Start with individual photos of your friends
A Tiny Desk AI video with friends works best when each selected role has its own reference. Ask each friend for one clear photo and decide their role before uploading. You can replace a singer, the saxophonist or the whole group. Everyone else stays as they appear in the source.
A group photo creates an ambiguous reference when several faces share the frame. Choose individual pictures, or prepare a separate crop for each person that retains enough hair, shoulders and clothing to describe their appearance.
Use a fixed five-person role map
Photo numbers belong to roles, even if earlier slots are empty. Uploading only to slot 5 still selects the saxophonist. The keyboard player does not move into slot 1 when the lead is unassigned.
| Slot | Band member | Source cue |
|---|---|---|
| 1 | Lead singer | Black hair, navy collared shirt and headset; opening and later close-ups. |
| 2 | Keyboard player | Green sweater over a white collar, seated at the red keyboard. |
| 3 | Left backing singer | Black hair, pink cardigan and cream top in the band shot. |
| 4 | Right backing singer | Blonde hair and light blue top in the band shot. |
| 5 | Saxophonist | Wavy brown hair, lavender top and gold saxophone in the later solo. |
Pick a casting plan that fits your group
For a birthday video, the birthday person can take the lead singer slot while the rest of the band remains original. Two friends can use vocals and saxophone for separate close-ups. A three-person group could choose the lead and both backing singers to appear across the main vocal sections.
- Solo lead: fill slot 1 and leave 2–5 empty.
- Vocals and sax duo: fill slots 1 and 5.
- Backing-vocal pair: fill slots 3 and 4.
- Full band: give all five slots a separate photo.
Choose the complete appearance you want
The edit requests the person shown in each reference, including face, hairstyle, visible body appearance and clothes. Source outfit descriptions locate the old performer. They do not require your friend to keep the old outfit. A photo cropped tightly around the face provides less information about clothing than an upper-body portrait.
Keep the face large enough to recognize. Do not sacrifice facial detail just to include shoes or a distant full-body silhouette. The template preserves the performance props, so an object held in your portrait is not a reliable way to replace the saxophone or keyboard.
Check assignments before and after generation
Write a short list such as “Alex: lead; Sam: sax” and compare it with the review dialog. Remove an image to return that role to Keep original. Empty roles are not instructions to reuse another friend’s photo.
After generation, pause at each person’s clearest appearance and check whether their identity and outfit stay consistent. A good opening frame alone does not prove that the full band mapping held through every cut.