Nooder/supertonic-2-coreml

By Nooder · ↓ 14 · ♥ 0 · Updated 2026-01-19

What's in this repo

This page lists the files in Nooder/supertonic-2-coreml, including duration_predictor.onnx, text_encoder.onnx.

2 ONNX models (.onnx files)
106 files, including 40 .json, 28 .mlmodel, 26 .bin, 5 .md, 2 .onnx, 1 .py
Last updated 2026-01-19

ONNX files

File	Size	Download
resources/onnx/duration_predictor.onnx	—	Download
resources/onnx/text_encoder.onnx	—	Download

How to use

These ONNX files can be loaded with ONNX Runtime. To inspect a model graph, upload the file to the Netron viewer. Download links point to the original files on Hugging Face.

README

Supertonic-2 CoreML

This repository provides CoreML exports of Supertonic 2 for macOS and iOS. It focuses on on-device inference with multiple >=8-bit quantization variants.

GitHub repo (code + demo app): https://github.com/Nooder/supertonic-2-coreml

Code & demo

The GitHub repo contains:

Swift demo app (CoreML pipeline + UI): supertonic2-coreml-ios-test/
CoreML tooling + tests: scripts/
Docs: docs/

What is included

models/: CoreML model packages by variant (>=8-bit only)
resources/: voice styles, embeddings, and text normalization assets
manifest.json: list of artifacts with checksums and sizes
SHA256SUMS: sha256 checksums for all files
tests/: smoke tests for CoreML model loading

Quickstart (iOS / macOS)

Pick a variant from models/ (see the quant matrix in docs/quant-matrix.md).
Bundle the corresponding CoreML packages and resources/ into your app.
Use the Swift demo app in the GitHub repo supertonic-2-coreml as the reference implementation.

Required files (checklist)

Bundle the following into your app:

CoreML packages for your chosen variant:
- duration_predictor_mlprogram.mlpackage
- text_encoder_mlprogram.mlpackage
- vector_estimator_mlprogram.mlpackage
- vocoder_mlprogram.mlpackage
resources/voice_styles/
resources/embeddings/
resources/onnx/unicode_indexer.json
resources/onnx/tts.json

Minimal iOS integration

// Example usage (see demo app for full UI + playback)
let service = try TTSService(computeUnits: .all)
let result = try service.synthesize(
    text: "Hello from CoreML!",
    language: .en,
    voiceName: "F1",
    steps: 20,
    speed: 1.0,
    silenceSeconds: 0.3
)
print("WAV file:", result.url)

To select a specific variant, update the CoreML folder name in TTSService (the demo defaults to coreml_int8).

Example: iOS 18 `int8_both`

This variant uses int8 weights for multiple stages on iOS 18.

Bundle these files in your app:

Resources/
  coreml_ios18_int8_both/
    duration_predictor_mlprogram.mlpackage
    text_encoder_mlprogram.mlpackage
    vector_estimator_mlprogram.mlpackage
    vocoder_mlprogram.mlpackage
  voice_styles/
  embeddings/
  onnx/
    unicode_indexer.json
    tts.json

In the Swift demo app, update the CoreML folder name to point at coreml_ios18_int8_both (the app defaults to coreml_int8).

Choosing a variant

Use the folder naming to select the right artifact:

coreml_int8: faster, lower fidelity
coreml_compressed: smaller memory (linear8)
coreml_ios18_*: for iOS 18 CoreML runtime (>=8-bit only)

4-bit variants are intentionally excluded due to quality.

Variant matrix (quick view)

Variant folder	Quantization (by name)	Intended target	Notes
`coreml`	full precision (mixed)	general	baseline quality
`coreml_int8`	int8 (all stages)	general	faster, lower fidelity
`coreml_compressed`	linear8	general	smaller memory
`coreml_ios18`	full precision (mlprogram)	iOS 18+	best quality on iOS 18
`coreml_ios18_int8_vocoder_only`	int8 (vocoder only)	iOS 18+	balanced
`coreml_ios18_int8_both`	int8 (multiple stages)	iOS 18+	fastest, more loss
`coreml_compressed_ios18`	linear8	iOS 18+	smallest memory

For deeper guidance, see docs/compatibility-matrix.md and docs/quant-matrix.md.

Steps vs. quality (quick guide)

Steps	Speed	Quality
10	fastest	lowest
20	balanced	good
30	slowest	best

Troubleshooting

Missing resource error: Ensure resources/ folders are bundled and named exactly.
Model not found: Confirm the CoreML folder name (e.g., coreml_ios18_int8_both).
Fails to load on device: Check iOS deployment target matches your variant.

Tests

The tests/test_coreml_models.py script runs a simple smoke test that loads all stages (duration predictor, text encoder, vector estimator, vocoder) with dummy inputs.

Attribution and license

This CoreML export is derived from Supertone/supertonic-2. Model weights are licensed under OpenRAIL-M (see LICENSE). Sample code is MIT-licensed (see NOTICE and UPSTREAM.md).

