coreai). Source and recipe revisions, omitted tensors and precision are embedded in `provenance.json`. Every bundle file has a SHA-256 entry in `SHA256SUMS`.