Whisper Live installation guide

Note: it is preferable to use a machine with NVidia GPU in order to use the most accurate whisper models. Else on CPU most probably you will be limited to 'base' or 'small' models.

Install Docker in it is not already installed. For Windows install Docker Desktop from here: https://www.docker.com/products/docker-desktop/
If you have NVidia GPU make sure you have installed the latest drivers from NVidia
Install and run Whisper Live:

- No GPU:

Powershell:

docker run `
 --name whisper-live `
 --restart=unless-stopped `
 -e WHISPERLIVE_API_KEY=mysecretkey `
 -e WHISPERLIVE_MODEL=base `
 -e WHISPERLIVE_MAX_CLIENTS=4 `
 -e WHISPERLIVE_MAX_CONNECTION_TIME=300 `
 -e WHISPERLIVE_USE_VAD=true `
 -v whisper-live-data:/var/lib/whisper-live `
 -p 9090:9090 `
 -p 8000:8000 `
 -d hwdsl2/whisper-live-server

Bash:

docker run \
 --name whisper-live \
 --restart=unless-stopped \
 -e WHISPERLIVE_API_KEY=mysecretkey \
 -e WHISPERLIVE_MODEL=base \
 -e WHISPERLIVE_MAX_CLIENTS=4 \
 -e WHISPERLIVE_MAX_CONNECTION_TIME=300 \
 -e WHISPERLIVE_USE_VAD=true \
 -v whisper-live-data:/var/lib/whisper-live \
 -p 9090:9090 \
 -p 8000:8000 \
 -d hwdsl2/whisper-live-server

- NVIDIA GPU:

Powershell:

 docker run `
 --name whisper-live `
 --gpus=all `
 --restart=unless-stopped `
 -e WHISPERLIVE_API_KEY=mysecretkey `
 -e WHISPERLIVE_MODEL=base `
 -e WHISPERLIVE_MAX_CLIENTS=4 `
 -e WHISPERLIVE_MAX_CONNECTION_TIME=300 `
 -e WHISPERLIVE_USE_VAD=true `
 -v whisper-live-data:/var/lib/whisper-live `
 -p 9090:9090 `
 -p 8000:8000 `
 -d hwdsl2/whisper-live-server:cuda

Bash:

sudo docker run \
 --name whisper-live \
 --gpus=all \
 --restart=unless-stopped \
 -e WHISPERLIVE_API_KEY=mysecretkey \
 -e WHISPERLIVE_MODEL=base \
 -e WHISPERLIVE_MAX_CLIENTS=4 \
 -e WHISPERLIVE_MAX_CONNECTION_TIME=300 \
 -e WHISPERLIVE_USE_VAD=true \
 -v whisper-live-data:/var/lib/whisper-live \
 -p 9090:9090 \
 -p 8000:8000 \
 -d hwdsl2/whisper-live-server:cuda

Change the values of the environment as per your needs, for example WHISPERLIVE_MAX_CONNECTION_TIME=300 allow sessions up to 300 seconds (5 minutes), so you may want this to be WHISPERLIVE_MAX_CONNECTION_TIME=3600 for example.
Important ones are the model and API key. For GPU the good results are achieved with large-v3-turbo.

Available models:

Model	Disk	RAM (approx)	Notes
`tiny`	~75 MB	~250 MB	Fastest; lower accuracy
`tiny.en`	~75 MB	~250 MB	English-only
`base`	~145 MB	~700 MB	Good balance — default
`base.en`	~145 MB	~700 MB	English-only
`small`	~465 MB	~1.5 GB	Better accuracy
`small.en`	~465 MB	~1.5 GB	English-only
`medium`	~1.5 GB	~5 GB	High accuracy
`medium.en`	~1.5 GB	~5 GB	English-only
`large-v1`	~3 GB	~10 GB	Older large model
`large-v2`	~3 GB	~10 GB	Very high accuracy
`large-v3`	~3 GB	~10 GB	Best accuracy
`large-v3-turbo`	~1.6 GB	~6 GB	Fast + high accuracy ⭐
`turbo`	~1.6 GB	~6 GB	Alias for `large-v3-turbo`

If you need to start the container with another parameters do:

docker rm -f whisper-live

and then start it again with the new parameters. Note that the downloaded models will be preserver in whisper-live-data volume.

In SubtitleNEXT use the following URL: ws://127.0.0.1:9090/ (if it is on different machine use its IP instead) and for API key the value you have set to WHISPERLIVE_API_KEY ('mysecretkey' in the examples above).

Note: On first use the configured model will be downloaded, so the API will not respond until it finishes. Wait either until it start return results or if connection is closed try again later.

Revision #5
Created 5 October 2026 09:33:23 by Pierre
Updated 5 October 2026 20:30:11 by Pierre