ACT900

Private Speech-to-Text STT & Text-to-Speech TTS, OpenAI & 52 voices

Faster Whisper

CORS allowed origins. Change to your domain for production.

Auto-generated API key. Find in Variables tab after deploy.

Logging verbosity: debug, info, warning, error.

AI models downloaded at startup. Includes STT (base) and TTS (Kokoro). See README for model options.

Required for volume write permissions. Do not change.

Seconds before unloading model from memory. -1 = never unload.

Quantization type for CPU inference. INT8 recommended.