Find why generated video looks fake

Diagnose why this generated video reads as fake.

Check each source separately — the feeling is vague but the causes are specific:
1. **Physics** — weight, momentum, how things fall, cloth and hair movement, objects passing through each other. The most common cause and the least fixable by wording.
2. **Lighting** — shadows that do not match the source, missing contact shadows where objects meet surfaces, light that does not change as things move.
3. **Anatomy and motion** — joints bending wrong, gait, hands, faces during speech, blinking.
4. **Texture** — surfaces too uniform, skin too smooth, no wear or dirt, everything equally in focus.
5. **Camera** — motion too perfect, no micro-instability, focus that never breathes.
6. **Timing** — actions at uniform speed with no anticipation before or settle after.

For each finding: what specifically is wrong, and **fixable by prompt / fixable in editing / model limitation.**

Then:
- The single thing a viewer notices first.
- What to change in the prompt for the fixable ones.
- ⚠️ For the model limitations: say so plainly so I stop rerolling. *Knowing to stop is worth more than another attempt.*
- Whether reducing ambition — shorter, simpler, less motion — would produce a usable shot faster than fixing this one.