How much footage does an H3 style LoRA need anyways?

I think I finally came up with a good idea for a video style I could (relatively) easily get footage of for the current contest, but I realized I have no idea how much content I need for it. I also realized I'm likely not the only one who wants to know that, so I figured I'd ask publicly. Also share any advice you have on training for H3 in the comments.

Picture unrelated: I had no idea what to put and when browsing my folder of old gens I remembered this one was fun to make: Bending LoRAs into doing things they were never intended to do is always good fun.

Edit: Submitted for training with 19 videos and 16 still images (character sheets and high res backgrounds). Lets hope this works out.

Edit: Training v2 with 34 videos and 16 still images. Even from epoch 1 and 2 the golf course sample image looks way better (Miku sample is still wrong camera).