islomov

/

navaistt_v2_medium

like 8

A Docker-based STT (Automatic Speech Recognition) system using faster-whisper with a Gradio web UI.
Features

• faster‑whisper with CTranslate2 for ~2× real-time transcription on GPU
• Lazy loading: model loads on first request (20–30 s), then stays in memory
• Token‑based auth: secure /transcribe with X-API-KEY
• UI Basic Auth: protect Gradio interface with username/password
• Multi‑format support: WAV, MP3, OGG/Opus, M4A, FLAC, AMR, etc.
• Health check: /healthz returns OK immediately

https://github.com/nek1987/stt-service-webui/