Generate VTT file from speech to text

Question

I was able to generate text from an audio file using huggingface, using this code

transcriber = pipeline(
  "automatic-speech-recognition", 
  model="fsicoli/whisper-large-v3-pt-mzcv16"
)

transcriber.model.config.forced_decoder_ids = (
  transcriber.tokenizer.get_decoder_prompt_ids(
    language="pt", 
    task="transcribe"
  )
)

transcription = transcriber("half0.mp3")

Is there a way to output a vtt file? I get the transcription but not the timestamps

Answer 1

Have you looked at using the return_timestamps keyword? It seems you can also get the timestamps on either the word on text segment level (for the Whisper model at least).