Repeated sentences in result text

[redhumus]

Hi, I'm using stable-ts to translate audio to text, but my use case isn't for subtitles, is for social research. But I'm having repeated sentences. What I'm doing wrong to have this issue?

This is the code I'm using:

import stable_whisper

model = stable_whisper.load_model('medium')
# modified model should run just like the regular model but accepts additional parameters
result = model.transcribe('my-file.m4a', regroup=True)

# srt/vtt
result.to_srt_vtt('my-transcription.srt')
print('Fin del proceso')

Thanks

[jianfch]

You can try the solution in openai/whisper#192.

[redhumus]

Thank you for your response. I checked the link and .srt file again and found that the result seems repeated but it is highlighting each word to be used in subtitles. Then, it must be a lack of clarity on my part about the use of the tool.

Then, I would like to confirm:

1. The solution is still to try lower --compression_ratio_threshold . How can I do this in a python script not in commandline?

2. Maybe there is another parameter I can change in my python script to modify the result to not export it for subtitles but for a plain text without repeating for highlighting text?

Thank you

This is an example of the .srt file exported:

0 00:00:00,200 --> 00:00:00,480 <font color="#00ff00">Bueno,</font> o sea yo le pregunto cómo es su proceso de liderazgo.

1 00:00:00,480 --> 00:00:00,500 Bueno, o sea yo le pregunto cómo es su proceso de liderazgo.

2 00:00:00,500 --> 00:00:00,640 Bueno, <font color="#00ff00">o sea</font> yo le pregunto cómo es su proceso de liderazgo.

3 00:00:00,640 --> 00:00:00,760 Bueno, o sea <font color="#00ff00">yo</font> le pregunto cómo es su proceso de liderazgo.

4 00:00:00,760 --> 00:00:01,000 Bueno, o sea yo <font color="#00ff00">le</font> pregunto cómo es su proceso de liderazgo.

5 00:00:01,000 --> 00:00:01,240 Bueno, o sea yo le <font color="#00ff00">pregunto</font> cómo es su proceso de liderazgo.

6 00:00:01,240 --> 00:00:01,560 Bueno, o sea yo le pregunto <font color="#00ff00">cómo</font> es su proceso de liderazgo.

7 00:00:01,560 --> 00:00:02,180 Bueno, o sea yo le pregunto cómo <font color="#00ff00">es</font> su proceso de liderazgo.

8 00:00:02,180 --> 00:00:02,280 Bueno, o sea yo le pregunto cómo es <font color="#00ff00">su</font> proceso de liderazgo.

9 00:00:02,280 --> 00:00:02,640 Bueno, o sea yo le pregunto cómo es su <font color="#00ff00">proceso</font> de liderazgo.

10 00:00:02,640 --> 00:00:02,860 Bueno, o sea yo le pregunto cómo es su proceso <font color="#00ff00">de</font> liderazgo.

11 00:00:02,860 --> 00:00:03,280 Bueno, o sea yo le pregunto cómo es su proceso de <font color="#00ff00">liderazgo.</font>

[jianfch]

To remove the highlight:

result.to_srt_vtt('my-transcription.srt', word_level=False)
# or
result.to_srt_vtt('my-transcription.srt', segment_level=False)

[redhumus]

Thank you for your help, that worked for me