v0.3.1

andimarafioti/faster-qwen3-ttsv0.3.1Jul 15, 2026by andimarafioti

AI Summary

This release introduces a unified public warmup API for both Torch and GGML backends to resolve interface mismatches, while ensuring backward compatibility. It also updates the demo Space deployment to use the published PyPI package.

Key Highlights

Added a public `model.warmup(prefill_len=100)` API shared by the Torch and GGML backends.
Maintained backward compatibility by keeping the former private `_warmup()` entry point as a compatibility alias.
Implemented CUDA graph capture for the Torch backend during warmup.
Updated internal, demo, example, and test callers to use the new public API.
Installed the published PyPI package in the demo Space deployment.

New Features

Public backend warmup API with prefill_len parameter.
CUDA graph support for the Torch backend.
PyPI package installation in the demo Space.

Full Release Notes

## Highlights

- Add a public `model.warmup(prefill_len=100)` API shared by the Torch and GGML backends.
- Keep the former private Torch `_warmup()` entry point as a compatibility alias.
- Capture CUDA graphs once on Torch while treating GGML warmup as a safe no-op until qwentts.cpp exposes a native preparation hook.
- Update internal, demo, example, and test callers to use the public API.
- Install the published PyPI package in the demo Space deployment.

This resolves the backend interface mismatch reported in [huggingface/speech-to-speech#354](https://github.com/huggingface/speech-to-speech/issues/354).

## Installation

```bash
pip install faster-qwen3-tts==0.3.1
```

For the optional GGML backend:

```bash
pip install "faster-qwen3-tts[ggml]==0.3.1"
```

## Changes

- [#128](https://github.com/andimarafioti/faster-qwen3-tts/pull/128) — Expose public backend warmup API
- [#122](https://github.com/andimarafioti/faster-qwen3-tts/pull/122) — Use the PyPI package in the demo Space

**Full changelog:** https://github.com/andimarafioti/faster-qwen3-tts/compare/v0.3.0...v0.3.1