AI Halloween Trend 2026: Make a Halloween Video from Your Photo
The AI Halloween trend: turn your photo into a costume video. A one-click ready trend, Seedance 2.5 prompts, AI Halloween costume try-on, steps and mistakes.
The AI Halloween trend is a short video or photo where you appear in costume inside a Halloween scene: at a house party, in a pumpkin field, or in front of a mirror whose reflection turns into someone else. All it takes is one photo of yourself: the model keeps your face and adds the costume, makeup, lighting and motion. The quickest route is the ready-made Halloween Makeover trend: upload a selfie, and a few minutes later party guests build your look on the fly and pull you into the dance. For a scene of your own, use Seedance 2.5 with one of the prompts below, try on an AI Halloween costume as a still in Nano Banana Pro, or copy the moves from someone else's clip in Swap mode.
What the AI Halloween trend is
Every October, feeds fill up with clips where someone walks into a party in ordinary clothes and turns into a vampire, a witch or a zombie within a couple of seconds. These videos spread fast for three reasons:
- A familiar face. The hero isn't an actor but you or a friend, which makes the clip easy to share.
- A transformation on camera. Watching the look change in real time holds attention better than a static portrait.
- A short format. 8–12 seconds, vertical, with music: a natural fit for Reels, TikTok and Shorts.
Search interest climbs through the second half of October and peaks in the final week before the 31st, so it pays to make your clips early. Posting a few days ahead gives the algorithm time to pick them up while people are still looking for costume ideas.
Two ways to make it: a ready trend or your own scene
A ready trend in one click. On the Halloween Makeover page the prompt is already written and tuned: you upload one photo, and Seedance 2.5 does the rest. The story: you arrive at a Halloween party with no costume, four guests dress you up piece by piece as you walk in, then drag you onto the dancefloor. Similar scenes in the same mood include Subway Escape and One Last Embrace, both with zombies; you'll find them and many more in the Trends section.
Your own scene in the studio. If you want a specific costume, location or mood, open a model in the studio, upload your photo and paste a prompt. Below is a table of which model fits which job, plus three prompts you can copy.
Which AI model to use for a Halloween video
| Task | Model | Why |
|---|---|---|
| A video scene with your face: party, transformation, dance | Seedance 2.5 | Keeps the face from your photo via @Image1, second-by-second storyboard, 4 to 30 seconds, vertical 9:16 |
| AI Halloween costume and makeup on a photo | Nano Banana Pro | Changes outfit, makeup and background while keeping your face and pose, up to 4K |
| Animate a finished costume photo | Kling 3.0 | Animates a start frame, 3–15 seconds, 720p or 1080p, aspect ratio taken from the frame |
| Copy a dance from a specific clip | Kling Motion Control | A proven way to transfer motion from a video onto the person in your photo |
| The same motion copy, but more precise and with sound | Swap | A 4–30 s clip plus a photo: the character repeats poses, expressions and lips, the original sound stays, no prompt needed |
A zombie makeover as a still photo is covered in detail in how to turn your photo into a zombie with Nano Banana, and dance trends with motion transfer are in the Kling Motion Control trends guide.
Which photo to upload
Your photo decides whether you'll recognize yourself in the result:
- One person in the frame. In a group shot the model may pick the wrong face.
- Face front-on, in even light. Skip sunglasses, hoods, hands near the face and beauty filters.
- Waist-up or full-body for dancing. With only a face, the model has to invent the body, and the motion looks less natural.
- Only your own photos, or photos of people who agreed. Celebrity photos are rejected by moderation before the generation starts.
Any decent phone photo works. Good light matters more than a high resolution.
Halloween AI video prompts for Seedance 2.5
Seedance understands English and Russian alike. @Image1 is your photo, the first one in the upload list. The costumes are described in plain words with no movie or character names, which keeps refusals down.
1. Costume change at a house party
Vertical 9:16 video, cinematic Halloween house party at night, warm orange string lights, carved pumpkins on the shelves, soft haze, film grain. @Image1 is the main character; the face, hairstyle and body of @Image1 stay identical in every shot. 0–3 s: @Image1 walks through the front door in everyday clothes, looks around and laughs. 3–6 s: the camera circles @Image1, and with each turn the outfit becomes a black velvet vampire cape with a tall collar and pale elegant makeup. 6–10 s: @Image1 stops in the middle of the room, lifts the cape with one arm and smiles into the camera while guests in costumes cheer in the background. Sound: upbeat spooky party music and laughter.
2. Witch in a pumpkin field
Vertical 9:16 video, moonlit pumpkin field at night, blue moonlight with warm candle glow, low ground fog, cinematic film look. @Image1 is a witch in a wide-brim black hat and a long dark coat; the face of @Image1 stays clearly recognizable in every shot. 0–4 s: slow dolly-in as @Image1 walks between rows of glowing jack-o'-lanterns. 4–8 s: @Image1 raises an old lantern, and candles light up one by one along the path. 8–12 s: @Image1 turns to the camera, wind lifts the coat, a flock of crows rises behind. Sound: wind, distant soft choir.
