This would set SL7 apart from the rest. feature req

I don’t own spectral layers because I use RX but I thought id put a post here on a feature that would set SL apart from everyone else.

As a full-time dialogue editor (mainly audiobooks) part of my job is to remove breaths. There is no tool out there in the market that does it properly. I have used them all inc rx de breath, waves de breath, gates, etc. None of them get it right, they can get you close but they cut off words and if you want to do it right, you still need to do a manual review.

What would be great, is if you can use the function to find similar events in SL. So you highlight or take samples of some breaths in the recording, and SL finds all the breaths in the recording. It would then select all entire breaths and show them in the SL editor,

then SL, goes through all the breaths it finds 1 by 1 so you can review and adjust the start and ends. Once they are all confirmed, you can then select a function to remove or reduce the volume.

no company is doing this, if you pull it off you will be saving people hours, and it would make people like me jump to SL for this purpose.

In other words, do a manual review. No breath is going to be exactly like another, and something would, inevitably, be mistaken for a breath by your automated feature, so there would be no way to automate the de-breathing the way you have in mind. You might as well use the standard RX NR module and do things one at a time.

Great idea - I have been looking for a good breath removal tool too.

In the meantime, I found that the approach outlined in this video works well. Using noise reduction and Auto Duck in Audacity, this approach works well to avoid cutting off the ending and beginning of words.

https://www.youtube.com/watch?v=emyGcinHUu0

A fricative could easily be mistaken for a breath by AI. Once removed, it would be impossible to replace, except by re-recording an entire section of the audio file. The only way to make sure AI hadn’t made a mistake in its diagnosis would be to review things the hard way.

With enough practice, it becomes easy to outsmart just about any AI feature in an audio program, unfortunately.

Poinzy,

If I use an aggressive method to remove breathing (like the one I mentioned in my last post) the problem of removal of a fricative does happen. So I keep a copy of the original track just above the one with breathing removed. I sync-lock them together and as I listen to the track with the breath removed in order to do other editing, if I notice anything that appears to be missing, e.g., beginning or ending of a word, or maybe a fricative, I just pull it down from the original track (with a macro) which avoids the problem of re-recording.

That has worked well for me.

nonsense, because izotope rx has a tool to find similar events with a probability slider, if you select a breath, it does a 90-95% accuracy job of locating them all, but what they don’t do and what i was asking for is, a program that will collect all the breaths, and show you them in a window 1 by 1. that way you can quickly review and accept or decline, modify the start and ends and then at the end, either select delete or reduce in volume to all reviewed areas, i think dude youre best not replying here, im fairly advanced doing this full time and know what im talking about, and your suggestions are limited but thanks anyway.

Thanks for handling constructive criticism so well. It’s very advanced and professional of you.

this is interesting! thank you

That’s interesting. I’ll keep this in mind. Thanks. (Oops, I mean, you pleb, don’t you know who I am??)