trentor

Did you read the paper? Because it doesn't look like it. There are actionable lists for all examples. Everyone with basic reading comprehension can do it.

You are exaggerating so hard it's impossible to take you serious. People can click a link on a website without hyperventilating.

You can just look at the airflow charts and buy accordingly?

It started at where2 and Google acquired them.

they aim for four nines nowadays.

Crazy. A company that actually exits the market they threatened to leave. Meta and Twitter should take note.

What moat?

I think the early models distorted the perception of what artists actually need. It's not about workflows. Image creation is inherently iterative. None of the platforms seem to understand that, while AI artists have somehow fallen in love with the unnecessary complexity of node-based tools. Apparently, 400 tangled noodles look impressive when the boss peers over your shoulder.

The big studios and agencies are all building custom tooling as they always have but now a lot faster.

Krea2 was released because the open model was an interims model. The hosted krea2 used the flux2 vae and open weights qwen1.

I would be with you if they didn't change the owner to private equity in the last year.

doubt.

Fedora

It was probably just a json with settings.

What trend? BFL was always non commercial. The "only" trend would be qwen image not being apache anymore.

Refreshing seeing a thread about a new model without talking about pelicans.

that's microfuchsia.

I hope they got their inference under control. Gemini has a lot of "overloaded" hiccups.

gov.uk is something completely different?

Looked romantic in Fight Club tbh.

Hypothetically how would you implement it? Every dev and their mom would be interested in a way to force people to enter their actual email address instead of whatever they want...

In all that the one thing I don't understand is how they can just change their product for existing customers without even an email. Must be illegal.

It’s an article about us forces leaving Irak seems reasonable to talk about how many they lost during the time they were there…

It's not the guy. It's the guys he attracts.

See I will never use anthropic because they run inference on spacex. Wat den een sien Uhl, is den annern sien Nachtigall.

Looks like they have still problems with caching. Prize is double the other providers for cache hits... which is most of what I do. :/

Sad but I have to say they just stood still for 20 years and tried catching up in the last 5 that never works out.

No, it's probably their rope implementation. They had a similar problem with the old qwen image but to be sure it needs some digging.

It's right there in the hugging face link?

latents go from 16ch @ 8x compression to 64ch @ 16x, so roughly the same total latent budget but much more channel heavy. It’s also deeper/wider, and the old 2x2 transformer patching is gone.

On some images it still produces artifacts but can't say if it's the transformer or the VAE yet.

They always had a fourier space mark in their models even without the VAEs are usually pretty easy to detect.

They finally fixed their VAE. It really held back their models over the last 2 years.

EDIT: It still produces artifacts it's better but unusable for production work. In midvalues you will see a slight dot pattern.