Spaces:

ali-vilab

/

modelscope-text-to-video-synthesis

like 1.45k

Starting on L4

so like ive been using this for like a good 2 months, and i have gained experience regarding using this tool, so heres the guide: experimental testing has shown that using 2 characters or people commonly seen together will work better, for example obama and biden, another thing to note is that i have been testing the timelapse function, and it works for making anime like fights. another function i like to use is flying/levitating/floating, now one thing to note is that floating often puts water in the scene. now, an experimental function i have figured out just now is sparring, that might work well in conjunction with fighting function (to often add more weight to the fighting aspect of the prompt) another function i like to use sometimes to add to the scene is the slow motion function, now this will make the fight land like 1 punch, however for quality and clarity purposes it will be more in depth and take a good 16 frames

GENERAL NOTES: for a general fight scene, use sparring and fighting in conjunction in order to maximize the chances of actually getting a good result.

do not add additional weights past 3 words, as this will possibly spoil your prompt weights, and will make it likely to outright fail.