# WhisperLive GPU

WhisperLive GPU is a local, near-real-time speech-to-text application for homelab users that transcribes live microphone audio or pre-recorded files with OpenAI Whisper on an NVIDIA GPU, supporting word timestamps,. It is open source and self-hosted, installable on Unraid.

Canonical page: https://ourlocalhost.com/app/3479/whisperlive-gpu

## Description

WhisperLive GPU is a local, near-real-time speech-to-text application for homelab users that transcribes live microphone audio or pre-recorded files with OpenAI Whisper on an NVIDIA GPU, supporting word timestamps, hotwords, and speaker diarization.

## Platform availability

Available on 1 of 10 platforms: Unraid.

## Health score

76/100 (Good), based on 4 of 4 factors.
- Maintained: 100 (last commit 19 days ago)
- Popular: 73 (4,347 GitHub stars)
- Easy to install: 35 (1 install platform, install notes, Docker image)
- Light to run: 70 (700 MB minimum RAM)

## Links

- GitHub: https://github.com/collabora/WhisperLive
- Unraid: https://ca.unraid.net/apps/whisperlive-gpu-0dgm9mx0gapea5?q=WhisperLive---GPU

## Common questions

### Which Docker image should I use if my server does not have an Nvidia GPU?

Use the standard CPU image instead of the GPU build, which is optimized for Nvidia GPUs with CUDA.

### Does WhisperLive provide live transcription?

It is described as a nearly-live implementation of OpenAI's Whisper.

### How can a self-hosted Whisper setup support near-real-time transcription from a browser?

One user built a websocket server that streams audio from a JavaScript front end to Whisper and added a voice-activity detector to improve accuracy.

### Why might Whisper not run consistently on an Nvidia GPU even if it worked before?

A user reports that Whisper worked on a Linux laptop but did not run consistently on a Windows system.

---
Source: Our Local Host (https://ourlocalhost.com/app/3479/whisperlive-gpu), the self-hosted and open-source app directory.
