pypi / piper-tts

piper-tts

What changed in piper-tts 1, from its official release notes, upgrade guides and advisories. Sources checked 2026-10-27.

2 breaking11 new APIs and options1 performance1 fixes and other

See which of these apply to your code, ranked, in your agent:

npx @unstale/cli init

piper-tts 115 changes

Run the g2pW model through piper.g2pw_onnx instead of g2pw.api; the model runs under onnxruntime without the prior torch-based batching.
1.6.1release notes
The zh extra no longer installs torch or requests, reducing its installed dependencies.
1.6.1release notes

Set this training option to thai to use the Thai phonemizer.
1.8.0release notes
Set this voice configuration field to thai to use the Thai phonemizer for synthesis.
1.8.0release notes
Adds a setup command option for Thai.
1.8.0release notes
Adds a th extra for installing the Thai phonemizer dependency.
1.8.0release notes
Adds TLTK-based Thai word segmentation and tone-aware phonemization; Thai voices remain compatible with the IPA-based (espeak) warmstart.
1.8.0release notes
Use --data.phoneme_type japanese for training with the Japanese phonemizer.
1.7.0release notes
Adds a ja extra with an OpenJTalk-based Japanese phonemizer using pyopenjtalk-plus.
1.7.0release notes
Set phoneme_type to japanese in a voice config to synthesize with the Japanese phonemizer.
1.7.0release notes
Adds piper_create_options to the libpiper API.
1.7.0release notes
Adds piper_create_with_options() to libpiper; piper_create() remains as a wrapper for ABI compatibility.
1.7.0release notes
Adds a setup command option for Japanese.
1.7.0release notes

g2pW runs 1.5–2x faster because it no longer forks DataLoader worker processes on every call.
1.6.1release notes

Parses full-context labels for pitch accent and maps them to IPA, keeping Japanese voices compatible with IPA-based (espeak) warmstart.
1.7.0release notes