Whisper STT local server

A high-performance bridge to a local Whisper STT server. Designed for real-time conversational AI agents where every millisecond counts.

Repository

https://github.com/fakehec/whisper-stt-local-server

Architecture

Redirects all transcription requests to a high-performance local GPU server at http://localhost:5000.

Usage

{baseDir}/scripts/whisper-bridge.sh /path/to/audio.mp3 large-v3-turbo en

Key Metrics

• Latency: ~0.2s for standard voice commands.
• Accuracy: High (Whisper large-v3-turbo).
• Concurrency: Hybrid thread-managed GPU architecture.