Am in the middle of genning the vid requests, so stay tuned, they might only appear in the next thread, idk how long it will take
>>98801177Set cfg to 5, change length to 81, steps to 20. Don't use tiled VAE decoder, just use the regular one, because you are already using a quant model. I suggest writing in the positive that "This is a digital animation." and writing something among the lines that it's a video. Using a model like florence to scan and describe the image works wonders in getting you a good base prompt, of which you can build on top of. Maybe try not to use the word "wobble", it might make think you are describing jelly. You can also make it so that you can see a rough preview of the video it is currently genning by setting the preview method to TAESD in the manager, and then enabiling it in the settings, so that you can stop a gen like 5 steps in if you see the movement is wrong.
