v0.4.0-rc.1

Phonon runtime preview

This release packages the Phonon runtime for Python, Rust, browsers, and Apple apps. Supply a checkpoint from a local folder or Hugging Face, including its own config, tokenizer, weights, and voices. Model and voice licenses are separate from the runtime's MIT OR Apache-2.0 license.
The model is currently not yet public, so one needs to supply the model path to a separately provided model.

Try it

Run the Python command with uv and a q8 checkpoint:

uvx --from 'ptts==0.4.0rc1' ptts --model /path/to/model --lang en --quant q8 "Hello from Phonon." -o out.wav

Install the browser preview:

npm install phonon-tts@next

Install the Rust command:

cargo install ptts --version 0.4.0-rc.1 --locked --features cli

Native downloads

CLI archives cover Linux x64/ARM64, Mac Apple silicon/Intel, and Windows x64. Check SHA256SUMS and the platform requirements. ptts.rb installs the matching CPU command with Homebrew.

Apple apps can use ptts-swift-0.4.0-rc.1.zip with its checksum-pinned PhononCore.xcframework.zip. Supply an exported Core ML model bundle as described in the Swift guide.

All integrations require an explicit model source and normalization language. Private Hugging Face checkpoints require access credentials. The selected public Phonon checkpoint will follow at model launch.