Replies: 4 comments
-
I've just tried the same thing... two different characters loras in Flux. In addition to taking over 10-minutes on my RTX 4070Ti, it also does not actually work. Both characters (one on each side of the screen) used the second hooked lora. I could not get the first lora to be recognized at all. |
Beta Was this translation helpful? Give feedback.
-
Hello ! What I understand so far is that the two prompts have to be the same and describe the whole scene, with trigger words. |
Beta Was this translation helpful? Give feedback.
-
Hi @uberneko, Thank in advance for any hints about it ! |
Beta Was this translation helpful? Give feedback.
-
Hello ! In March, I realized that it has been divided by almost 2, like 4:30 mins... And I realize now, with ComfyUI 3.29, it went down again to 2:30 mins ! So again almost divided by two. I don't know what you did guys, within the code, for Masking, but that was really great !! Thank you ! |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
That was a good news, https://blog.comfy.org/masking-and-scheduling-lora-and-model-weights/, and decided to give it a try -
With some adjustments... no checkpoint but changing with Model (Flux1.dev), CLIP and VAE loaders - also with two "Create Hook LoRA" - as I wanted to make one specific person on the right and one specific on the left with the basic black/white mask.
Although usually my 4070Ti Super takes about 30s to create a picture, it took more than 14mins with a poor result (it seems to work since I can see the two people I needed) but it is quite blurred. Reducing to less steps would be far worse.
Did you experience the same slow-down ?
Beta Was this translation helpful? Give feedback.
All reactions