search

ali-vilab / VGen

Official repo for VGen: a holistic video generation ecosystem for video generation building on diffusion models
https://i2vgen-xl.github.io
2.97k stars 265 forks source link

multiple videos for each text input during inference #96

Open saba-vaez opened 8 months ago

saba-vaez commented 8 months ago

Hi, when I do inference using fine tuned modelscope t2v model, it produces 3 or 4 videos for each text prompt. I was wondering is this behaviour normal and what is the difference between the videos?

Screenshot 2024-03-12 at 10 35 47
lky-ang commented 3 months ago

round = 1