search

kabachuha / sd-webui-text2video

Auto1111 extension implementing text2video diffusion models (like ModelScope or VideoCrafter) using only Auto1111 webui dependencies
Other
1.28k stars 107 forks source link

inpainting, specify an initial frame and the video will be extrapolat… #74

Closed nagolinc closed 1 year ago

nagolinc commented 1 year ago

…ed from there

inpainting_Screenshot 2023-03-27 091243

https://user-images.githubusercontent.com/7775917/228009298-6bcbae51-f58f-4da9-b781-80270b54e026.mp4

kabachuha commented 1 year ago

Nice! Thank you for your work.

As for the future additions, what do you think of making a keyframing map to enable images insertion in various points of time. And it will allow for 'looping' the animation

JSON is probably the best format for making such a map. Here is how it was done for vanilla Deforum https://github.com/deforum-art/deforum-for-automatic1111-webui/pull/247

nagolinc commented 1 year ago

Nice! Thank you for your work.

As for the future additions, what do you think of making a keyframing map to enable images insertion in various points of time. And it will allow for 'looping' the animation

JSON is probably the best format for making such a map. Here is how it was done for vanilla Deforum deforum-art/deforum-for-automatic1111-webui#247

Agreed. It should be relatively straightforward to modify what I’ve done. The key variables are the sequence of images and the sequence of mask weights.

I think a straightforward way to do it is if the file is a video we use the frames for the video. Otherwise we just repeat the single image for all of the frames and then as you said we can have the sequence of mask weights be optionally input as json or the user can select the options beginning, end and loop.

Natotela commented 1 year ago

so excuse me for the dumb question, but this is from April, says merged, and I got today's most updated extension, so how come I don't have that tab?

kabachuha commented 1 year ago

It's inside the txt2img tab, at the bottom