New Technologies (AI)

Just for anyone who might be interested to give Kaedim a go: You have to pay even for the trial, and it’s one of those borderline scammy deals:

“The 6$ trial covers 1 generation per day for 3 days on any chosen plan: Light, Pro or Startup. At the end of the 3-day period, your trial will automatically convert to the chosen plan unless cancelled. Advanced features not included.”

The cheapest monthly price is $539!!! And that limits you to 30 models / 12 iterations.

Ergo there isn’t even a free trial so one could thoroughly assess the quality of the models this creates to see whether this might be worth becoming a part of one’s standard workflow.

3 Likes

Wow that’s insane!

https://mobile.twitter.com/AKubrakiewicz/status/1560163966562799616

https://mobile.twitter.com/andrewpprice/status/1560140616126066689

2 Likes

Thank you piranha4D. I’ll wait a couple of days, if they don’t defend themselves in a credible manner i will delete my post.

Edit: i decided it is better to have a warning than to delete. That way the bad information is not lost.

4 Likes

Yeah, not just “borderline scammy”, but seriously scammy. Mechanical Turk more than AI. Here’s an article on Medium by the company founder in reaction to the suspicions, claiming that because their AI “is far from perfect”, they have “quality control engineers” go over the models to fix them up: https://medium.com/kaedim/response-from-kaedim-re-our-ai-931d3ef39c33

Uh. Right. That old job posting is pretty damning, that’s not looking for “quality control engineers”, but for people who can crank out simple models in, surprise, the exact amount of time it now takes for a user to get the result. I’ve also since seen some of the wireframes, and they don’t look computer-generated, they show huge differences between different models, which is another indication that they are made by different humans.

I mean, we’re definitely at the doorstep of this being possible for real, but this company doesn’t inspire me with confidence because they’re not transparent, no researchers have access to the actual technology, and this smells like a potential rip-off. Maybe they really are into ML, and they really have an AI, but they’re faking it until they’re making it, to get at money.

@Bullit, yes, I agree, having a warning is better than deleting it.

3 Likes

Hi, a new denoise and more from RAW images with AI.
The article is in German but the video is stunning:

Cheers, mib

2 Likes

Reminds me of that ad campaign, Life’s too short for the wrong job :


What a miserable venture anyway

3 Likes
3 Likes

Unemployment would become a complete non-issue at least, if only society could overcome the challenge of getting people to actually sign on to such a job (to eliminate repetitive, boring, or otherwise dangerous positions people do not want to take anyway is one reason why we have machines).

Yes, no. Unemployment is only a problem because we, collectively, can’t be ****ed to redistribute wealth. If only society could overcome the whims of the few, more like

4 Likes

I meant that in jest actually, but recent years has actually seen a growing luddite movement amid the explosion in AI-powered technology (from what we are discussing in this thread to applications for the smart home).

The calls to remove the automated looms (in order to bring back the seamstress) is long gone, but the anti-technology foundation is still there. In this case, it will be calls to preserve the jobs of photo retouchers, technical directors, texture artists/processors, modelers, rotoscopers, ect…, no one should have the ability to make a Pixar-level movie or AAA-level game without one hundred people apparently.

The loss of manufacturing jobs in the US was driven primarily by automation, despite what the ‘they’re taking our jarbs’ crowd would have you believe.

Until there is a way for people’s needs to be met, finding ways around industry’s need for people will only result in poverty.

3 Likes

Dall E again, but look how the cloths can be changed in an animation albeit simple.

https://twitter.com/karenxcheng/status/1564626773001719813

3 Likes
2 Likes
4 Likes

Oh, they released! I didn’t pay attention; that was faster than I expected. Thanks for your news keeping me updated even when I am lazy!

From the release announcement:

We hope everyone will use this in an ethical, moral and legal manner and contribute both to the community and discourse around it.

Oh the sweet summer children. :wink:

3 Likes

Thanks for posting. Can you upload a render to it for example and it generates variants of that image based on prompts?

Sorry, i did not saw the tutorials.

Yes, you can use an image as input.

Here’s somebody showing examples. More can be found by doing an image search with “stable diffusion img2img”.

2 Likes

Ok. No worries, man :slightly_smiling_face:

Thanks man :+1:

1 Like