I would love to add img 2 img, prompt rewrite and other features to my ComfyUI Hunyuan Image 3 nodes
Only wish that you would release the instruct weights so I could start work on those. I think Hunyuan Image 3 is great and under appreciated.
I've already started working on updates.... Should push them to the GitHub in the next days
hello
I've already started working on updates.... Should push them to the GitHub in the next days
Let us know when it's up 😀
Update: I just pushed updates tonight to my github and huggingface. The new nodes support text to image generation with instruct and instruct-distil, image to image (image edits) and multi-image fusion. I made NF4 and INT8 models for both Instruct and Instruct-distil to go with my ComfyUI nodes. Models are here https://huggingface.co/EricRollei
and comfyui nodes for Hunyuan Image 3 base, and Instruct are here: https://github.com/EricRollei/Comfy_HunyuanImage3
The new models are really cool. The recaption and think-recaption options are slow, but it turns out you don't really need it for image editing or really text to image. Try the distil models first as they are much faster. Slightly better quality from the instruct model but slower.
Thanks @EricRollei