Skip to content
ModelListedExternal
Download

Supertonic 3 CoreML

First-party CoreML export of Supertonic-3's four non-autoregressive flow-matching graphs, for on-device iOS / Apple Neural Engine. Built by our own pipeline (speech-models/stmodels): weights lifted from the...

Sourcehuggingface.co/aufklarer/Supertonic-3-CoreML/tree/d159f984aa1c4ccaafd0541076463e422cf75408
KindOtherBase modelSupertone/supertonic-3VersionRevision d159f98Bgrade

Model source

Kind
Other
File size
3.42–243 MB
Base model
Supertone/supertonic-3
Version
vRevision d159f98
Updated
Oct 19, 2026
Downloads
2.9K
Source
Hugging Face

Source description

SupertonicTTS-3 — CoreML (.mlpackage, iOS)

First-party CoreML export of Supertonic-3's four non-autoregressive flow-matching graphs, for on-device iOS / Apple Neural Engine. Built by our own pipeline (speech-models/stmodels): weights lifted from the Supertone/supertonic-3 ONNX initializers → PyTorch nn.Module → coremltools (mlprogram, FP32, iOS18+).

Graphs & parity (FP32, vs ONNX Runtime)

Modulemlpackageparity max|Δ|
Duration predictorDurationPredictor.mlpackage7.2e-06 ✓
Vector estimator (ODE denoiser)VectorEstimator.mlpackage2.5e-03 ✓
VocoderVocoder.mlpackage3.0e-04 ✓
Text encoderTextEncoder.mlpackagemean 2.5e-04 (max 2.5e-2 at isolated positions)

Text/duration use fixed T=128 (relpos attention has T-dependent pad widths — pad/segment text to 128); vocoder + vector-estimator use a dynamic latent-length RangeDim. The host runs the flow-matching ODE loop (vector_estimator ×total_steps) — the graphs contain no control flow. Assets to drive them: tts.json, unicode_indexer.json (G2P-free tokenizer table), voice_styles/*.json.

FP32 = parity reference. For ANE residency, use the mixed-precision Supertonic-3-CoreML-FP16 — vocoder + duration FP16, text-encoder + vector-estimator FP32; measured transparent at 47–51 dB mag-STFT SNR.

Attribution & license

  • Weights: derivative of Supertone/supertonic-3 (commit 3cadd1ee6394adea1bd021217a0e650ede09a323), Supertone Inc., arXiv:2503.23108 — OpenRAIL-M (use-based restrictions carry over: no non-consensual impersonation/deepfakes, etc.).

Other Supertonic-3 formats

Ecosystem

  • soniqo.audio — website / use-case explorer (transcription, voice cloning, live ASR, voice agents).
  • speech-core — C++ orchestration library; Supertonic plugs in as a TTSInterface CoreML model.
  • speech-swift — Apple Silicon MLX + CoreML runtime.
  • speech-android — Android SDK consuming on-device LiteRT bundles.

Other CoreML models

Sources

1 source
Hugging FacePrimaryVersion d159f984aa1c4ccaafd0541076463e422cf75408
huggingface.co/aufklarer/Supertonic-3-CoreML/tree/d159f984aa1c4ccaafd0541076463e422cf75408

Verified Aug 29

Model artifacts

4 artifacts

DurationPredictor.mlpackage/Data/com.apple.CoreML/weights/weight.bin

bin · 3.42 MB · SHA-256 714d7d6bcf0c…6625 · Hugging Face

Download

TextEncoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin

bin · 34.4 MB · SHA-256 9618a7595441…e8d7 · Hugging Face

Download

VectorEstimator.mlpackage/Data/com.apple.CoreML/weights/weight.bin

bin · 243 MB · SHA-256 593cb7df14ef…e1fd · Hugging Face

Download

Vocoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin

bin · 96.6 MB · SHA-256 f0f325c9e90f…ac6d · Hugging Face

Download

Source excerpts

2 excerpts
ReadmeAug 29huggingface.co/aufklarer/Supertonic-3-CoreML

--- license: openrail base_model: Supertone/supertonic-3 tags: - text-to-speech - tts - coreml - on-device - ios - ane - supertonic - flow-matching library_name: coreml pipeline_tag: text-to-speech --- # SupertonicTTS-3 — CoreML (.mlpackage, iOS) **First-party CoreML export** of Supertonic-3's four non-autoregressive flow-matching graphs, for on-device iOS / Apple Neural Engine. Built by our own pipeline (`speech-models/stmodels`): weights lifted from the [`Supertone/supertonic-3`](https://huggingface.co/Supertone/supertonic-3) ONNX initializers → PyTorch `nn.Module` → `coremltools` (`mlprogram`, FP32, iOS18+). ## Graphs & parity (FP32, vs ONNX Runtime) | Module | mlpackage | parity max\|Δ\| | |---|---|---| | Duration predictor | `DurationPredictor.mlpackage` | 7.2e-06 ✓ | | Vector estimator (ODE denoiser) | `VectorEstimator.mlpackage` | 2.5e-03 ✓ | | Vocoder | `Vocoder.mlpackage` | 3.0e-04 ✓ | | Text encoder | `TextEncoder.mlpackage` | **mean 2.5e-04** (max 2.5e-2 at isolated positions) | Text/duration use **fixed T=128** (relpos attention has T-dependent pad widths — pad/segment text to 128); vocoder + vector-estimator use a dynamic latent-length `RangeDim`. The host runs the flow-matching ODE loop (`vector_estimator` ×`total_steps`) — the graphs contain no control flow. Assets to drive them: `tts.json`, `unicode_indexer.json` (G2P-free tokenizer table), `voice_styles/*.json`. > FP32 = parity reference. For **ANE residency**, use the mixed-precision > [`Supertonic-3-CoreML-FP16`](https://huggingface.co/aufklarer/Supertonic-3-CoreML-FP16) — vocoder + duration FP16, > text-encoder + vector-estimator FP32; measured transparent at **47–51 dB** mag-STFT SNR. ## Attribution & license - **Weights**: derivative of [`Supertone/supertonic-3`](https://huggingface.co/Supertone/supertonic-3) (commit `3cadd1ee6394adea1bd021217a0e650ede09a323`), Supertone Inc., [arXiv:2503.23108](https://arxiv.org/abs/2503.23108) — **OpenRAIL-M** (use-based restrictions carry over: no non-consensual impersonation/deepfakes, etc.). ## Other Supertonic-3 formats - [**Supertonic-3 — CoreML (FP16)**](https://huggingface.co/aufklarer/Supertonic-3-CoreML-FP16) — mixed-precision ANE variant (47–51 dB). - [**Supertonic-3 — ONNX (INT8)**](https://huggingface.co/soniqo/Supertonic-3-ONNX-INT8) — server / desktop (ONNX Runtime). - [**Supertonic-3 — LiteRT**](https://huggingface.co/soni...

NoteAug 29huggingface.co/aufklarer/Supertonic-3-CoreML

Source context: 2860 downloads · 0 likes · Pipeline text-to-speech · Library coreml · Repo aufklarer/Supertonic-3-CoreML