I seen a faceswap on Reddit that literally changes the shape of the face and everything. So, it's more of a head swap, but everything matches to where it looks like it was the original. Would be cool of there was a video swapper like that. Comfy is way better now that I had the time to learn it better. All the models are already there.
Super helpful, thank you! One thing I find is that if I change the text prompt the generation takes a Cheer takes a long time . After that first gen it is quick again. Is there a way to make changing the text less impactful to the generation time at all? Cheers!
So just tested a few options, its possible - great tools :D Just transferring an image with inpainting into an area without recreating it... still searching for that :D
@@Razidurgh yeah, I had 32 GB, too and upgraded to 128 GB last week. It's a day and night difference even though the GPU (3090ti / 24 GB) didn't change. Shifting models around got so much smoother.
Using the FLUX Fill model, if you use large masks (for instance, in the case of outpainting), the model performs poorly, adding a significant amount of noise to the resulting images, Did you experience those effects?
@@TheFutureThinker Yes, I have. The example you show in the video is not exactly what I mean. I get the entire expanded area covered by noise, and in some cases, even the original area is affected
Example Workflow And AI Models : comfyanonymous.github.io/ComfyUI_examples/flux/
Thank you! Another excellent video and workflows. Flux tools is so hot right now 🔥
Glad you like them!
Can we use LORA with FILL?
Does any of these new tools work for Flux schnell?
Thanks for the info. Would like to try Fill , it looks promising.
yes , it's nice model
12:22 Yes, you can see demarcation. Look at the wall different texture on the right. Can be fixed, but more work...
Yes, so base on this as a start and make another refiner sampling.
But at least the skin tone is not 2 tones. Happened in other inpaint model.
Do you think I can make a scene 360 seamless panoramic with this outpaint?
On cevitai someone modified flux fill fp8
Please make a video review about it for those who have GPU vram 16 GB or less its size is only 12 GB
I seen a faceswap on Reddit that literally changes the shape of the face and everything. So, it's more of a head swap, but everything matches to where it looks like it was the original. Would be cool of there was a video swapper like that. Comfy is way better now that I had the time to learn it better. All the models are already there.
it's possbile by doing img2img batch using this. but without motion model like AD , it will flicker a lot
I have a small, stupid question. What this small fox means in comfyui ? I did not saw it in mine Comfy :).
It means, that this node with the fox is from Comfyui. Other nodes without the fox are custom nodes. :)
Super helpful, thank you! One thing I find is that if I change the text prompt the generation takes a Cheer takes a long time . After that first gen it is quick again. Is there a way to make changing the text less impactful to the generation time at all? Cheers!
Looks like your case are loading the Dual Clip again, that cause another model loading time.
@@TheFutureThinker I'll take a look thanks!
@@TheFutureThinker I had the clip models in the dual clip loaded in the opposite order. Seems to be working fine now.
😉👍
Where are Workflows? I can't find them in provided link :)
On each section, those anime girl images are workflow, you can drag it onto your desktop, and load it in Confyui
@@TheFutureThinker wow :) nice. One more thing, I just lose my "start" button :D Where can I find it? :D
Found Ctrl+Enter ;-)
@@lukaszpodgorski oh ok haha. yes hotkey hidden
Which graphic card are you using ?
Does it work with 8 GB v ram
yes it work in 8 GB Vram.
Can I do that using AMD gpu? 7800xt
Can you use loras für in-outpainting? I tried quickly but it was not working. I mean like a character or something.
i haven't try Lora, but Redux reference another image works.
@@TheFutureThinker Not with a character, its similar but not like a lora :)
So just tested a few options, its possible - great tools :D Just transferring an image with inpainting into an area without recreating it... still searching for that :D
My 32 GB RAM is struggling. I need an upgrade.
@@Razidurgh yeah, I had 32 GB, too and upgraded to 128 GB last week. It's a day and night difference even though the GPU (3090ti / 24 GB) didn't change. Shifting models around got so much smoother.
Using the FLUX Fill model, if you use large masks (for instance, in the case of outpainting), the model performs poorly, adding a significant amount of noise to the resulting images, Did you experience those effects?
Have you seen my examples have those effects?
@@TheFutureThinker Yes, I have. The example you show in the video is not exactly what I mean. I get the entire expanded area covered by noise, and in some cases, even the original area is affected
an official black forest openpose and then it will be perfect !
Oh yes looking forward to that
Esa es mi prima