---
license: other
pipeline_tag: text-to-speech
library_name: phoonnx
tags: [phoonnx, optispeech, tts]
---
# phoonnx-optispeech
OptiSpeech voices for [phoonnx](https://github.com/TigreGotico/phoonnx) — non-autoregressive (FastSpeech2-style)
acoustic + GAN vocoder, single ONNX (waveform out). One voice per subfolder
(`model.onnx` + native `config.json`). Mirror of
[mush42/optispeech](https://huggingface.co/mush42/optispeech), en-US, 24 kHz.
Part of the [phoonnx community mirror](https://huggingface.co/collections/OpenVoiceOS/phoonnx-community-mirror-6a2204d302e0e7497870cd68).

| voice | acoustic | speaker |
|---|---|---|
| `mush42__optispeech-lightspeech-en-us-emily` | LightSpeech | emily |
| `mush42__optispeech-lightspeech-en-us-mike` | LightSpeech | mike |
| `mush42__optispeech-convnext-en-us-emily` | ConvNeXt-TTS | emily |

The native `config.json` is generated from OptiSpeech's embedded ONNX metadata.
See [docs/optispeech.md](https://github.com/TigreGotico/phoonnx/blob/dev/docs/optispeech.md).