I used the workflow where Krea2 rendered the draft in Text--to-Image. SDXL model used as a textures refiner. Worked well for me some time until I've trained the Krea2 with good textures.
I think tandem of Qwen Image 2.1 with Krea2 or SDXL might be viable and usable for some cases.

skill issue, change CFG to 3.5

Explain how it's the best edit model? The Qwen VAE is bad comapred to the Flux 2 VAE

@elevendr lol, Qwen 2.1 doesn't use the shitty Qwen 1 VAE... But I'm sure you knew that, right?!

After spending a long time messing around with K2 edit, i can actually say it's pretty powerful. It just takes a really long time to figure it out. So ease of use .. no .. but it is very capable.

How does it compare to Klein9b for edits?

>Nothing special for other purposes.

You mean like EDIT that it's supreme at?

@wkwjhskssksbg465 AI-edit is a very special topic, sometimes old good Photoshop much simplier and better :)

@mphobbit bro stfu, with a ton of skill no one is photoshoping the realism edit's I'm making with one hand.

@wkwjhskssksbg465 why are you so nervous XD? We just have different tasks. In your tasks AI editing is supreme, in my tasks it shows satisfactory results, that's all. Will I use it? Yes, but as one among others. For me nothing is better than straight hands and watchful eye.

"Nothing special for other purposes." - mphobbit

What a buffoon. Completely clueless, tone deaf comment.

It's shills and fanbots I think.

I'm not deleting Krea2, but this model is insane for edit and reference.

@wkwjhskssksbg465 It's good to have a variety of models for different tasks.

Anima for example, is real (fun:3) if you enjoy (the process of:2.5) really strange (prompting:-1.5) to get the desired wacky outcome. A little puzzle-solving exercise with keywords.

Those are all SUPER AI and plastic. Can you point to ming doing photographic?

Ming Image gives me major ERNIE‑Image v2 vibes...

The edit capability is also a separate checkpoint (unlike Qwen 2.1). It'd probably be generous to say it's a generation behind Qwen Image 2.1

Like ERNIE, it seems like Ming have chased a bunch of score metrics without really thinking so much what the actual output is like beyond that. & it shows.

The Text Encoder int8 is also 18.17 GB (in addition to the verbose prompting/their LLM you kind of need to go along with it). Frankly speaking; nothing I've seen on the Ming image page looks impressive; & what I've generated isn't comparing favourably to the competition either. I'll test further to see the maximum it's capable of, but I don't think it's going to favour well (unless you like the fake style).

@ShadowCell One point for the Ming Image is that is compatible with Z-image LORAs

The shills and emotionally attached to old model people almost got you! Never listen to anyone when a new model comes out.

Most haters = "I can't run it well, therefore it sucks and I don't want anyone else to use it either". It's either jealousy, or a skill issue lol.

What I'm generating so far (& it's early days); is dramatically more realistic than even something like ChatGPT is capable of (a massive cloud model you have to pay for....). In that context, Qwen image 2.1 is insanely good, & an immense technical feat. The edit capability just absolutely blows everything else out of the water, too.