# Faster-whisper

Faster-whisper is a reimplementation of OpenAI's Whisper speech-to-text model built on CTranslate2, a fast inference engine for Transformer models. It is open source and self-hosted, installable on ZimaOS, Docker Hub and Unraid.

Canonical page: https://ourlocalhost.com/app/1062/faster-whisper

## Description

Faster-whisper is a reimplementation of OpenAI's Whisper speech-to-text model built on CTranslate2, a fast inference engine for Transformer models. This container runs faster-whisper as a Wyoming protocol server, delivering up to 4× faster local transcription with lower memory usage for voice assistant setups.

## Platform availability

Available on 2 of 9 platforms: Unraid, ZimaOS.

## Health score

74/100 (Good), based on 5 of 5 factors.
- Maintained: 75 (last commit 16 days ago)
- Popular: 88 (25,776 GitHub stars)
- Easy to install: 45 (2 install platforms, install notes, Docker image)
- Light to run: 60 (2 GB minimum RAM, runs on ARM)
- Security check: 95 (no serious findings, CI, tests)

## Links

- ZimaOS: https://github.com/bigbeartechworld/big-bear-casaos/tree/master/Apps/faster-whisper
- Docker Hub: https://hub.docker.com/r/linuxserver/faster-whisper
- GitHub: https://github.com/linuxserver/docker-faster-whisper
- Unraid: https://ca.unraid.net/apps/faster-whisper-18k9dlc1fdj1bw?q=faster-whisper

## Common questions

### How do I install faster-whisper for use in a self-hosted script?

Install the stable faster-whisper package with pip. It provides the transcriber and required dependencies so your program can start and load a model.

### Can I use a pre-downloaded faster-whisper model folder instead of fetching a model name?

Yes, point WhisperModel at the local converted model folder. This makes faster-whisper use your local copy instead of fetching a model name from an online source.

### What model settings should I use on an NVIDIA GPU?

Start the model with device set to cuda and compute_type set to float16. Using those settings on a CPU-only machine can fail or run poorly.

### How can I expose faster-whisper as an API server?

Use an OpenAI-compatible faster-whisper server implementation, which wraps faster-whisper as the transcription backend. This lets you call it through an OpenAI-style transcription API.

---
Source: Our Local Host (https://ourlocalhost.com/app/1062/faster-whisper), the self-hosted and open-source app directory.