epic fail? xD, bad mine. solved, Reflections, Ideas

[zephirusgit]

Just a scare. Ahh I was very scared because I started really very slowly, the comfyui, to the point that the visualization of images was disgust, it was unincoronized, etc. I didn't know if it was a problem at the GPU or Comfyui. But apparently it was a velocity parameter in the browser, which limits speed, so that it does not consume more of a certain bandwidth. So it was nothing, bad mine. I take this opportunity to tell you about a very good checkpoint model, "Epicphotogasm_lastunicorn.safetensors" of Civitai Pretty complete, I am using that model with another Lora and Wow, some impeccable images .. With some other more distinguished me between a man and a woman, very poorly balanced. With this everything goes perfect. By the way, I need more xD nodes. Does anyone know if there is any project like Diffusers, or Videocrafter for Comfyui? or who encourages the image as if they spoke? or the example of the dog or leon, or voice creators. Like Bark? I know there are already avatars and things like that, but I feel that there will be great things very soon. I congratulate each person who helps create these wonderful things. I make a consultation, long ago when I recently started in Comfyui He had several tools that detected a person and their pose, With Controlnet, and I generated the skeleton, bones, and with it, I could create very good things, then when updating certain things, that stopped walking, (luckily I leave a backup) But then the controlnet that followed, no matter how good it is very good for certain things, it is fatal to others, where the skeleton was doing very well. Question open to geniuses, because I am a simple user. Does anyone manage to make it walk again? a utility that is missing if it exists, It would be a node that turns a video to images, example Load Video >> Frame by Frame Loop, and from there, for example, detect the skeleton with controlnet, and recreate it with stablediffusion, picture to picture, having the option that this frame exit, is in a folder, and then rearma the video .

Load Video folder to put the frames

Image

CONTROLNET FOLDER (LOOP) Bone skeleton.

Image >> Latent >>> mask >>

Post processing, reface, Fixface Swapface. etc

Save to Folder MP4 MP4 final ..

mm what do you think, is there something like that? For example with FFMPEG, manage to pass a video to photograms And that folder if opening it from Comfyui make reactor and save them. and externally looking for the output images, That was a problem because I have thousands, and it was not a clean sequence, (I made a program in Basic to rename) to a new sequence. and then reensamble.

Likewise, this took me hours, And try to use the roop externally and nothing, it was a few minutes. But regardless of the times, it would be great to be able to do it in Comfyui!

One could be filmed by doing something, and reconverting an avatar box by picture ..

I also admit that I am having many problems to maintain the same character and clothes in different paintings, there the character managed to keep it but clothes always varied

I still can't perfect that, unless certain garments explicitly clarify, but then I need a logic, randomly generating layers with their clothes, choosing between a list, and trying to maintain that prompt until the end, Is it possible to make a node of (hair styles) or clothing styles? I think at some point I saw how a reface, but for clothes? Does anyone remember it? What was a trained model but only with clothes? If you know please comment. If you can create a basic node, for example you have a CSV For example dog breeds, example Bulldog, Doverman, Chihuahua, Caniche, And you can create a node that is, Load CSV, Random (3), >>> ejm Doverman. And so with colors etc .. in a program is simple But I do not fall an idea of how it is here, xD

[zephirusgit]

[dmarx]

the nice vibes are always appreciated, but this isn't the place for them. you should close this issue