Work in progress: An extension for StableDiffusion WebUI - Automatic1111 and Vladmandic´s cut.
It takes any image sequence and process some interpolation between them in order to generate animations into movie file.
First processors will be Google F.I.L.M to interpolate semantically between images in a soft way. Second idea is to have the interpolation routine from InfiniteZoom on board, to rework/rerun a video if there small glithces to be fixed.
Very poor, yet: Development is currrently using Vladmandics WebUi-Fork. The contained "install.py" is somehow not called or just not processing the requirements.txt.
- install extension via url
- in settings of that extension is a URL, download the model-zip (from civitai. https://civitai.com/models/54606)
- unzip it to this_extensionfolder/scripts/processors/FILM (creates pretrained_models folder)
- open terminal, cd to vlad/
venv\Scripts\activate
cd extensions/stable-diffusion-webui-sequencor
pip install -r requirements.txt
(needs torchvision, some cli tools)- restart vlad/auto1111
- In Sequencor-Settings (i Settings tab) define a path to your FFMPEG-installtion (need there a ffprobe-executable)
good luck
nothing to set so far.
- after interpolation FILM, VRAM is occupied by this model. need a unload-button or - mechanism
- Infinite Zoom not implemented