Character consistency in AI video - Has anyone actually solved this?

I'm working on a feature-length AI-assisted film and character consistency is killing me. I've tried Runway, Pika, and even custom LoRAs in Stable Diffusion, but I can't get the same character to look identical across 100+ shots.

The face changes slightly, the lighting affects it differently, and don't even get me started on wardrobe consistency.

Has anyone actually cracked this? I'm willing to try anything at this point - even if it means manual frame-by-frame cleanup.

What's your workflow? Are you using reference images? Training custom models? Giving up and hiring actors?