How to fix a transcript that is repeating text in the output that is not repeated in the source audio?

#4
by vfresh - opened

Anyone have a problem with this model repeating text fragments starting at a random point?

Perhaps I need to change a parameter ??

Here is the an example output. The EXAMPLE SENTENCE is said one time from [02:49.140 --> 02:57.140] but is repeated 3 times in the output with subsets of the time stamp listed.

[02:49.140 --> 02:55.140] EXAMPLE SENTENCE.
[02:55.140 --> 02:55.640] EXAMPLE SENTENCE.
[02:56.140 --> 02:57.140] EXAMPLE SENTENCE.

mlx_whisper with --condition-on-previous-text False is helpful for me.

mlx_whisper with --condition-on-previous-text False is helpful for me.

I would just like to add that this seems to have resolved the repeating issue for me also. Thank you for the help!

Your need to confirm your account before you can post a new comment.

Sign up or log in to comment