Good old missionary.
No trigger words, just describe. Look at training data / untouched sample videos with workflow attached for sample prompts.
2/5 videos generally work. I will probably retrain this at a higher resolution with a better dataset.
Using diffusion-pipe on Linux, 4090, 12 videos, 256 resolution, 32 frames, 20 epochs, 2 hrs.
Description
FAQ
Comments (2)
Very good. THANK YOU SIR,I hope lora in the upside-down position
Would really love to see a 2.2 hi-lo version of this