Kolors is a large-scale text-to-image generation model based on latent diffusion, developed by the Kuaishou Kolors team. Trained on billions of text-image pairs, Kolors exhibits significant advantages over both open-source and closed-source models in visual quality, complex semantic accuracy, and text rendering for both Chinese and English characters. Furthermore, Kolors supports both Chinese and English inputs, demonstrating strong performance in understanding and generating Chinese-specific content. For more details, please refer to our technical report.
Description
FAQ
Comments (27)
No files
any prompt gives me nothing but static as output
Unbelievable, this model do exactly what i prompted, and it can handle Hands much better than SDXL or AuraFlow. Gamechanger, i think.
Do You promp in English or chinese? Also, what Sampler do You use?
Error: Model type not recognized.
no prompt................................
how to use this ??? anyone?
minuszoneai nodes for comfyui
Very good Model !
Nice Model, made a WF for comfyui with friendly civitai metada ! https://civitai.com/models/608622/kolors-basic-workflow-civit-friendly-metadata
好奇怪呀，这个模型画男的好像非常困难，都明确说了是青年男子、男子、男人都是画出女的，或者创画出长头发偏女性的男
试试胡歌
What to do with the attached checkpoint (kolors_v10.safetensors)?
I dont find it very good at realism but oh boy this model in extremely good in so many other aspects, paired up with IPadapter is magic.
I find prompting like "a high-resolution photograph of thing" helps a lot to steer it away from generating more artistic 2dish images.
@diffusionfanatic1173 hey, thanks for tip, i will for sure try it out
does this work in forge?
It would be great if you could get this worked out with civitai so it would auto-tag your model.
IPAdapter is Really good with this Model ;) Give it a try !
We are banned to use it profesionally. I can't even put an image made with it in a game under "CC by" open licence...
I say : That icon should be in RED ! And blinking.
what do you mean exactly? I don't see anything in their license in that regard beyond them stating that if you run a service that has greater than 300 million monthly users then you need to contact them for a special commercial agreement.
I downloaded Kolors from HuggingFace, and that license allows commercial use in most situations. Only here, for some reason, a prohibition of commercial use.
FYI, the "Sliding Window Guidance" custom Comfy node works great with this, at about half the default scale setting (so around 2.5 instead of 5). I usually run that with either DPM++ SDE GPU at CFG 5.0, or DPM++ 3M SDE GPU Exponential at CFG 4.0.
this is a YT vid of how to install in comfyui
watch?v=ra1yv5VDHFo
Wow, looks amazing, just like real photos
Returns an error AttributeError: 'ChatGLMConfig' object has no attribute 'use_cache'
What could be the problem? (ComfyUI)
Details
Files
kolors_v10.safetensors
Mirrors
kolors.fp16.safetensors
diffusion_pytorch_model.fp16.safetensors
diffusion_pytorch_model.fp16.safetensors
diffusion_pytorch_model.fp16.safetensors
diffusion_pytorch_model.fp16.safetensors
diffusion_pytorch_model.fp16.safetensors
kolors_v10.safetensors
kolors.fp16.safetensors
kolorsfp16.safetensors
kolor_unet_fp16.safetensors
diffusion_pytorch_model.fp16.safetensors
diffusion_pytorch_model.fp16.safetensors
kolors.fp16.safetensors
diffusion_pytorch_model.fp16.safetensors