Aye, that's what it's doing here. Similar to Gen2 without a prompt.
@fofrai
-
AI-Generated Video Game Images Show Promising Results
By
–
Trying out some video game images and the results are good. pic.twitter.com/fhpoDcz9md
— fofr (@fofrAI) 21 novembre 2023Trying out some video game images and the results are good.
-
Open Source Image-to-Video Tool Achieves Strong Results
By
–
Open source image-to-video at last!
— fofr (@fofrAI) 21 novembre 2023
Results are really good at 840×360 resolution.https://t.co/2EPdTZWgaN https://t.co/ANzQsoIYk4 pic.twitter.com/qVj6HaqxK6Open source image-to-video at last!
Results are really good at 840×360 resolution. https://
replicate.com/p/t74n2u3b3ua7
tjicrhawllyaa4
… -
Promising AI Project: Code and Models Expected Soon
By
–
This looks really promising from @giuvecchio95
— fofr (@fofrAI) 21 novembre 2023
Hopefully code and models soon. https://t.co/JNdi4PbahAThis looks really promising from @giuvecchio95 Hopefully code and models soon.
-
SDXL LCM and HED ControlNet Generate Creative 2004-Style Images
By
–
Using my bad photos from 2004 lora with SDXL LCM and an HED controlnet. 4.4s "A photo from 2004 in the style of TOK, an astronaut riding a unicorn in a kitchen" https://
replicate.com/p/mnbrgidbgpiy
r6mnq537yprpq4
… -
SDXL LCM Multi-ControlNet LoRA Model Released
By
–
The model is here, but still a little WIP: https://
replicate.com/fofr/sdxl-lcm-
multi-controlnet-lora
… And the code is here: -
ControlNet helps control generative AI model outputs
By
–
I'd add that controlnet helps reign it in a bit
-
Optimal Guidance Scale for AI Model Prompting Experiments
By
–
You definitely need to prompt harder. I ran some experiments to work out best guidance scale too and settled on 1.1.
-
SDXL Model with LCM and ControlNet Achieves 4x Faster Generation
By
–
This is a result from a model that combines LCM, SDXL, custom loras and multi-controlnet, ~4x faster. 1024×1024, 8 steps, HED controlnet: 3.86s