Qwen Image is slow

Steps reduced to 10; now 4X+ faster.

2 minutes and 42 seconds on an RTX 5080 with 16 GB VRAM 64GB RAM?

@zzkszzks603 ah, use Lightning then

use lightning lora 8 steps

@vxtamins https://huggingface.co/Wuli-art/Qwen-Image-2512-Turbo-LoRA

Only 3 minutes?

Mine take 4.5-6 minutes (30 steps) on RTX 5070.

considering the model is 39 GB I'd say that's pretty good lol

facts bro, it stupid, i tried this on god damn int4 and the quality and prompt adherence was still better than flux 2 max, chatgpt image or nano banana 2. and it looks better than all of them, especially the hands

This checkpoint is literally so damn good, it actually feels like cheating using it....lol

It turns even low effort, garbage prompts into gold lmao.

I have the thing running right now in the background, & it's spamming out 1584x1072 images one after the other in like 15-20seconds like a golden goose. It's becoming redundant favouriting the good images, because virtually all of them are.

You almost have to actively, & intentionally go out of your way to try & generate bad anatomy on purpose before it'll do it.

PS - I went away, & then came back after posting that last comment...... & in that brief time, it generated another 117 high resolution images; of which, only literally 3 images had minor anatomy defects! All the rest had perfect fingers, hands, etc.

PS, PS - I ended up generating over 2,000 images tonight.....lol. But I've never had so much fun using a checkpoint before. My ribs are literally aching from pain I've been laughing so hard generating so many insane things ^^

@ShadowCell Could you point me towards a decent workflow? Are you using it with the lightning lora?

@Themacsback  https://huggingface.co/Wuli-art/Qwen-Image-2512-Turbo-LoRA/tree/main Wuli-Qwen-Image-2512-Turbo-LoRA-4steps-V3.0-bf16.safetensors

Try V3 - It seems to be quite a step up over others.

(As for workflows) Here's something I posted elsewhere........

"I use a lot of things, but I'm moving more & more over to SwarmUI. It's biggest advantage is that it has a fully working ComyfiUI backend which you can switch over to in 5 seconds (or without even firing Swarm up). & SwarmUI/Comfyi gets the latest/most updates to support new things/optimisations.

Personally, I'm not a fan of all this workflow/node bullshit lol. It's just too time consuming, & clunky most of the time (especially if you need to edit someone else's spaghetti workflow). With SwarmUI, you have the added control of that, or if you don't need it, & just want to change settings quickly, then you have the Forge like UI frontend of SwarmUI you can use instead.

https://github.com/mcmonkeyprojects/SwarmUI If anyone is curious; scroll down & "Download The Install-Windows.bat file". You basically just double click that, & it auto-downloads everything you need in 10mins or less. It's the easiest setup ever. For n00bs; It'll also automatically auto-download any clips or VAE, etc, you need for checkpoints if you have the wrong one, or none at all."

if the model is 19 gb big, you need a minimum of 19 gb of memory. In practice, your operating system needs to run too, so you probably want a 24 gb vram gpu.

Спасибо бро, очень помогло!

Рад помочь, товарищ

Looks like you're posting again (I can see your images in feed in the last 6 hours). Like you, I kind of thought I was shadow banned or something for some unbeknownst reason. Looks like it's civitai servers just being absolute potato shit again though lol.

My images were constantly going missing, or showing as 'not posted', etc.