Lumina-Image 2.0 : A Unified and Efficient Image Generative Model
Is a 2 billion parameter flow-based diffusion transformer that uses Gemma 2 2B for its text encoder.
Before using the model for the first time, be sure to update your ComfyUI instance. c:
ComfyUI workflow:
https://comfyanonymous.github.io/ComfyUI_examples/lumina2/
FAQ
Comments (6)
Hi karely,
thank you for adding the model! Please be aware that you need a png file to embed workflows. So either replace the jpg image or add the link to ComfyUI examples: https://comfyanonymous.github.io/ComfyUI_examples/lumina2/
Fun fact: I get a bit different result with the same settings, and obtain too much fingers on her other hand...
And i only see a little bit of color bleeding from the glass on her dress
CheckpointLoaderSimple Error when loading this safetensors
1. Cool.
2. Can this function in Webui-Forge?
How fast it is compared to SDXL and Flux?
I don't understand why I have to wait so long for generation. Yes, 2080 Super 8Gb... But: 2 images 1152*896, gradient_estimation beta, CFG3.5 10steps takes 246 seconds. Moreover, both BF16 and GGUF-q5 give the same speed. But the same images with the same parameters, only 20steps, are generated on Flux.dev in 208 seconds, on Flex.1-alpha in 169 seconds, SD3.5 Large in 208 seconds.
ERROR: Could not detect model type