Whisper audio import tips?

I’m getting an error trying to import an MP3 I downloaded from YouTube. It’s a Korean girl talking for about half an hour about her plans to study abroad in Germany. It seems kind of interesting but I need to read it to understand it :joy:. Any tips for importing audio and avoiding the errors and failed import messages?

1 Like

mp3 files causes errors if they are too big (more then 60MB 200MB) or if the bitrate is odd. You can transcribe the mp3 using faster whisper xxl, for example.

You can try this website to make the conversion. It’s the one I use.

If it is a bitrate problem, this one will correct that issue. If it’s too big, as Obsttorte points out, that’s another issue.

2 Likes

@Obsttorte FYI, the audio file limit size has been increased to 200MB, we changed that a few weeks ago.

@kraemder What error message do you see?

4 Likes

Good to know. Seems I missed that.

Just an import failed message. I got it to work. The original file size was 30 MB. I used Audacity to shrink it down 8 MB and that got it to go through. I don’t think the system can handle anything near 200 MB.

1 Like

The file size definitely wasn’t the issue in this case, sounds more like some issue with that specific audio file.

2 Likes

I wonder what the issue is then? Do you have any ideas? The original file I downloaded from YouTube plays fine in my MP3 player.

1 Like

What is the bitrate? Did you try to convert it (from mp3 to mp3, sounds odd, but might get rid of anomalies)?

It was 128kbps MP3 and I converted it to 35kbps MP3. I suppose it may have gotten rid of an anomaly but I’m wondering if there’s any examples of anomalies that cause problems. Maybe I don’t have to reduce it to 35kbps next time.

1 Like

I would have expected the 128kbps version to work better. 35kbps is an odd bitrate, and such a reduction might not work well for the transcription, too. Did the converted file work?

In regards to “anomalies” I was thinking of the codec and the heartrate used. Not sure whether there are more properties that can differ among mp3 files.