Peter Michael Gits
Use official HuggingFace documentation approach to override model detection
2a50bb0
metadata
title: STT GPU Service Python v4
emoji: 🎙️
colorFrom: blue
colorTo: green
sdk: docker
app_port: 7860
hardware: l4
sleep_time_timeout: 1800
suggested_storage: small
pinned: false
app_file: app.py
models: []
datasets: []

STT GPU Service Python v4

Real-time WebSocket STT streaming service using full Moshi model.

Features

  • WebSocket streaming (80ms chunks at 24kHz)
  • REST API endpoints
  • FastAPI backend with real-time transcription
  • L4 GPU acceleration (30GB VRAM)
  • Full Moshi model for high-quality STT

Endpoints

  • / - Web interface for testing
  • /ws/stream - WebSocket streaming endpoint
  • /api/transcribe - REST API endpoint
  • /health - Health check