Silero

Specify input files in wav/weba/webm/opus/vorbis format and it will transcribe them to txt in an optional output directory.

Dependencies

• libonnxruntime
• libopus

AI model

• https://github.com/snakers4/silero-models.

Download the AI model using models/en/download.sh.

What to expect

• installing libonnxruntime is a pain, but tract can't load the model currently
• the matroska library is poor and will fail to parse many valid webm files
• only english is currently supported

License

Apache-2.0 + MIT