Spaces:
Running
Hear me out
For me, Klein seems to take a bit more than 15-30 minutes to generate an image, but maybe that's just because there isn't a lightning LoRA yet. Also, I'd love to see a Qwen-Image space running perfectly fine on a CPU. I tried it out and got a runtime error.
I purposely use klein-4b because of its small size on CPU, this CPU space use flux-2-klein-4b-Q4_K_M.gguf (2.6 GB: 4B parameters); I could try one day once i don't get "Internal Error" when create space.., but it will defintively be painfully slow as qwen-image-2512-Q4_K_M.gguf is 13.2 GB (20B parameters) and that even with lightning LoRA https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/blob/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors
A ZeroGPU space looks more realistic but has already been made by many.
For some reason generating an image and editing an image give me the same blank squares or patterns and I have no idea why. Maybe it's because I haven't set my image resolution high enough?
UPDATE: I found that downscaling the image I want to edit works, although since Klein is running on a CPU instead of a T4 or ZeroGPU (which is understandable), it doesn't exactly make image generation any faster
For me, Klein seems to take a bit more than 15-30 minutes to generate an image, but maybe that's just because there isn't a lightning LoRA yet. Also, I'd love to see a Qwen-Image space running perfectly fine on a CPU. I tried it out and got a runtime error.
Klein 4B don't need a lightning LoRA, it already run at 4 steps fine:
Our fastest distilled model for sub-second image generation.
I created for you https://huggingface.co/spaces/WeReCooking/Qwen-Image-CPU and if waited over 9000s+! That's cuz I did not finish... ^^ I will ping you once i do.
I know Qwen ZeroGPU is free quota is limited per day but i know this spaces/selfit-camera/Omni-Image-Editor with use omnieditor.net, or opensourcegen.com that has more relax gen.