kinda a style?
Samples came out as a very bland and generic 3d lora, but I'm hopeful that with the right negative prompt this can work well. This is testing what I saw in my other lora: https://civitai.com/models/840795/vft-test
That lora was made from frames captured from a single 3d animated video, and the lora ended up able to output other things in the style of the animation used. In this one, two videos were used. Whatever characters appeared within these frames were not tagged.