Short answer
Face-swap apps keep the footage but struggle with bodies, hair and light. One-call video models regenerate the whole clip and drift from the original cuts. Rendering each shot separately from the reference — what this site does — keeps the edit intact and handles up to five identities, at the cost of a slower render.
Make my band videoTiny Desk AI vs other ways to make the meme
Three approaches, three very different failure modes.
Updated 2026-10-29
There are three realistic ways to make this meme. They fail in different places, so the right choice depends on what you care about.
Side by side
| Approach | Keeps the edit | Multiple people | Typical failure |
|---|---|---|---|
| Face-swap app | Yes, frame by frame | Weak beyond one face | Hair, light and body do not match; seams on turns |
| One-call video model | Sometimes | Usually misses 1–2 people | New camera moves, shifted cuts, background redraw |
| Per-shot reference render (this site) | Yes, frame-locked | Up to five seats | Slowest render; back-row faces are harder |
Why one call over 19 seconds breaks
Video models cap the combined length of the reference and the output. Nineteen seconds of reference plus nineteen seconds of output exceeds that cap, so the clip has to be split. Splitting only works if you cut on hard cuts — which this reference happens to have, at frames 33, 161, 312 and 393.
Why face swap alone is not enough
A face swap pastes a face onto existing pixels. It cannot invent the back of a head that turns, fix a hairline that does not match, or relight a face to the room. When the shot is a close-up with a laugh, those seams are the whole result.
Where this approach is weaker
- It is slower: two to five separate renders instead of one.
- People at the edge of wide shots are still the hardest seats.
- Clothes and instruments are inherited from the reference, not invented for your cast.
- There is one clip. No song choice, no custom scene.
If that trade-off sounds right, make one. If you want the cost model first, read is Tiny Desk AI free.