text2vid img2vid H3 Minimax - Fully uncensored local AI video generator released today

Using Ref2VA to replace a character in a video.

Source video: (use a 360p resolution source video for faster generation times)


Final result:


Image of Chun-Li for
Source video above for
 
I know this has been asked a lot but is anyone able to edit a video where it will just remove the clothes but keeps everything the same like the movements and backgrounds? Is it better to use another model for it or does minimax does a great job for this purpose?
 
You should really look into in painting the subject in the reference video green so that no features from them bleed into the transfer. There are workflows for it
 
i havent tried others --since minimax came out and none before had the reference input
but that is a pretty easy thing to do --- might need some xxx loras from civitai to get the body right --- it does a decent job at assuming what is underneath
 
Yes and I have an example that I can’t share because real people

You basically just tell the model to remove the woman’s clothes
 
Dude I updated comfyUI and now it doesn't work at all, it took 17+ minutes for 15 secs at 0.5 megapixel on my 5090 and the result was basically the same video i gave it as a reference. Yesterday it was like 3-5 minutes and it worked fine. Anyone with the same problem?
 
How are you able to add a reference video node instead of the original 2 image reference that came with the original ref2vid template? Do you mind sharing the workflow with the video ref node?
 
Seeing how I credibly intensive this all is is making me realize how hard it is going to be to replace Somato 😭 I just wanted to learn because I was tired of gatekeeping but damn I don't even have a rig for something that would last 10 sec it seems 💀💀💀
 
I get a Error invoking remote method 'run-action': Error: EPERM: operation not permitted, mkdir 'D:\' message when I open my instance. It worked fine yesterday when I was trying to learn to use it
 
Lewdrawts on ref2video default, next to where it's loading in 2 images, add a "load video" node, then a "get video components" node. The output from the first goes into the second. Then you take the "images" output from the second (and audio if you want it) and drag it to the "minimax H3 reference to video" where the images are already going.
 
PC config: RTX 5060ti 16GB, R7 5800 and 16GB ram, I have stability matrix and comfy ui running on it.

Can someone suggest me some great I2V models for local generation that I could use. I'm currently using wan 2.2 A14B model and want better quality and more prompt adherence.
 
I dont know what is going on with mine either after the latest update. Using the basic template for i2v, it was taking 15minutes to just get to 15% before I cancelled it.
 
You must log in or register to reply here.
Back
Top