Fi prefer to use latest dev branch
Ayes im using latest.. coulndt find the save method.. got it now
FStable Diffusion XL (SDXL) vs PIXART-α (pixart-alpha github io) with my auto installer modified and improved gradio
PIXART-α is definitely better and only OSS…
PIXART-α is definitely better and only OSS…
BI asume you didn't train yet with Pixart? They just updated their repo and now collaborate with the LCM Team
(
New) Nov. 30, 2023. 
PixArt collaborates with LCMs team to make the fastest Training & Inference Text-to-Image Generation System. Here, Training code & Inference code & Weights & Demo are all released, we hope users will enjoy them. Refer to docs for more details. At the same time, we update the codebase for better user experience and fix some bugs in the newest version.
(
New) Nov. 30, 2023. PixArt collaborates with LCMs team to make the fastest Training & Inference Text-to-Image Generation System. Here, Training code & Inference code & Weights & Demo are all released, we hope users will enjoy them. Refer to docs for more details. At the same time, we update the codebase for better user experience and fix some bugs in the newest version.Fi havent trained yet
Fthey ddont have any example how to train lol
KPixart is fast! Same time to render a 512x512 as 2048x2048.... Wierd!
KAlways seem to get this "dotted" images though
Fuse my auto installer i have no such issue
KI have used your installer
Fi made amazing comparison list here
FFwhat is your prompt i can try
KFord mustang driving fast down a runway
KJust tried something
KNothing fancy, nothing well thought through
KSA-Solver almost seem better?!
KI don't get how it works 
512x512 takes the same time as 2048x2048 and produces the same image
512x512 takes the same time as 2048x2048 and produces the same imageKAlso resolution has NO impact on VRAM
KThis is somewhat impressive
Fye this prompt is broken :d
F
Faccurate
Ftry this prompt
Fbeautiful, vibrant phoenix portrait in the dark sky spreading wings, highly detailed, hyper realistic, best quality
K
Fi think because when you increase resolution
Fit still generates 1024x1024
Fand upscale haha
Kbut when you generate 512... It downscales?!
KIt takes the same time, so
Fi didnt test 512 yet
Fye it doesnt generate higher res
KIt's the same, just better quality
KExplains things
Ftesting right now
KI really, REALLY wish we had access to Midjourneys upscale function!
K256x256 didn't downscale though
KHorrible quality
K512x512 is really sharp though
Fthat would be nice
KSDXL Turbo then? What's your impressions
KI have seen some models already out on Civitai
KI think speed is not the biggest problem we have though. It's quality
Fit is lower quality
KSo when do we get models with higher quality I wonder. Focus seem to be on speed
AI'm trying to do an exorcist style image, with the girl kind of upside down, but the face wont attach or make the same pose, any advice on how to do this
KI'm impressed with PixArt, but I can't really use it for anything, so I'm sticking with A1111 and SDXL for now
KStabel Diffusion sucks at up-side-down!
KTry in Dalle-3 and then inpaint in A1111