Model info

text-to-speech supertonic

supertonic coreml onnx ios macos tts mlprogram text-to-speech en ko es pt

↓ 14 ♥ 0

Created: 2026-01-19

Updated: 2026-01-19

View on Hugging Face

Files (106)

.gitattributes

LICENSE

NOTICE

README.md

SHA256SUMS

UPSTREAM.md

config.json

docs/compatibility-matrix.md

docs/quant-matrix.md

docs/release-checklist.md

manifest.json

models/coreml/duration_predictor_fixed.mlmodel

models/coreml/duration_predictor_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml/duration_predictor_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml/duration_predictor_mlprogram.mlpackage/Manifest.json

models/coreml/text_encoder_fixed.mlmodel

models/coreml/text_encoder_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml/text_encoder_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml/text_encoder_mlprogram.mlpackage/Manifest.json

models/coreml/vector_estimator_fixed.mlmodel

models/coreml/vector_estimator_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml/vector_estimator_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml/vector_estimator_mlprogram.mlpackage/Manifest.json

models/coreml/vocoder_fixed.mlmodel

models/coreml/vocoder_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml/vocoder_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml/vocoder_mlprogram.mlpackage/Manifest.json

models/coreml_compressed/vector_estimator_mlprogram_linear8.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_compressed/vector_estimator_mlprogram_linear8.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_compressed/vector_estimator_mlprogram_linear8.mlpackage/Manifest.json

models/coreml_compressed/vocoder_mlprogram_linear8.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_compressed/vocoder_mlprogram_linear8.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_compressed/vocoder_mlprogram_linear8.mlpackage/Manifest.json

models/coreml_compressed_ios18/vector_estimator_mlprogram_linear8.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_compressed_ios18/vector_estimator_mlprogram_linear8.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_compressed_ios18/vector_estimator_mlprogram_linear8.mlpackage/Manifest.json

models/coreml_compressed_ios18/vocoder_mlprogram_linear8.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_compressed_ios18/vocoder_mlprogram_linear8.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_compressed_ios18/vocoder_mlprogram_linear8.mlpackage/Manifest.json

models/coreml_int8/duration_predictor_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_int8/duration_predictor_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_int8/duration_predictor_mlprogram.mlpackage/Manifest.json

models/coreml_int8/text_encoder_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_int8/text_encoder_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_int8/text_encoder_mlprogram.mlpackage/Manifest.json

models/coreml_int8/vector_estimator_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_int8/vector_estimator_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_int8/vector_estimator_mlprogram.mlpackage/Manifest.json

models/coreml_int8/vocoder_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_int8/vocoder_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_int8/vocoder_mlprogram.mlpackage/Manifest.json

models/coreml_ios18/duration_predictor_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_ios18/duration_predictor_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_ios18/duration_predictor_mlprogram.mlpackage/Manifest.json

models/coreml_ios18/text_encoder_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_ios18/text_encoder_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_ios18/text_encoder_mlprogram.mlpackage/Manifest.json

models/coreml_ios18/vector_estimator_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_ios18/vector_estimator_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_ios18/vector_estimator_mlprogram.mlpackage/Manifest.json

models/coreml_ios18/vocoder_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_ios18/vocoder_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_ios18/vocoder_mlprogram.mlpackage/Manifest.json

models/coreml_ios18_int8_both/duration_predictor_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_ios18_int8_both/duration_predictor_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_ios18_int8_both/duration_predictor_mlprogram.mlpackage/Manifest.json

models/coreml_ios18_int8_both/text_encoder_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_ios18_int8_both/text_encoder_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_ios18_int8_both/text_encoder_mlprogram.mlpackage/Manifest.json

models/coreml_ios18_int8_both/vector_estimator_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_ios18_int8_both/vector_estimator_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_ios18_int8_both/vector_estimator_mlprogram.mlpackage/Manifest.json

models/coreml_ios18_int8_both/vocoder_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_ios18_int8_both/vocoder_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_ios18_int8_both/vocoder_mlprogram.mlpackage/Manifest.json

models/coreml_ios18_int8_vocoder_only/duration_predictor_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_ios18_int8_vocoder_only/duration_predictor_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_ios18_int8_vocoder_only/duration_predictor_mlprogram.mlpackage/Manifest.json

models/coreml_ios18_int8_vocoder_only/text_encoder_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_ios18_int8_vocoder_only/text_encoder_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_ios18_int8_vocoder_only/text_encoder_mlprogram.mlpackage/Manifest.json

models/coreml_ios18_int8_vocoder_only/vector_estimator_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_ios18_int8_vocoder_only/vector_estimator_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_ios18_int8_vocoder_only/vector_estimator_mlprogram.mlpackage/Manifest.json

models/coreml_ios18_int8_vocoder_only/vocoder_mlprogram.mlpackage/Data/com.apple.CoreML/model.mlmodel

models/coreml_ios18_int8_vocoder_only/vocoder_mlprogram.mlpackage/Data/com.apple.CoreML/weights/weight.bin

models/coreml_ios18_int8_vocoder_only/vocoder_mlprogram.mlpackage/Manifest.json

resources/embeddings/char_embedder_dp.fp32.bin

resources/embeddings/char_embedder_dp.shape.json

resources/embeddings/char_embedder_te.fp32.bin

resources/embeddings/char_embedder_te.shape.json

resources/onnx/duration_predictor.onnx ONNX

resources/onnx/text_encoder.onnx ONNX

resources/onnx/tts.json

resources/onnx/unicode_indexer.json

resources/voice_styles/F1.json

resources/voice_styles/F2.json

resources/voice_styles/F3.json

resources/voice_styles/F4.json

resources/voice_styles/F5.json

resources/voice_styles/M1.json

resources/voice_styles/M2.json

resources/voice_styles/M3.json

resources/voice_styles/M4.json

resources/voice_styles/M5.json

tests/test_coreml_models.py

Related models

More from Nooder

No other models by this author.

More text-to-speech models

Browse all text-to-speech models