Multilingual text-to-speech on ONNX Runtime
Hebrew · English · Spanish · Italian · German

 

---

Quick start

git clone https://github.com/maxmelichov/BlueTTS.git
cd BlueTTS
uv sync
uv run hf download notmax123/BlueTTS2.5-onnx --repo-type model --local-dir ./onnx_models

Voice JSONs ship in voices/, so you are ready to synthesize:

import soundfile as sf
from blue_onnx import BlueTTS

tts = BlueTTS(onnx_dir="onnx_models", style_json="voices/noa.json")
samples, sr = tts.synthesize("שלום, זהו מודל דיבור בעברית.", lang="he")
sf.write("out.wav", samples, sr)

Numbers, dates, prices and codes are spoken as words automatically. Mix languages inline with <en>…</en>:

samples, sr = tts.synthesize("שלום לכולם, <en>welcome to the presentation</en>.", lang="he")

Hebrew grapheme-to-phoneme is handled by RenikudPlus, which downloads its own weights the first time you synthesize Hebrew.

Documentation

src/blue_onnx/	Inference API — entry points, text normalization, language spans, accelerators
examples/	Runnable scripts for every feature
exports/	New voices, ONNX export, TensorRT engines
training/	Dataset prep and the three training stages

Install

Requires Python 3.12+ and uv.

git clone https://github.com/maxmelichov/BlueTTS.git
cd BlueTTS
uv sync

Optional extras are documented per use case: accelerators (OpenVINO, CUDA), --extra export for voice/ONNX export, and --extra tensorrt.

Models

Current — notmax123/BlueTTS2.5-onnx

uv run hf download notmax123/BlueTTS2.5-onnx --repo-type model --local-dir ./onnx_models

Ships the core graphs (text_encoder, vector_estimator, vocoder, duration_predictor_style), the runtime tts.json / vocab.json, stats.npz and uncond.npz, a reference_encoder for reference-audio conditioning, and five voice JSONs under voices/. Guidance comes from uncond.npz rather than a baked cfg_scale input, and the vocoder takes the de-normalized latent — blue_onnx detects both from the graphs, so nothing to configure.

This bundle is also the input to create_tensorrt.py: its file names line up with the engines blue_trt expects.

Previous — notmax123/blue-onnx-v2. Still supported and still the only bundle with the zero-shot voice-conversion graphs (codec_encoder, style_encoder, duration_style_encoder) that examples/zero_shot.py and blue_onnx.style need.

Neither bundle includes per-voice style JSON — use voices/*.json from this repo, or export your own from a reference clip (needs the PyTorch checkpoints at notmax123/blue-v2).

Citations

@ARTICLE{2025arXiv250323108K,
       author = {{Kim}, Hyeongju and {Yang}, Jinhyeok and {Yu}, Yechan and {Ji}, Seunghun and {Morton}, Jacob and {Bous}, Frederik and {Byun}, Joon and {Lee}, Juheon},
        title = "{SupertonicTTS: Towards Highly Efficient and Streamlined Text-to-Speech System}",
      journal = {arXiv e-prints},
     keywords = {Audio and Speech Processing, Machine Learning, Sound},
        pages = {arXiv:2503.23108},
}
@article{kim2025training,
  title={Training Flow Matching Models with Reliable Labels via Self-Purification},
  author={Kim, Hyeongju and Yu, Yechan and Yi, June Young and Lee, Juheon},
  journal={arXiv preprint arXiv:2509.19091},
  year={2025}
}
@misc{yi2025robustttstrainingselfpurifying,
      title={Robust TTS Training via Self-Purifying Flow Matching for the WildSpoof 2026 TTS Track},
      author={June Young Yi and Hyeongju Kim and Juheon Lee},
      year={2025},
      eprint={2512.17293},
      archivePrefix={arXiv},
      primaryClass={cs.SD},
      url={https://arxiv.org/abs/2512.17293},
}

License

MIT.

Voice cloning and responsibility

This software can produce speech that mimics a reference voice. The maintainers and contributors are not responsible for what you do with it — compliance with law, consent from voice owners, and ethical use are entirely your responsibility. Do not use it to deceive, impersonate without permission, or infringe anyone's rights.