With LoRA on, your prompt picks up to 3 LoRAs from this pod's library. Tick some here to choose them yourself instead (up to 3). Automatic goes back to matching the prompt.
Create a character
Add a character
15 to 30 clear pictures of the same person, varied angles and light. The model learns this person and can then draw them in any scene, face and body. Tip: on a movie you like, open it and use "Frames for a character" to fill this automatically with sharpened frames.
Picture LoRA: about 20 to 40 minutes. Video LoRA: 1 to 3 hours, and videos pause while it trains. Both: the picture one first. Stop any of them from Characters.
Load a saved character
Pick a .safetensors character file you saved before. The name and the trigger fill in from the file name. Keep the trigger exactly as it was when the character was trained, or the person will not show up.
Your characters
Character
Generate expressions
The reference picture fills the six windows, and each makes a short clip. To run a window from a different picture, tap the picture then the window, or drag the picture onto the window. On a window, Keep takes three sharpened frames into the tray and Redo makes it again.
Source pictures, tap one then a window, or drag onto a window