An experimental lora that works quite well. I wanted to see how well Wan could handle the technique I described here: https://huggingface.co/blog/neph1/framepack-camera-control-loras (TL;DR: Single training video camera concept).

It adds a subtle shake to the camera, dogma style, which is associated with a life-like, documentary feel.

Only High noise needed.

Requires a high strength, showcase images use around 1.7. May work without trigger. It's actually trained on "shaky, handheld camera", but I would recommend not having "camera" in the prompt.