Skip to content

Latest commit

 

History

History
85 lines (62 loc) · 2.14 KB

WHISPER_NOTES.md

File metadata and controls

85 lines (62 loc) · 2.14 KB
Raw

Running Whisper

The following notes have little to do with whisper_server, but might be of interest if you want to run Whisper directly.

Whisper

Follow the setup instructions at Whisper on GitHub:

  • install Python
  • pip install -U openai-whisper

Linux

sudo apt update && sudo apt install ffmpeg

Windows

choco install ffmpeg

Whisper in Docker

docker build -t nalbion/openai-whisper .
docker run --rm -e "PULSE_SERVER=/mnt/wslg/PulseServer" -v /mnt/wslg/:/mnt/wslg/ nalbion/openai-whisper

Whisper.cpp on the Web

whisper.cpp can also be run in the browser.

The online examples are not too impressive on a laptop.

Whisper.cpp in Docker

Memory usage

Model Disk Mem
tiny 75 MB ~125 MB
base 142 MB ~210 MB
small 466 MB ~600 MB
medium 1.5 GB ~1.7 GB
large 2.9 GB ~3.3 GB

Build

Choose a model from the table above and run:

MODEL=base & LANG=".en" & docker build -f Dockerfile.cpp -t nalbion/whisper-cpp-$MODEL --build-arg MODEL=$MODEL$LANG .

or on Windows:

set MODEL=base& set LANG=.en& docker build -f Dockerfile.cpp -t nalbion/whisper-cpp-%MODEL% --build-arg MODEL=%MODEL%%LANG% .

Run Whisper

docker run --rm -v /dev/snd:/dev/snd nalbion/whisper-cpp-base

or on WSL:

docker run --rm -e "PULSE_SERVER=/mnt/wslg/PulseServer" -v /mnt/wslg/:/mnt/wslg/ nalbion/whisper-cpp-base

or for an interactive shell:

docker run --rm -it -v "$(pwd)/build:/artifacts" --entrypoint /bin/bash nalbion/whisper-cpp-base

or from Windows:

docker run --rm -it -v %cd%/build:/artifacts --entrypoint /bin/bash nalbion/whisper-cpp-base

From the shell you can copy the wasm files:

cp -r bin/stream.wasm/* bin/libstream.worker.js /artifacts/wasm