Introducing RunDiffusion XL Beta!
RunDiffusion.com has done it again! One of the worlds first SDXL Models!
Join our 15k Member Discord where we help you with your projects, talk about best practices, post exciting updates and news, and share the amazing art we generate.
Why? Isn't SDXL already amazing?
Yes! We feel this is a step up! SDXL has an issue with people still looking plastic, eyes, hands, and extra limbs. We couldn't solve all the problem (hence the beta), but we're close! We tested hundreds of SDXL prompts straight from Civitai.com (using ComfyUI) to make sure the pipelines were identical and found that this model did produce better images! See for yourself here.
The model tends to favor closeups or "headshots" more than full body unless you explicitly prompt for that. Not sure if that is a desired feature. Open for feedback on that.
Strengths
• More photorealistic for animals, cars, and people.
• Different Model, different generations. Try it with your favorite prompt!
• Better portraits. SDXL likes to put the subject way far out. This brings things closer.
• Eyes and hands are a bit better (from our testing)
• Better prompt control. Make sure to use commas and break apart subjects and styles!
Weaknesses
• Still a few concepts that are "plastic" (Food, but please test your prompts and let us know!)
• Can sometimes be too "person bias"
• Creativity is a little bit hindered. This could be considered a "Photorealistic" model. But again, TEST!
• Bias towards "sports cars" lol
Notice the RDXL follows the prompt a bit better. When asking for "upper body" it tends to follow that.
How to Use
• Prompt as usual. Use commas to separate subjects and styles. If you aren't getting what you want, use the token again. e.g. iron man made of lava, suit of armor infused with molten lava, ...etc if you don't get Iron Man in your generation, then put that token somewhere else, `iron man made of lava, suit of armor infused with molten lava, marvel iron man super h