3. The mirror reflection
Vertical 9:16 video, dark bathroom lit by a single candle, retro camcorder texture, gentle flicker. @Image1 stands in front of a mirror; the face of @Image1 stays the same in the room for the whole clip. 0–3 s: medium shot, @Image1 calmly looks at the reflection. 3–6 s: the reflection slowly changes into a zombie version of @Image1 with gray skin, dark circles under the eyes and a torn suit, while @Image1 in the room stays still. 6–9 s: the reflection tilts its head and smiles, the candle flickers. 9–10 s: quick cut to black. Sound: low drone, a single sharp music hit at the end.
What you can swap without breaking the scene: the costume (skeleton, werewolf, ghost bride, pirate), the light color, the location and the final gesture. For how @-references and second-by-second storyboards work, see Seedance 2.5 prompts.
How to make a Halloween AI video from a photo: 5 steps
1. Pick your route
If you want a clip today with no prompt writing, open the ready trend and jump to step 2. If you want your own scene, open Seedance 2.5 in “References” mode.
2. Upload your photo
One sharp photo with your face front-on. In the studio, the first uploaded photo becomes @Image1.
3. Paste the prompt
Copy one of the prompts above and change the costume or location to suit you. Describe what should be in the frame rather than what shouldn't: the model often draws exactly the things you list as forbidden.
4. Set format and length
Vertical 9:16, 8–12 seconds. Run the first test in 480p, and once the face and scene look right, repeat it in 720p or 1080p.
5. Change one thing at a time
If you don't like the result, adjust a single detail: the costume, the light or the moment of transformation. Change everything at once and you won't know what helped. To see the costume as a still first, build the frame in Nano Banana Pro and then animate it in Kling 3.0.
If you want to recreate someone else's Halloween clip
Viral dances and transitions are easier to recreate from the source video than from a prompt. The proven option is Kling Motion Control: upload the clip with the motion and your photo, and the model transfers the moves onto you. For a better result there's the Swap mode: a 4 to 30 second clip plus a photo, and the character repeats the poses, facial expressions and lip movements, while the original sound stays. No prompt needed.
Common mistakes
| Mistake | Cause | Fix |
|---|---|---|
| The face in the clip doesn't look like you | Profile shot, filter or sunglasses in the photo | Use a sharp front-facing portrait in even light |
| The generation is refused | A movie title, character name or actor named in the prompt | Describe the costume in your own words: “a black velvet cape with a tall collar” |
| Something you ruled out shows up in the frame | The model draws words from “forbidden” lists | Write only what should be in the frame |
| The costume drifts between shots | The transformation stretches across the whole clip | Give the change 2–3 seconds, then keep the costume fixed |
| The clip comes out horizontal | The default format was left on | Pick 9:16 before you start |
| The model picked the wrong person | Several people in the photo | Crop the photo down to one person |
What's next
Start with the ready Halloween Makeover trend: one generation shows how well the model keeps your face. Then try your own scene in Seedance 2.5 or an AI Halloween costume on a still photo in Nano Banana Pro. Get it done before October 31: the final week is when these clips get the most views.
Don't want to fiddle with prompts?
Ready-made scenes: upload your photo and get the result in one click.
Models in this article
Frequently asked questions
How do I do the AI Halloween trend with my own photo?
The fastest way is the ready-made Halloween Makeover trend: upload one photo, and Seedance 2.5 builds a clip where party guests put a costume on you on the fly and pull you into the dance. For a scene of your own, open Seedance 2.5 in the studio, upload your photo as @Image1, paste a second-by-second prompt, and set 9:16 and 8–12 seconds.
Can I try on an AI Halloween costume without making a video?
Yes. In Nano Banana Pro, upload your photo and describe the costume and makeup: what to change (outfit, makeup, background) and what to keep (face, hair, pose). You get a still photo in costume that you can post as is or animate later in Kling 3.0.
How long should a Halloween AI video be, and in what format?
For TikTok, Reels and Shorts use vertical 9:16. Seedance 2.5 renders from 4 to 30 seconds, and a costume transformation scene needs about 8–12. Run your first test in 480p and the final version in 720p or 1080p. The cost is shown on the model card before you start.
Why does the AI refuse to make a video from my photo?
Usually it's the photo or the prompt: a profile shot, sunglasses, a heavy filter, several people in the frame, a celebrity, or a movie title or franchise character named in the text. Use a sharp front-facing portrait and describe the costume in your own words. If a generation fails because of a provider error, your tokens are refunded automatically.
How do I recreate a viral Halloween dance from someone else's clip with my face?
Use the Swap mode: upload a 4 to 30 second clip and your photo, and the character repeats the poses, facial expressions and lip movements, while the original sound stays. No prompt needed. Kling Motion Control is the proven older option, but Swap gives a more precise result.
Every model in this roundup is available in the IIshki studio at reduced club prices. Club members get tokens for generations every month.
Join the clubMore articles
- Seedance 2.0 reviews: what it actually does in 2026Seedance 2.0 reviews checked against practice: references, faces, moderation, 15-second clips, quality, and how to run Seedance AI without a VPN.
- AI Winter Photoshoot From Your Photo: 6 Ready Scenes and PromptsHow to make an AI winter photoshoot from your own photo: 6 one-click snow scenes, Nano Banana Pro prompts, keeping your face, and turning snowfall into video.
- AI Dance Video Generator: How to Make a Photo Dance in 2026How to make an AI dance video from a photo: a one-click dance trend, copying moves from any clip with Swap, Seedance 2.5 and Kling 3.0 prompts, steps and fixes.