amrmantawi/bmo-paroli-tts

BMO Paroli TTS (ONNX)

Custom English text-to-speech model for the local-llm BMO assistant.

Files

File	Description
encoder.onnx	Paroli encoder (phoneme → latent)
decoder.onnx	Paroli decoder (latent → audio)
config.json	Piper/Paroli config (phoneme map, inference settings)

Usage

Designed for use with the Paroli TTS backend in local-llm. Requires espeak-ng phonemization (en-us voice).

Inference defaults from config.json:

• Sample rate: 22050 Hz
• noise_scale: 0.667
• length_scale: 0.8
• noise_w: 0.8

License

MIT (adjust if your training data or base model requires a different license).