Character LoRA Training

After creating and rigging a character, you can train a Character LoRA to help preserve that character’s appearance more consistently when generating videos, especially across changing views, poses, framing, and shot composition.

Once training is complete, the Character LoRA is saved for that character and applied automatically whenever you generate videos with the same character.

You only need to train a Character LoRA once per character. It is especially useful when consistency matters, such as when the character appears from different views or distances, moves closer to or farther from the camera, turns around, or appears across multiple cuts.

Character LoRA Training can help maintain the character’s face, body, clothing, colors, accessories, and overall design. It may also help preserve important visual details such as patterns, symbols, or logos, although results can still vary depending on the shot, prompt, motion, and generation context.

Character LoRA Training is optional, but once trained, it is used automatically for that character.