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.
17 lines
548 B
YAML
17 lines
548 B
YAML
# Minimal compose for running chatterbox-fast. `docker compose up --build`.
|
|
# Mount a HuggingFace cache so the ~6 GB Turbo weights survive container recreate.
|
|
services:
|
|
chatterbox-fast:
|
|
build: .
|
|
image: chatterbox-fast:latest
|
|
restart: unless-stopped
|
|
gpus: all
|
|
ports:
|
|
- "${CBF_PORT:-8197}:8197"
|
|
environment:
|
|
- CBF_DEFAULT_VOICE=${CBF_DEFAULT_VOICE:-}
|
|
volumes:
|
|
- ${HF_CACHE:-./hf_cache}:/app/hf_cache
|
|
# Optional: mount your own voices over the baked starter set.
|
|
# - ./my-voices:/app/voices
|