Adobe Project Clean Take makes it easy to change what people say.

[Jim Feeley]

Project Clean Slate appears to let you easily do stuff like change speakers' inflections and words, do some extreme (demo) noise reduction, break a single video/audio track into stems, and change the emotion of VO. It looks like a late-stage research project right now, but I'll guess we'll see at least some of these features in Adobe apps in 2026. Not the only game in town, but Adobe's demo makes the tasks look simple and they can get this stuff into the hands of lots of people...

 

"Editing dialogue just got smoother. Project Clean Take uses AI to correct mispronunciations, isolate voices, remove noise, and refine delivery—all in seconds. It’s a powerful assistant for podcasters, filmmakers, and anyone seeking studio-quality sound without the studio."

 

A nine-minute video of a demo presented last night:

 

[Philip Perkins]

Another opportunity for overuse of dialog fix tools before the tracks end up in a mix, where it is usually discovered that given the rest of the track that much processing was not needed for the dialog.  As an RRM of mostly verite documentaries I dislike nearly all of what is described.  Mispronunciations and vocal inflections are part of how a real individual person (ie not an AI) communicates in their unique way. The acoustic environment they are speaking in is part of their story.  And over and over I have the issue of production sound or editorial folks improving location dialog into a ruin because A: they are not hearing that dialog in the context of an eventual mix and B: when you have a (new) hammer in your hand everything looks like a nail.  When I get tracks that have been victims of this kind of "one and done" or "one size fits all" NR passes I usually find that really only the BG noise has been dealt with--they haven't done the nitty gritty work of working over clothing noise, tongue drops, nose whistles, chair creaks, crew footsteps, nearby car door slams etc etc, they've just done the easy stuff.  If you are working on a dramatic show shot in a modern 360 virtual set with monitor and camera fans, crew noise, HVAC, prop noise etc etc then by all means carpet bomb your production DX with this kind of thing.  But if you are trying to tell a story about a real person in a real environment leave the dialog the hell alone.

[The Immoral Mr Teas]

1 hour ago, Philip Perkins said:

... leave the dialog the hell alone.

 

+1 to everything Philip has to say, but happy to quote just this last bit.

 

Rather than (potentially useful) tools to allow a craftsperson to do their craft this will indeed just be yet another one-stop-shop for 'fixing' that dialog - - bad location? bad recordist? bad performance? not any more! - - and 'helping' (bad) productions and (bad) producers achieve a whole new set of self- congratulatory aims.

 

As for adobe, I'm reading an awful lot of pdfs at the moment and having used 'acrobat' for as long as I can remember for the last few days I've started opening them in an all-new-song-n-dance version which tells me everything I read looks long and complicated and would I like it to guess the important bits for me? I guess I can probably turn all these pop-up crap things off if I stopped to work out how but I've taken the easier route for now and just try to remember to open them in another less irritating program.

 

Fingers crossed though our more 'professional' software doesn't all go down the 'sit back I'll do that for you' route too.

 

J

[Philip Perkins]

Be sure that Avid et al are working on this sort of thing every day.

 

[Johnny Karlsson]

Well, I am probably not the only one who has noticed that everything always works perfectly in these demo presentations.

But - when you try it on a real thing, it never really works great, or takes a lot of manual tweaking to get a decent result.

(Anyone remember how great Siri worked years ago in a presentation…?)

 

 

 

 

[Lucas]

Maybe in a few years time it will be able to add rustle, door slams and mic handling noise as if it was recorded by a real sound person.

[Olle Sjostrom]

I second everything everyone said. I just want to add that I think AI processes makes us all just lazy and stupid, and this is sometimes the results of that, for real: 

 

[berniebeaudry]

I was talking to a crew that was shooting some testimonials in the corridor outside of a college basketball game.  No one was wearing headphones or even ear buds.  Being a location sound person I asked them about the high level of noise from the people in the corridor in addition to the crowd roar and pa inside the venue.  They had no concern at all and even cited a project they shot previously inside a sporting venue with loud music playing over the pa.  They were able to take the music out in post so they had no worries about the current location.

[Derek H]

This is how AI hurts us. On many smaller jobs where a skilled recordist would have been understood as essential to pull useable dialog out of a bad situation now maybe it’s just a camera mic or a clip-on with a run through whatever plugins and it’s good enough.  Producer pockets the ~$1000 they would have paid a sound mixer and gets a feather in their cap for coming in under budget. 

[Philip Perkins]

I'm sure that the editors of that testimonial project had whatever NR feature came with the last update of their NLE app.  They would be very unusual indeed if they had made a greater effort than that, as in buying and learning the NR etc plugins we use in doco audio post these days.  Wishful thinking distorts reality for a lot of people.  I have clients who send me verite docs that they believe "sound fine" saying they just want me to do a few tweaks, when in fact the soundtrack is an unlistenable, non-QC-passable mess that would cause viewers to bail on the show early after constantly working the volume control on their remote.   Some of these folks aren't listening closely, some of them don't think those audio issues are a cause of concern, and a lot of them are just plain cheap and lazy.

[Izen Ears]

Yep. And no one will care. I blame YouTube.

[Philip Perkins]

3 hours ago, Izen Ears said:

Yep. And no one will care. I blame YouTube.

Oddly, I've found that they start to care if they get into a festival and the film is shown in a theatre, with a theatre sound system playing at theatrical levels.  But that kind of screening is far from assured for most indies anymore.