To make a Halloween video with AI, start with a photo or a short scene description, generate one clear action, then add your music and titles in an editor. A costume reveal needs only one good clip. A party invitation can use three short shots joined together.
Start with the guided costume-reveal example below. It gives you a starting frame, a matching scene recipe, specific ReelMuse settings and a short checklist. Then follow the 15-second editing exercise using three existing clips, or choose another effect from the six references.
How to make a Halloween video with AI: follow this example
The example changes an everyday outfit into a black gothic gown during one turn. Watch it once, then use the steps below to plan your version.
The opening and closing frames below come from the example clip. The settings and prompt are a suggested ReelMuse recipe, not the recorded settings or original production prompt for that clip.
1. Prepare a full-body starting photo
Use a portrait of one adult standing still, with their head, hands and feet visible. Choose a simple background and leave room around the body for the turn. You can start with everyday clothes; this recipe asks the video model to change the outfit.
These frames show the composition and transformation. For your own video, upload your own single photo rather than a before-and-after collage.
2. Select these starting settings in ReelMuse
Open Image to Video, sign in and upload the photo. Keep Image to Video selected and use this setup:
| Control | Select |
|---|---|
| Model | Seedance 2.0 Mini |
| Aspect ratio | 9:16 |
| Duration | 5 seconds |
| Resolution | 480p |
| Last Frame | Off for this single-photo attempt |
At the current configuration, that image-to-video setup costs 60 credits per generation. Check the number beside Generate before submitting, because changing the model, duration or resolution changes the cost. Editing a still image first is a separate generation.
This is a concrete starting setup, not a tested claim that Mini reproduces the reference exactly. Stay at 480p while checking the action; only increase the resolution once you are happy with the shot.
3. Paste this costume-reveal prompt
The adult in the reference photo makes one slow turn in place.
As their back faces the camera, a dark cape sweeps across their body,
briefly hiding the change of clothes. When they face forward, they
wear a long black gothic gown with lace details and a short black
cape fastened with a red jewel. Keep the same recognizable face,
hairstyle and body proportions. Preserve the stone doorway and floor
layout. The fabric settles as they stop. Use one continuous shot,
with a gentle forward camera move after the costume is revealed.
Replace "stone doorway and floor" if your photo has a different setting. Keep the one-turn structure on the first attempt. Do not add another costume, flying, a new character or a second camera move.
4. Generate, inspect and download the clip
Press Generate and open the finished task when it is ready. Check these points before making another version:
- The face remains recognizable before and after the turn.
- The outfit changes once, while the cape hides the body.
- The hands, legs and background do not duplicate or bend unexpectedly.
- The character stops in a readable final pose.
Use the task's download action to save a usable take. If the download dialog offers a watermarked version or an upgrade for the original, choose the version you want before continuing to the editor.
If the clothing change melts or the face drifts, prepare the finished costume in Image Editor first and animate a small turn of that still. To retain a before-and-after reveal, put the everyday and costumed takes next to each other in the editor and cut when the subject faces away. You do not need a perfect transformation inside one generation to finish a reveal.
5. Finish a short edit
A single good clip is enough for a quick costume post. To make a longer Halloween reel, follow the three-clip editing exercise below. It uses the existing reference videos, so you can practise the cut and title before generating anything.
Try the same workflow with another Halloween idea
The costume reveal shows the basic workflow: prepare an image, describe the action, generate a clip and check the result. You can apply those steps to the ideas below. For each one, start by deciding what should happen on screen.
| Your idea | Start with | Keep the action simple |
|---|---|---|
| Costume reveal | A clear, full-body photo | One turn and one outfit change |
| Haunted photograph | A still containing a framed portrait | One movement inside the frame |
| Miniature witch | A prepared image with a figure on a palm | A small wave or bow |
| Witch transformation | A portrait or full-body photo | One spell and one final costume |
| Vampire reveal | A clear head-and-shoulders photo | A slow look toward the camera |
| Pumpkin-head character | A waist-up or full-body photo | One transformation, then a held pose |
For a simpler variation, try a small movement inside a haunted photograph or animate a costume already visible in your starting image. Preparing the look as a still lets you work on the appearance and motion separately, as the optional step below explains.
Optional: prepare the Halloween look before animating
Use this approach when you want a gently moving Halloween portrait, or when you want to settle the costume and background before generating motion. If your starting photo already has the look you want, continue to the video prompts below. Otherwise, open Image Editor, upload your photo and describe the costume and setting. For a new character without a reference photo, use Text to Image.
Replace the bracketed parts in this image-editing prompt:
Edit the uploaded photo into a Halloween portrait. Preserve the adult
subject's facial features, skin tone, hairstyle and body proportions.
Change the outfit to [COSTUME] and place the subject in [SETTING].
Use [LIGHTING] with realistic fabric and natural skin texture.
Keep [FRAMING] and leave space around [HAT / CAPE / HANDS].
Make this a single still image with a coherent background.
For example, choose a black velvet witch outfit, a candlelit doorway, warm amber side lighting and a full-body composition. Inspect the still before animating: a bent hat or extra finger can become more distracting when it moves.
This illustration includes an inset before-image and an arrow; use the final single image you generated as the video input, not the illustrated comparison.
Upload the finished still to Image to Video and ask only for a blink, a small glance or moving fabric. Keep the final prompt focused on one action while you troubleshoot. Runway's video prompting guide provides further guidance on building up motion gradually.
Six Halloween video templates with copyable prompts
The first recipe is the costume reveal you have just followed; the other five apply the same workflow to different effects. Each one explains what image to start with, what motion to request and what to simplify if the result drifts. These are scene recipes, not one-click presets. The videos illustrate the effects, and the suggested test length is independent of the source video's length.
1. A costume reveal during one turn
Use the guided costume-reveal example above for the full clip, opening and closing frames, prompt and suggested settings. Change the final outfit description to a vampire coat or another single Halloween look while retaining the same person and one-turn action.
2. A framed photograph that starts moving
Start with: an image where the entire photograph, frame and surrounding tabletop are visible. A bare selfie does not tell the model which part should be a physical picture.
Starting-image prompt:
A small wooden photo frame on a worn desk, holding a portrait of an
adult looking toward the camera. A single candle stands beside it.
The whole frame and its corners are visible. Cool evening light enters
from the left, with a small warm reflection on the glass. Realistic still.
Video prompt — try a 5-second clip:
Locked close-up of the framed photograph. Only the person inside the
picture moves: their eyes slowly turn toward the viewer, then they blink
once. The wooden frame, desk and room remain still. The candle flame
gently wavers. The portrait stays flat behind the glass and inside the
frame throughout. End on the person's quiet, direct gaze.
If it fails: remove the camera move and any instructions about the character coming out of the frame. The effect depends on a small change inside a stable object.
3. A tiny witch standing on your palm
Start with: an edited still showing an open palm and the miniature character already standing on it. Generate the arrangement before asking for motion; creating a new hand, person and scale relationship at once is a much more complicated request.
Image-editing prompt:
Keep the adult from the uploaded photo recognizable. Show their open
palm near chest height. Place a tiny witch figurine on the center of the
palm, wearing a plum-colored cape and holding a short wand. The figure
has its own clearly visible feet and a coherent small scale. The adult
looks toward it. Warm studio background, clear separation around the hand.
Video prompt — try a 5-second clip:
The tiny witch standing on the open palm gives one small wave, then
lowers her arm. A few warm sparks drift from her wand and fade.
The adult smiles slightly while holding the palm steady. Keep the hand,
fingers and miniature character at the same scale and position.
The camera remains still, with focus on the little witch.
If it fails: remove the sparks and keep only the wave. Avoid asking the character to jump or fly until the model can maintain the hand and the figure's scale.
4. A witch spell and costume transformation
Start with: a full-body image with visible hands and space around the shoulders. Pick a single visual mechanism, such as mist, rather than combining a portal, flight and several costume changes.
Video prompt — try a 5-second clip:
The adult in the reference image slowly lifts one hand. A thin ribbon
of amber mist circles their shoulders, briefly covering the clothing.
As it clears, they wear a dark emerald witch cloak and a tall hat.
Preserve the same face, hairstyle and body proportions. Their feet
remain on the ground. The cloak moves in a light breeze while the
camera makes a gentle forward push, ending in a calm final pose.
If it fails: remove the forward push and shorten the mist effect. For a simpler version, use a finished witch portrait and ask only for the hand gesture and moving cloak.
5. A close-up vampire reveal
Start with: a vampire portrait prepared in Image Editor. Establish the makeup, eyes and coat in the still so the video prompt can concentrate on expression.
Image-editing prompt:
Restyle the adult in the photo as a gothic vampire while preserving
their face shape and recognizable features. Add restrained pale makeup,
deep burgundy eyes and a high-collared charcoal coat. Keep the lips
relaxed. Place them beside a dim window with cool blue side light
and one warm candle in the background. Head-and-shoulders composition.
Video prompt — try a 5-second clip:
The vampire turns their head a little toward the camera and slowly
raises their eyes to meet the viewer. A faint smile reveals two small
fangs. Their face, hair and collar remain consistent with the image.
The candle behind them flickers softly. Use a locked close-up and end
with a steady gaze, retaining natural skin texture and subtle movement.
If it fails: keep the mouth closed and use the gaze alone. Teeth and large mouth movements add another difficult detail to an otherwise simple portrait animation.
6. A glowing pumpkin-head reveal
Start with: a waist-up photo and a clearly visible neck and shoulders. This effect intentionally changes the head, so do not combine it with an instruction to preserve every facial feature.
Video prompt — try a 5-second clip:
An adult stands still in a dark coat. A brief swirl of autumn leaves
passes around their head. When the leaves clear, the head has become
a single carved jack-o'-lantern with a soft amber glow inside.
Keep the same coat, shoulders and body position. The pumpkin rests
naturally above the collar. The character gives a small head tilt,
then holds the pose. Locked camera, cool evening light, playful mood.
If it fails: make the pumpkin-headed still first and animate only the tilt. Leave fire and large body movement out of the test; add more complex effects only after the silhouette stays stable.
Two more looks to make as a still, then animate
The next two ideas use the optional still-first approach above: finish the portrait or pet costume in Image Editor, then animate a small movement in Image to Video. Each has a separate prompt for the image and the motion. These additional ideas do not have matching video previews in this article.
A moonlit portrait with violet braids
Use this in Image Editor with an adult portrait:
Preserve the adult's recognizable face, skin tone and proportions.
Give them long violet braids with small silver clips and a black velvet
Halloween jacket. Place them in an old garden lit by moonlight and
one warm lantern. Keep the portrait photographic, with natural skin
and fabric texture. Leave space around the shoulders and braids.
Upload the resulting still to Image to Video and try a short take:
The subject slowly turns their head toward the camera, then holds
a quiet gaze. A light breeze moves the ends of the violet braids.
Thin mist drifts behind the shoulders. Keep the camera locked and
preserve the face, costume, hairstyle and garden from the image.
If the braids or face distort, reduce the turn to a small glance. Once that looks stable, try a gentle push-in as a separate variation.
A pet in a soft pumpkin costume
Use this in Image Editor with a clear photo of your pet:
Preserve the pet's breed, fur color, distinctive markings, eyes and
natural body shape. Add a soft pumpkin costume fitted around the body,
with a small green leaf detail on the back. Keep the face, ears and
paws visible. Place the pet on a warm autumn studio backdrop with
two small pumpkins nearby. Realistic fur, fabric and natural shadows.
Then animate the edited still:
The pet blinks once and gives a small curious head tilt, then settles
back into its starting pose. Keep the paws planted and the pumpkin
costume resting naturally on the body. Preserve the animal's markings,
face and proportions. The camera and background remain still.
Begin with this small reaction rather than a walk or dance. It gives you a simpler result to inspect before asking the model to move several limbs at once.
Make a 15-second Halloween edit with the existing clips
Once you have a few short clips, you can join them into a longer Halloween reel. To practise that final step now, use the three reference downloads below, then replace them with your own clips when ready. The editing example combines existing clips; it does not require another AI generation.
Editing example: three reference clips, trimmed to five seconds each, joined with straight cuts and a title over the last three seconds. Download the reference edit.
| Timeline | File to import | Keep |
|---|---|---|
| 0–5 seconds | Costume reveal | The first 5 seconds |
| 5–10 seconds | Witch transformation | The first 5 seconds |
| 10–15 seconds | Pumpkin-head reveal | The first 5 seconds |
Put the clips together in CapCut Desktop
- Open New project, then Import the three downloaded files.
- Set the canvas to 9:16. Drag the clips onto one video track in the table's order, with no gaps.
- Trim each clip to 5 seconds. The timeline should finish at 00:15. Leave out transitions for this first edit so they do not overlap the reveals.
- Add a text layer reading Happy Halloween from 00:12 to 00:15. Place it where it does not cover the face. For an invitation, add your date and place as a separate line you can edit.
- Listen to the full cut. Keep the clips' audio, or mute them and add one licensed soundtrack under the whole timeline. Trim the soundtrack to the end and fade it out.
- Choose Export and save an MP4 / H.264 file. For these reference clips, keep 720 × 1280 at 24 fps, matching the sources. Export your own generations at their actual resolution and frame rate; selecting 1080p does not restore detail missing from a 480p source.
The CapCut desktop guide explains its import, timeline and export controls. Labels can vary slightly by version. The downloadable reference edit was assembled from the same files with an equivalent cut-and-title workflow, not recorded from a CapCut session.
Check the file before sharing
Open the exported MP4 outside the editor. Confirm it is vertical, lasts about 15 seconds, includes all three shots without black gaps, and shows the title for the last three seconds. Watch it once with sound and once muted. For an invitation, read the actual date and address at normal playback speed; extend the closing shot if they disappear too quickly.
Fix common generation problems
| Problem | Change to try |
|---|---|
| The face changes halfway through | Use a clearer starting photo, reduce the turn and remove the camera move |
| A hat or hand is cut off | Widen the starting still before animation; leave room for the action |
| A transformation melts between outfits | Cover the change with one physical action, or join two prepared looks in an editor |
| The background drifts | Use a locked camera and specify the objects that should remain still |
| A photo character leaves the frame | Describe motion inside the printed picture and keep the surrounding room static |
| Separate clips look unrelated | Reuse the same character image and repeat the same lighting and material descriptions |
| The invitation text is misspelled | Generate the visual without lettering and add the text in your editor |
| The prompt keeps getting longer | Return to one subject, one action and one camera instruction, then retest |
Frequently asked questions
Can I make a Halloween video from one photo?
Yes. A portrait can become a short character animation in Image to Video. If you want a different costume or setting, prepare that look in Image Editor first and upload the edited still. For multiple shots with the same person, reuse the same reference rather than describing a new person each time.
Do I need first and last frames for a costume transformation?
Not necessarily. You can attempt a change from a single photo and prompt, or edit together two separately animated looks. If the selected model supports a last-frame input, supplying a prepared final costume can help specify the ending. Check the model's available controls rather than assuming every model offers the same inputs.
Which model should I choose?
The guided example uses Seedance 2.0 Mini as a starting point. If you try another model, check that it supports your input, aspect ratio and intended duration. There is no tested winner for the six prompts above: compare a short take with the same starting image before spending credits on a longer version.
How many credits will a clip use?
The cost depends on the selected model, mode, duration and resolution. Check the displayed credit cost before generating and use Pricing for current plans. Image preparation and video generation are separate operations; a prompt that asks for higher resolution does not change the selected resolution or its price.
Are the sample videos the results of these exact prompts?
No. They are reference examples showing the kinds of effects discussed. The copyable prompts are original recipes for making your own version, and results will vary with the model and input image. Use your own photos for your final clip.
Start with one clear shot
Pick one of the six recipes, prepare the still and run a short test in Image to Video. Once the character and movement hold together, make the next shot or add your invitation text in an editor.
References
- Morphic: How to make a Halloween video with AI — workflow reference for planning, generating and editing separate clips.
- DomoAI: AI Halloween creation guide — additional photo-to-video and seasonal scene ideas.
- CapCut: desktop editing guide — import, timeline and export controls used in the editing instructions.
- Runway: Gen-4 video prompting guide — primary guidance on iterating simple motion instructions.
