MPSE Cuts Ties With “Once-Strong Partner in Sound”

[Alex H.]

An interesting development this week. A letter went out to Motion Picture Sound Editors members, announcing that they had cut ties with an unnamed organization over a product release related to generative AI. From the letter:

One such product hit the market last week and caused quite a stir (rightfully) in our community. A vigorous and thoughtful examination of this product, particularly its marketing strategy, forced us to break ties with and no longer support a once-strong partner in sound.

While they did not name the “once-strong partner in sound”, there’s wide agreement, speculative as it may be, that this refers to the recent Krotos/Adobe Premier announcement. I’m fairly certain that Krotos is the one that’s been dropped.

I am not a member of MPSE, but this was featured in one of the latest YT videos from “Mastering Post-Production Sound”:

I’ll post my thoughts on Krotos soon.

 

[Jim Feeley]

Ya, I recall the chatter when Krotos previewed/demoed their video-to-sound plugin at NAB a few months ago. Krotos says it doesn't use generative AI, but uses AI to identify what's in a shot and then find and sync appropriate real-world sfx to the shot. Not sure that's a whole lot better for the sound community's livelihood. It's clever though. And I'd guess it'll totally appeal to lots of editors, producers, and clients. This is all just top of my head.

Has MPSE taken a stance about AI-driven noise reduction, EQ, and the like (which arguably is generative as much as manipulative)?

I look forward to seeing your considered thoughts, Alex.

One-minute marketing overview of Krotos' latest:

 

[Alex H.]

Here are my thoughts on Krotos:

I chatted with them at NAB. And yes, they were adamant that Krotos Studio doesn’t use generative AI, but rather uses AI to automate MIDI trigger placement that draws from pre-recorded sound libraries. They also pointed out that users can utilize their own sound libraries.

I still had my reservations about Krotos Studio after speaking with them. It takes the art out of the process. A friend of mine here in Knoxville, another sound designer who works primarily in reality/home improvement made the argument that Foley is used in those shows not to tell the story or help drive it forward, but simply to fill in empty spaces. Further, those shows just don’t (won’t) have the time or budget for proper Foley work, so in those cases something like Krotos Studio may make sense. I bought that argument.

However, this announcement with Adobe betrays all of that. Generative or not, they’re outright spotlighting that this does all the work for you. Enter some prompts, show it the footage, and let it do the rest. Qualitatively, the results still aren’t great. I wish that was enough to make me feel better about it, but it’s not. The bar has been lowered, and it’s the studios and producers that are going to see the bottom line, the cost savings, and decide that lower standards and lower quality are perfectly “good enough” since it doesn’t cost much to achieve.

I was speaking yesterday with another sound mixer friend of mine, out of NYC, and he made a very valid point that the market for video editors who would use this is much more expansive than the market for actual sound designers and Foley artists, so Krotos are going where the money is with no regard for the jobs they’re going to destroy.

I’ve been using Dehumanizer for a while, but I have zero intention of investing any money, or even time, into Krotos Studio. There have been quite a few sound editors posting in the groups on FB that they’re completely uninstalling all things Krotos in response to this announcement. Can’t say I blame them. I’m slightly torn, only because Dehumanizer is basically a composited set of plug-ins (modulators, delays, phasers, etc.) and not an AI-based plug-in. They already have my money on that one.

Still, the direction they’re taking poses serious threats to our sector of the industry.

Where is the art when there is no artist?

 

[Alex H.]

Has MPSE taken a stance about AI-driven noise reduction, EQ, and the like (which arguably is generative as much as manipulative)?

That, I do not know. I will say that plug-ins like Izotope, DXRevive, Clarity, Hush, etc. aren’t generative AI. They’re machine learning, neural network, but they aren’t scraping the ‘Net for the content they spit out. They also aren’t prompt-based. They’re user-controlled down to very finite settings.