vh
76314624bb
ci / test (push) Has been cancelled
Sub-second streaming TTS on Chatterbox-Turbo via adaptive buffer-ratchet chunking. First audio in ~0.5s (vs ~5s one-shot) with no quality compromise — chunk joins land on natural sentence pauses and the stream converges to one large near-full-context chunk within 2-3 joins. Works because the engine runs faster than realtime; the no-starvation guarantee is proven in a GPU-free simulation (tests/test_scheduler.py). - chatterbox_fast/scheduler.py: the adaptive-chunk scheduler (pure logic, no GPU) - chatterbox_fast/app.py: FastAPI server (POST /tts streaming, /voices, /health) - bench.py: streaming client (ground-truth TTFB + starvation check) - Self-contained Dockerfile (slim base + chatterbox-tts from PyPI) - Three public-domain LibriVox starter voices baked in (see voices/ATTRIBUTION.md) MIT licensed.
878 B
878 B
Starter voice attribution
The voices shipped with chatterbox-fast are short clips from LibriVox, whose recordings are dedicated to the public domain in the USA (https://librivox.org/pages/public-domain/). No attribution is legally required; it's provided here out of respect for the volunteer readers.
All three are excerpts from LibriVox Short Poetry Collection 001 (https://archive.org/details/short_poetry_001_librivox):
| voice | reader | from |
|---|---|---|
catharine |
Catharine Eastman | "Dover Beach" — Matthew Arnold |
peter |
Peter Yearsley | "Lament of the Irish Emigrant" — Lady Dufferin (H. Selina) |
kara |
Kara Shallenberg | "How They Brought the Good News from Ghent to Aix" — Robert Browning |
Each is a ~20-second mono excerpt, used as a voice-cloning reference. The underlying poems are themselves in the public domain.