This workflow on my rtx 4060 ti 8 Vram, MAKE 1 MINUTE OF GENERATION WITH UPSCALE TOGHETER.
To build this wf, takes a lot of weeks, with a help of a lot of people, we made this wf possible.
there a lot of custom nodes, a lot of them helps nunchako generate images better than standard one.
is higly recommended for you to have the last version of comfyui.
This workflow has first pass, second pass(upscale), qwen edit (optional) and can save your images with prompt for civit ai posting.
some examples of generation.
Description
FAQ
Comments (6)
Hi, first of all; love your work!
Question about the workflow, I don't know what I'm doing wrong but I can't seem to get the same character when I put an image in the image to image section. Seem like if I put an anime woman, It output a random semirealistic anime woman.
you need to set denoise of the first sample bellow to .3
Still not working... I see that you should reconnect the latent image to image. What should i conect to what exactly? Becaus it seems it's not recdognizing my input image at all.
@yannberg240398 even in denoise zero? you should connect the vae encode latent to first sample
@luisa_pinguin denoise 0 output the same image, but isn't this workflow made to edit images or is it just an upscaler? Maybe I didn't understand the utility of the workflow, I'm trying to change poses, background, etc.
Really well done, thanks for making and posting! In the standard workflow, I always get an error on 3rd image just FYI (I assume because Qwen Edit piece is removed, but then the 3rd image section should be deleted entirely - correct?).