Whisper
Installation⚑
Usage⚑
Extend the whisper with your custom words⚑
There many solutions:
- Tweak the prompt
- Train your own model
- Postprocess the result to match your desired output
Tweak the prompt⚑
You can use the --initial-prompt to add a list of words you want to be detected. For example with python:
model = whisper.load_model("base")
result = model.transcribe("audio.mp3", initial_prompt='Tensorflow pytorch' )
Then the model is more keen to detect those keywords
Train your own model⚑
Follow this tutorial