Direct answer
Photo 1 is the listener: left side, headphones, the one who hears the leak. Photo 2 is the speaker: right side, folders, the one who says the line. The prompt names people by position, so the upload order is the casting.
Make my videoPhoto order and casting
The one direction that matters, and the failure modes that come from getting it wrong.
Published November 10, 2026
The mapping
| Upload | Seat | Position | What they do |
|---|---|---|---|
| Photo 1 | Listener | Left, by the mirror | Wears the headphones; listens; touches one cup; reacts. |
| Photo 2 | Speaker | Right, by the doors | Holds the folders; hears the leak; says the line; smiles. |
What happens if you swap them
The model does not know which person is “the funny one”. It follows the seat mapping. If you upload the speaker first, the person you wanted behind the headphones will be the one holding the folders, and the line will land on the other face. The video still renders; the joke just lands on the wrong person. Swapping back and re-rendering is the only fix.
Photo requirements, per seat
- One person per photo. A second face in the frame can be copied into the scene.
- Head-and-shoulders or chest-up. Full-body photos are cropped automatically, but the face loses detail.
- Neutral expression is fine; the model supplies the reaction. Do not upload a photo that already has headphones or folders unless you want that object duplicated.
- No heavy filters, sunglasses, masks, or hats that cover the hairline.
- Both people must consent. No celebrities, no minors, no photos you scraped from someone else’s profile.
Can I use the same photo twice?
No. The pipeline uploads two identities and asks the model to keep them separate. The same image in both seats creates a duplicate-face conflict; the render usually fails or produces two copies of one person. If you want a solo video, this template is not the right tool.