Repeating segments being removed.

[radzionc]

I'd like the output to retain all segments from the input, but it seems some recurring portions are being omitted.

Here are the segments from the output.

2:  Upon acceptance, it progress.
3:  Stop.
4:  To reveal age-cowd model. Instead of leaning on the traditional switch case architecture,

But in the input, the fourth segment starts with the same text as the second segment.

4:  Upon acceptance, it progress to reveal age-cowd model. Instead of leaning on the traditional switch case architecture,

[jianfch]

What were the arguments you used?

[radzionc]

@jianfch only defaults of model.transcribe

import stable_whisper

def transcribe(input_path):
  model = stable_whisper.load_model('base')

  result = model.transcribe(input_path)

  return result

[radzionc]

In context, my aim is to edit a video recording for my development-focused YouTube channel. I want to eliminate pauses and segments where I misspoke. To streamline this process, I've been saying "Stop" after any misstep, hoping that this would make it easier to programmatically identify and remove the sections leading up to the "Stop" marker.

[jianfch]

The model has the tendency to omit repetitions. You can try temperature=0. If you only need to locate "stop", you also try locate(). See https://github.com/jianfch/stable-ts#locating-words.

# count=0 means look for all
model.locate(input_path, ' stop', 'English', count=0, verbose=True)