Character Consistency Issue with LoRA in ComfyUI

[kozmoss]

Hello everyone,

I’ve been working on generating consistent character images in ComfyUI. However, no matter how much detail I added in the prompt, the images still showed slight variations. Even if some results looked close to my character, others ended up looking like entirely different people in different frames.

After doing some research, I discovered that I could achieve better consistency by creating a LoRA . “https://www.youtube.com/watch?v=grtmiWbmvv0&t=197s” I found a free workflow online and set up all the required custom nodes, checkpoints, and models. The workflow runs without errors, but for some reason, none of the generated images are rendering in the Preview Image node.

What could be causing this issue?

[RSa]

Looks like your sampler has a red border around it, maybe missing some component?

Can you post the .json? It will help a lot to understand. Can’t really read the image.

ty!

[kozmoss]

Hello, the issue is most likely related to MVAdapter. Even after correcting the incorrect contexts, as far as I remember, MVAdapter requires a high VRAM to function properly. Since my GPU has only 6 GB VRAM, I’m encountering errors and it cannot generate a 6-image output.