Reducing interference/distortion over broad spectrum

I’ve looked through several posts here and have seen many good solutions and learned a lot. However I still need to post here because I could not help myself any further in an unfortunate situation:

I need to transcribe this recording (see spectrogram view) which has some horrible interference on a broad frequency spectrum most of the time. So if I listen closely it’s barely to impossible to recognize what’s said - for the full length of this recording it’s ugly to impossible work for me/my ears.

  1. So I isolated unaffected frequency bands but can’t understand enough (too small unaffected bands). Is it possible to isolate multiple unaffected frequency bands?

  2. I also tried noise reduction by identifying a noise profile (where noise only). This reduces the ugly interference significantly but I just could not find parameters and/or noise profiles such that enough of the spoken words remain. Maybe I’m simply too noobish. Do you have any advice?

Give AI a go (while it’s free)

Great idea, thanks! :slight_smile:

After playing a bit with this AI tool I could not achieve desired results…especially the parts that remained impossible to understand vanished completely.

Btw. for a closer look on lower frequencies:

So if any ideas for enhancing this or the noise profile recognition come up, please let me know!

Extracting understandable speech from incomprehensible noise only happens in the movies. The human brain is currently better than AI at comprehending speech: if you cannot understand it there’s little hope of success.

Thank you anyway! :slight_smile:

After playing excessively with the filter curve EQ followed by snipping together several noise only parts for better noise profile identification I got most of it fairly understandable. Quite some time investment is required but if needed Audacity has amazing capabilities! :smiley: