Guides

How to make a Halloween AI video from a photo

· 6 min read

Get the iPhone app

Short answer: to make a Halloween AI video from a photo, upload a clear photo to an image-to-video model and describe one spooky action in a sentence, such as "her costume changes into a vampire cape as the lights flicker". In Dora that takes a few minutes: choose a model that accepts photos, write the prompt, pick a vertical shape and generate.

Most guides for this agree on the recipe: a clear source image, one change you want to see, and a short seasonal action in the prompt (Seedance, Media.io). The rest of this guide shows it with Dora's real controls.

Step 1: Pick a photo the model can work with

The video starts from your photo, and the model redraws every frame from it, so a weak photo gets guessed again and again.

  1. Use a front-facing photo with even light and the face fully visible. Sunglasses and a hand over the face cause trouble.
  2. For a costume change, a half-body or full-body shot works better than a close-up, so the outfit is in the frame.
  3. One person per photo is the safest start. Groups give more faces to get wrong.
  4. For a pet or a house, any sharp, well-lit photo will do.

Step 2: Choose the effect

  • Costume transformation: your outfit changes into a witch, vampire or skeleton.
  • Ghost or fade: the person becomes see-through, or a ghost appears behind them.
  • Cursed portrait: a still portrait blinks, turns its eyes and smiles slowly.
  • Zombie or monster reveal: the face changes at the end of the clip.
  • Spooky scene: the same person placed in a graveyard, a haunted house or a pumpkin patch.

Step 3: Make the video in Dora

  1. Open the create page and switch to video.
  2. Pick a model that accepts a photo. Seedance 2 Mini, Kling 3.0 Turbo, Kling Omni and Gemini Omni all do. The models page shows each one's cost in coins before you spend anything.
  3. Upload your photo.
  4. Write the prompt with the subject, the one action, the mood and the camera (examples below).
  5. Choose 9:16 for TikTok, Reels and Stories, or another shape if you need it. Seedance 2 Mini offers 9:16, 16:9, 3:4, 4:3, 1:1 and 21:9.
  6. Keep it to 4 to 6 seconds, then generate. The result lands in your gallery.

Want the same person in a different scene? Same Character takes up to 7 photos of one character and places them in any scene you describe, which suits "me in a graveyard" better than a single-photo animation. SeeDance 2 and Gemini Omni also accept up to 7 reference photos.

Halloween prompts to copy

  • Costume change: "The woman in the photo spins once and her outfit turns into a black witch dress and pointed hat, orange glow from pumpkins behind her, camera holds steady. Keep her face the same."
  • Ghost: "The man slowly becomes transparent and a pale blue glow appears around him, fog on the floor, slow push-in."
  • Cursed portrait: "The portrait blinks slowly, the eyes follow the camera, a faint smile, candlelight flickers. Static camera."
  • Zombie reveal: "The person stays still, then the skin turns grey and the eyes go white in the last second, dark street, handheld camera."
  • Pet in costume: "The dog sits and a tiny pumpkin costume appears on it, it wags its tail, warm porch light."
  • Graveyard scene (Same Character): "The same person walks through a foggy graveyard at night holding a lantern, moonlight, slow tracking shot."

Tip: say what must stay: "keep the face and the background the same". One guide from Media.io makes the same point for its ghost effect.

Make it look better

  1. One change per clip. A costume swap, a ghost fade and a zombie face in four seconds will fail. Make three clips and cut them together in your editor.
  2. Say the light. "Flickering candle", "cold moonlight" and "orange pumpkin glow" give the clip a mood more than the word "scary".
  3. Try two models with the same photo and keep the better take.
  4. Add sound if you want it: ElevenLabs SFX v2 in Dora makes sound effects from a text description, such as a creaking door or wind.
  5. Check the face against the original photo before you post.

What can go wrong

  • The face drifts. The more the prompt changes, the more the face can change. Add "keep her face the same" and shorten the clip.
  • Costumes look generic. Name the details: colors, fabric and props.
  • Hands and text. Fingers and any lettering are the usual weak spots.
  • Filters. Some models refuse photos they flag as unsafe or of public figures; another model may accept the same photo.
  • Consent. Only use photos of people who agreed. Making a friend or a child look like a zombie is a joke only if they are in on it.
  • Labels. TikTok requires creators to label AI-generated content that shows realistic-looking scenes or people (TikTok Community Guidelines). A clearly fake monster still deserves the label if it uses a real person's face.

Other tools for Halloween videos

Dora is a general video generator and has no Halloween template pack. If you want one-click spooky effects, these are worth a look:

Where Dora is different: you choose from several video models, you control the prompt, and the cost in coins is shown before each generation. Where template tools are better: a guided look with no prompt writing.

Make your Halloween clip →

For a gentler movement, see how to animate any photo, and for more prompt wording, the AI video prompt guide.

Frequently asked questions

How do I make a Halloween video from a photo with AI?

Upload a clear photo to an image-to-video model, describe one spooky action such as a costume change or a ghost fade, pick a vertical shape and generate a 4 to 6 second clip. In Dora, Seedance 2 Mini, Kling 3.0 Turbo, Kling Omni and Gemini Omni all accept a photo.

What is the best photo for an AI costume transformation?

A front-facing, evenly lit half-body or full-body photo with the face fully visible. Avoid sunglasses, hands over the face and busy group shots.

Can I put myself in a graveyard or haunted house scene?

Yes. In Dora, Same Character takes up to 7 photos of one person and places them in any scene you describe. SeeDance 2 and Gemini Omni also accept up to 7 reference photos.

Can I make my pet a Halloween video?

Yes. Use a sharp, well-lit photo of the animal and describe one small action, such as a costume appearing and a wagging tail.

Do I have to label an AI Halloween video?

TikTok requires creators to label AI-generated or significantly edited content that shows realistic-looking scenes or people. Check the rules of each platform you post on.

Sources

Get the iPhone appUse it in Claude or ChatGPT

Keep reading

GuidesHow to animate any photo into a videoBring a still photo to life — the image-to-video models on Dora and how to prompt them.GuidesHow to turn a photo into a video on iPhoneA Live Photo saved as video, a slideshow or memory movie with music, or a still photo brought to life by AI. The steps for each, and which one you actually need.PromptsAI video prompt guide: 30 prompts that actually workA repeatable prompt formula plus 30 copy-paste prompts for cinematic, product, and social video.