lightx2v

/

Wan2.1-T2V-14B-StepDistill-CfgDistill

like 127

"I2V-480P Training
DataSet Preparation
Generate a series of videos using the original Wan2.1 model."

I used I2V-14B-480P-step_cfg_distilled model to generate about 3w videos. 8steps+cfg1.
The trained distilled i2v-14b-bi model (training is followed by "I2V-480P Training" readme) inferenced noise videos.

1、Should I use T2V-14B model to generate the videos as the training set?
2、Can I use your distilled T2V-14B model for faster inference as the training set?
3、Can I use the real world videos as the training set?