How to Make the 24K Magic AI Video From 2 Photos (4 Steps)
What’s new on this page
- — Published: our first 24K Magic AI video (Wan 3.0, 14 s, 480p, 280 credits, first try, about 5 minutes), the full prompt, the task screenshot, who plays which role and photo tips.
A 24K Magic AI video puts you into Dracula's cruise ship deck dance from Hotel Transylvania 3 as a real person: a grin and a collar tug in a cabin doorway, then the whole routine across a sunny deck. Ours adds a second act, where a friend crawls in and takes over. We made the clip above on 3 November 2026 from two photos. Here's exactly what we did, what it cost and how to make your own.
What happens in a 24K Magic AI video?
| Seconds | What happens |
|---|---|
| 0–2 | The dancer grins in a cabin doorway and tugs their collar |
| 2–4 | Wide shot: the dance starts on the sunny deck, sea behind |
| 4–6 | Low close-ups of feet sliding and kicking on the boards |
| 6–10 | Kicks, knee drops and arm swings by the white stairs |
| 10–11 | The friend crawls into the foreground |
| 11–14 | The friend takes over the deck by the ship's funnels |
How we made our first 24K Magic AI video
1. We started from a 14-second remake of the deck dance
The scene came back on TikTok on 11 October 2026, when @jst.daf posted it as a Hotel Transylvania 3 mashup, and live-action AI remakes followed. We used a 14-second live-action remake as the reference video, so the shots, cuts and footwork were already there.
2. We used one photo per person, each with a loud look
Photo 1 was a man with a green mohawk, a studded leather vest and red tartan trousers; photo 2 a man with tall orange hair, mirrored sunglasses, a teal flame shirt and white shorts. Each image paired a face close-up with a full-body shot, so the model saw both the face and the whole outfit.
3. We told the model who plays which role
The remake has two people, so the prompt names them: the main dancer becomes Image 1, the person who crouches in the foreground near the end becomes Image 2, and the original racing suit is dropped. We ran Wan 3.0, 480p, 14 seconds, sound on:
Replace the people in the reference video with the people in the reference photos. Keep the reference video's camera work, cuts, timing, choreography, ship-deck setting and lighting exactly the same. - The main dancer (doorway close-up, then the shuffle dance on the ship deck) becomes the person in Image 1: same face, hairstyle, body shape and outfit as Image 1. Do not keep the original racing suit. When the dancer pulls up a hood, show it only if Image 1's outfit has a hood; otherwise skip that motion naturally. - The second person who crouches in the foreground near the end becomes the person in Image 2: same face, hairstyle and outfit as Image 2. Both people's identities stay consistent in every shot, including the low-angle footwork close-ups. Realistic skin and fabric, natural motion blur, no face distortion, no swapping between the two people.
4. It rendered in about 5 minutes
We submitted at 09:28 UTC and the video was ready at 09:33, first try, 280 credits. The mohawk, studded vest and tartan trousers held in every shot, including the low footwork close-ups, and the friend kept his hair, sunglasses and flame shirt. Two things didn't come from the photos: the dancer wore the routine's teal sneakers instead of his black boots, and in the last second the friend briefly appears twice in the frame. That run is what the template now uses.
How to make a 24K Magic AI video in 4 steps
Step 1: Open the template
Go to the 24K Magic AI page. The template is already loaded at the top, with the model and prompt set up.
Step 2: Add the dancer
The Dancer slot gets the doorway close-up and the whole deck routine, so use a sharp, full-body photo with the face to the camera.
Step 3: Add the friend
The Friend slot crawls in around 10 seconds and finishes the dance. Nobody willing? Click Use example under that slot and our sample person crawls in instead.
Step 4: Generate and download
Press Generate. The button shows 280 credits before anything is charged. The video lands in My Videos, so you can close the tab. Download the MP4, which has no watermark, and post it.
Which photo plays which role in the 24K Magic AI video?
| Slot | Role | Best photo |
|---|---|---|
| Photo 1 | Dancer: doorway close-up, deck routine, 0–10 s | One person, full body, face clear |
| Photo 2 | Friend: crawls in at 10 s, dances to the end | One person, full body |
What photos work best for a 24K Magic AI video?
- One person per photo. A face cut out of a group shot can't hold up in the opening close-up.
- Full body. The low footwork shots show your legs; a face-plus-full-body image worked well for us.
- A loud outfit. A punk on a white cruise deck is half the joke.
- Clear face. No caps pulled low or hands over the face; the grin in the doorway is the shot everyone remembers.
- Adults only. Photos of yourself or adults who said yes. No children, celebrities or public figures.
Do you turn into Dracula in the 24K Magic AI video?
No. You stay a real person with your own face, hair and clothes; there's no cape, fangs or pale makeup unless they're in your photo. The cruise deck, the cuts and the dance come from the template; the people come from your photos.
That's why we packaged it: the role order, the no-racing-suit line and the model are already set, so you don't pay to rediscover them. Make your 24K Magic AI video with the same settings.
Is 24K Magic AI free?
Not completely. New accounts get 75 free credits and one video takes 280, so you need the one-time $4.99 pack (540 credits). That covers one video at about $2.59 and leaves 260 credits, with no subscription. See pricing. Failed renders are refunded automatically.
How long is the video and how much does it cost?
About $2.59 for one video: 280 credits for a 14-second, 480p widescreen clip with sound. Ours rendered in about 5 minutes.
What song is the 24K Magic AI trend?
Bruno Mars's 24K Magic, the 2016 single Dracula dances to in Hotel Transylvania 3. The AI video comes with music; to use the exact trending sound, add 24K Magic from TikTok's library when you post.
Can I make it with only one photo?
Yes. Upload yours as the dancer and click Use example under the friend slot, so our sample person crawls in at the end.
Why doesn't the dancer look like my photo?
Usually it's a small face cropped from a group shot, a side angle or a cap pulled low. Use a sharp, full-body photo of one person facing the camera and run the 24K Magic AI template again.
Is this official Hotel Transylvania content?
No. It's a fan-made template, not affiliated with Sony Pictures Animation, Hotel Transylvania, Bruno Mars or TikTok.
Want three of you next time? Our Chipmunks Dance AI tutorial walks through a trio dance template.