How to Make an Elf Video From a Photo: Five Recorded Clips, Prompts and What Went Wrong

In short: one photo, one fixed scene, five seconds. We made five elf videos from single AI-made sample photos with MiniMax H3 Max on Cicadas, two on the AI elf video generator and three on the elf dance page. All five kept the face close to the photo and gave a clean costume with no lettering. What went wrong was smaller: neither workshop clip had the jingling bells we asked for (the dance clips are shown without sound), and one wink came out half-formed. A 5-second clip is 63 credits on Cicadas; each of ours cost about $0.25 at the provider's list price.

We make Cicadas. The clips below were generated on 14 and 27 October 2026, and each one is listed with its request ID. Everything we say about a clip comes from its recorded description and from frame sheets we pulled from the file, not from memory. If you want to know how we record and check every clip before it goes on a page, read what 255 samples cost us and the flaws we logged.

1. Pick the photo first: the rules

The photo decides more than the scene does. Both elf pages take exactly one photo, and the rules are the same:

• An adult only. You, or an adult who agreed to appear. Photos of children are not accepted, and neither are celebrities or public figures. You tick a confirmation before each generation.
• Clear and front-facing. Good light, the whole face visible, nothing covering it. We only tested clean, front-facing portraits, so start with one.
• Your rights. The photo is used for that clip only, and the finished MP4 stays private in My videos until you share it.

We used three of the AI-made, fictional sample portraits used on many Cicadas pages: a woman with dark curls, a bearded man and a young man.

2. Choose the page: workshop or dance

The two pages run the same model in the same way, reference to video from one photo, 9:16, 5, 10 or 15 seconds. Neither page takes a prompt of your own: each has fixed scenes.

• AI elf video generator: two scenes in a toy workshop. Workshop dance: you dance between workbenches and grin at the camera. Wrapping presents: you wrap a box at top speed, stack it and wink.
• Elf dance page: three dances. Tree-side jig beside a lit tree, snow shuffle outside the workshop at dusk, high-kick line in front of a stack of presents.

Pick the workshop page for a card-style clip with a task in it, and the dance page if the point is the dancing. If you would rather have Santa walk into a photo of your own living room, with no person in the photo at all, that is the AI Santa video generator.

3. The five clips, prompt by prompt

Every prompt follows the same pattern: vertical phone footage, one continuous shot, the person from Image 1 in an elf costume (a green tunic and a pointed green hat with a bell; four of the five add red-and-white striped stockings), one action, "Keep the face and hair as in the photo", a sound line, and "No other people, no text, no logos."

Workshop dance (sample woman, request 01a13904-d888…)

The prompt asks for "a happy little dance between the workbenches", kicked-up heels and a grin. She dances on the spot in the aisle between toy-covered workbenches, under garlands, lifting her heels with her fists pumping at shoulder height. Her face and curls stay close to the photo, nobody else is in the frame and no lettering shows. The handheld camera we asked for barely moves.

Wrapping presents (sample man, request 01a13904-dd75…, back in 56 seconds)

The prompt asks him to wrap a present at top speed, set it on a tall stack and wink. He folds red paper over a box, ties a red ribbon, and then holds a stack of four presents; in the frames we pulled, the stack is not on the bench before that moment. His face and beard stay close to the photo. The wink is half-formed, closer to a lopsided squint.

Tree-side jig (sample woman, request 01a17db2-f604…)

Quick hops from foot to foot, heel kicks with her arms swinging, a full spin that turns her back to the camera, then a lean toward the camera with a grin. Two hands and two feet throughout, no lettering on the costume. The spin frames are blurred.

Snow shuffle (sample man, request 01a17db2-fb25…)

Side steps with loose arms, a shoulder shimmy, a knee lift and a spin with both arms out, on packed snow outside a wooden cabin at dusk with lit windows. The prompt asked for snow puffing up around his boots; the snow shows footprints instead. The clip ends mid-spin with his side to the camera, so the last frame is not a face.

High-kick line (sample young man, request 01a17db3-003a…)

Hands on hips, alternating high kicks, a clap over his head, a jump and a grin, in front of a tall stack of presents and shelves of wooden toys. No lettering on the costume or the presents. For the clap his hands pass in front of his face for a moment, so pick a different frame if you want a still for a card.

4. What went wrong, and what you can do about it

• No bells. Every elf prompt asks for jingling bells; the automated listeners heard none in our two workshop clips. Add a bell sound or a song where you post the clip.
• The wink. Small facial actions are the first thing to go. Our wink came out half-formed. Do not choose a scene for a single small gesture.
• Objects that appear. The present stack in the wrapping clip arrives all at once. It reads as a cut, not as a mistake, but watch for it.
• The ending. The snow shuffle ends side-on. If you want the clip to end on the face, the jig, the high-kick line and the workshop dance all ended facing the camera in our runs.
• Motion blur. Spins blur at 480P. 768P costs more (100 credits for 5 seconds), but we have not recorded an elf clip at 768P, so we cannot show you the difference.

If the face does not look like the person, try a sharper, front-facing photo with even light before another scene.

5. What it costs

On Cicadas, a 5-second 480P clip is 63 credits, about $0.62 on the $9.90 first pack. 10 seconds is 125 credits and 15 seconds is 188; at 768P they are 100, 200 and 300. The quote shows before you generate, a failed generation returns its credits, and a finished clip you do not like is still charged. Each of our five clips cost us about $0.25 at the provider's list price for MiniMax H3 Max ($0.05 a second at 480P); that is an estimate from the list price, not a reconciled bill.

The same one-photo method makes other characters. Our human pigeon AI video test put the same sample portraits on a pigeon, with six tries and the prompts that changed the head size.

Ready to try it? Start on the AI elf video generator with one clear photo.

FAQ

How do I make an elf video from my photo?

Add one clear, front-facing photo of yourself or of an adult who agreed to appear, pick a scene on the AI elf video generator (workshop dance or wrapping presents) or a dance on the elf dance page, confirm the photo and generate. The vertical MP4 lands in My videos.

Does it paste my face onto a cartoon elf?

No. MiniMax H3 Max draws a new scene of the person in the photo wearing an elf costume. In all five of our clips the face and hair stayed close to the sample photo; the costume, the room and the movement are the model's.

Can I use a photo of my child?

No. Both elf pages accept photos of adults only: you, or an adult who agreed to appear. Celebrities and public figures are not accepted either, and you tick a confirmation before each generation.

How much does an elf video cost?

A 5-second 480P clip is 63 credits, about $0.62 on the $9.90 first pack; 10 seconds is 125 credits and 15 seconds is 188. At 768P they are 100, 200 and 300. Each of our five recorded clips cost us about $0.25 at the provider's list price.

Can I write my own elf prompt?

Not on these two pages. The AI elf video generator has two fixed scenes and the elf dance page has three fixed dances, so the costume and the scene are known before you pay. You choose the photo, the length and the resolution.

Will the clip have jingling bells?

Do not count on it. Every elf prompt asks for bells, but in our two workshop clips the automated listeners heard fast tapping like footsteps, rustling paper and two thuds, and no bells. The three dance clips are shown with the sound removed.