Hi guys, does someone know a great video-to-video model which is NOT change the background?

[warm viper]

Hi guys, does someone know a great video-to-video model which is NOT change the background? Or it's not possible without training custom LoRa?

At least any information would be helpful. Thanks

[echo copper]

idk about video 2 video, but the best text 2 video that fits on mainstream RTX gpus is here https://github.com/lllyasviel/FramePack