zlacker

I've done some preliminary testing with Z-Image Turbo in the past week.

Thoughts

- It's fast (~3 seconds on my RTX 4090)

- Surprisingly capable of maintaining image integrity even at high resolutions (1536x1024, sometimes 2048x2048)

- The adherence is impressive for a 6B parameter model

Some tests (2 / 4 passed):

https://imgpb.com/exMoQ

Personally I find it works better as a refiner model downstream of Qwen-Image 20b which has significantly better prompt understanding but has an unnatural "smoothness" to its generated images.

>>vunder+(OP)
So does this finally replace SDXL?

Is Flux 1/2/Kontext left in the dust by the Z Image and Qwen combo?

>>echelo+A
SDXL has been outclassed for a while, especially since Flux came out.

>>trippl+I2
Subjective. Most in creative industries regularly still use SDXL.

Once Z-image base comes out and some real tuning can be done, I think it has a chance of replacing it for the function SDXL has

>>vunder+(OP)
On fal, it takes less than a second many times.

https://fal.ai/models/fal-ai/z-image/turbo/api

Couple that with the LoRA, in about 3 seconds you can generate completely personalized images.

The speed alone is a big factor but if you put the model side by side with seedream and nanobanana and other models it's definitely in the top 5 and that's killer combo imho.

>>echelo+A
Yeah, I've definitely switched largely away from Flux. Much as I do like Flux (for prompt adherency), BFL's baffling licensing structure along with its excessive censorship makes it a noop.

For ref, the Porcupine-cone creature that ZiT couldn't handle by itself in my aforementioned test was easily handled using a Qwen20b + ZiT refiner workflow and even with two separate models STILL runs faster than Flux2 [dev].

https://imgur.com/a/5qYP0Vc

>>aeon_a+33
Source?

>>vunder+(OP)
> It's fast (~3 seconds on my RTX 4090)

It is amazing how far behind Apple Silicon is when it comes to use non- language models.

Using the reference code from Z-image on my M1 ultra, it takes 8 seconds per step. Over a minute for the default of 9 steps.

>>vunder+(OP)
China really is keeping the open weight/source AI scene alive. If in five years a consumer GPU market still exists it would be because of them.

>>amrrs+f4
I don't know anything about paying for these services, and as a beginner, I worry about running up a huge bill. Do they let you set a limit on how much you pay? I see their pricing examples, but I've never tried one of these.

https://fal.ai/pricing

>>Scrape+Lf
Most of the people I know doing local AI prefer SDXL to Flux. Lots of people are still using SDXL, even today.

Flux has largely been met with a collective yawn.

The only thing Flux had going for it was photorealism and prompt adherence. But the skin and jaws of the humans it generated looked weird, it was difficult to fine tune, and the licensing was weird. Furthermore, Flux never had good aesthetics. It always felt plain.

Nobody doing anime or cartoons used Flux. SDXL continues to shine here. People doing photoreal kept using Midjourney.

>>venuse+Vv
It works with prepaid credits, so there should be no risk. Minimum credit amount is $10, though.

>>vunder+(OP)
If that’s your website please check GitHub link - it has a typo (gitub) and goes to a malicious site

>>soonti+VD
Thanks for the heads up. I just checked the site through several browsers and proxying through a VPN. There's no typo and it properly links to:

https://github.com/Tongyi-MAI/Z-Image

Screenshot of site with network tools open to indicate link

https://imgur.com/a/FZDz0K2

EDIT: It's possible that this issue might have existed in an old cached version. I'll purge the cache just to make sure.

>>vunder+AE
The link with the typo is in the footer.

>>rprwhi+uF
Well holy crap - that's been there for about forever! I need a "domain name" spellchecker built into my Gulp CI/CD flow.

EDIT: Fixed! Thanks soontimes and rprwhite!

>>tethys+YB
This. You can also run most (if not all) of the models that Fal.ai directly from the playground tab including Z-Image Turbo.

https://fal.ai/models/fal-ai/z-image/turbo

>>tarrud+Cv
The diffusion process is usually compute-bound, while transformer inference is memory-bound.

Apple Silicon is comparable in memory bandwidth to mid-range GPUs, but it’s light years behind on compute.

>>nialv7+Rv
Pretty sure the consumer GPU market mostly exists because of games, which has nothing to do with China or AI.

>>p-e-w+7I
> but it’s light years behind on compute.

Is that the only factor though? I wonder if pytorch is lacking optimization for the MPS backend.

>>echelo+A
SDXL has long been surpassed, it's primary redeeming feature is fine tuned variants for different focus and image styles.

IMO HiDream had the best quality OSS generations, Flux Schnell is decent as well. Will try out Z-Image soon.