30 Halloween AI Photo Prompts You Can Copy in 2026
A structured Halloween prompt library for selfies, costumes, pets, couples and backgrounds, with a formula that works across Gemini, ChatGPT and image models.

A Halloween prompt is not a magic spell. It is a short brief with four jobs: protect the subject, describe the costume, explain the light, and exclude the failures you do not want. Once you see the four jobs, you can write a hundred prompts without copying a hundred templates.
Below are 30 prompts organized by use case. They are written for image models that accept long instructions, including the models inside Animate AI's image workspace. You can also paste them into Gemini or ChatGPT and ask for a cleaned-up version. The Halloween AI prompts page has the same prompts in an editable lab.
The four-part formula
Every prompt below uses the same skeleton:
Identity lock + costume + scene and light + camera and exclusions.
The identity lock is essential for selfies and pet photos. For text-to-image, replace it with a character description. The scene and light sentence is what makes the image look photographed instead of pasted. The exclusion sentence is what saves you a retry.
A useful default exclusion line is:
Avoid extra limbs, duplicated fingers, distorted anatomy, text, logos and watermarks.
For pets, add “avoid extra animals, unnatural paws and duplicated limbs.” For couples, add “keep exactly two people and avoid face swapping.”
Try a prompt before you read the list
Prompt lab
6 tested directions
Pick a preset, edit the prompt, then copy it or open the workspace with the prompt already loaded.
531 characters · Die Generierung verbraucht gekaufte Credits
The lab above lets you pick a category, edit the prompt and copy it. The “open generator” button carries the prompt into the workspace, so you do not have to paste it twice.
10 Halloween selfie prompts
- Moonlit selfie. A photorealistic Halloween selfie in a moonlit garden, soft fog behind the subject, warm lantern light from camera left, natural skin texture, 3:4 vertical crop.
- Lantern portrait. A close portrait lit by a single carved pumpkin, warm orange key light, dark autumn porch background, gentle falloff, no gore.
- VHS flash. A 1980s VHS-style portrait with cold direct flash, mild grain, slight scan lines, a dark room behind the subject, face readable.
- Candlelit hallway. A portrait in a long hallway with wall sconces, deep shadows, a cool background and a warm face light.
- Rainy window. A portrait beside a rainy window at night, cool blue ambient light, a warm interior lamp, soft reflections.
- Graveyard dusk. A portrait in an old graveyard at dusk, stone angels, low mist, silver rim light, subdued colors.
- Neon candy. A playful portrait with pink and green neon signs, soft haze, clean skin tones, dark candy-shop background.
- Attic dust. A portrait in a dusty attic with a single skylight, floating dust, warm wood tones, quiet mood.
- Forest path. A portrait on a misty forest path, ferns, fireflies, soft green light, shallow depth of field.
- Studio black. A clean studio portrait on a black backdrop with a single hard rim light and a subtle costume detail.
8 costume prompts
- Gothic lace. A modest black lace gown, high collar, silver jewelry, candlelit stone hall, cinematic period lighting.
- Cosmic royalty. A midnight-blue gown with star embroidery, a silver moon crown, a dark observatory, blue-silver palette.
- Emerald armor. A modest green metallic suit, a simple mask, no logos, a dark industrial set, green rim light.
- Woodland witch. A moss-green cloak with leaf embroidery, a woven belt, a misty forest, natural wool texture.
- Editorial skeleton. Clean white and black face paint, a modest black outfit, a dark studio, fashion-editorial mood.
- Vampire royal. A high-collar velvet cape, antique silver details, a candlelit hall, refined non-gory horror.
- Retro disco. A modest 1970s outfit with a wide collar, warm dance-floor lights, film grain, golden highlights.
- Haunted mascot. A plush horn, a pastel-and-charcoal hood, stitched details, a foggy carnival at dusk.
5 pet prompts
- Pumpkin pet. A soft pumpkin costume with a leaf collar, warm autumn studio, miniature pumpkins, realistic fur.
- Glow dinosaur. A gentle green glow, a dinosaur hood and back plates, a dim studio, no covered eyes.
- Baked potato. A padded brown costume, a butter-pat hat, silver foil accents, warm kitchen light.
- Stick of butter. A soft yellow costume, a simple wrapper label, a bright studio, a warm yellow backdrop.
- Tiny ghost. A soft white sheet with safe eye holes, a dim hallway, cool moonlight and a warm floor lamp.
4 couple prompts
- Cowboy and angel. A modest western outfit and a soft white celestial outfit, a moonlit desert, warm string lights.
- Gothic duo. A black pinstripe suit and a black lace gown, a candlelit hall, deep shadows and red accents.
- Space duo. Two original flight suits with no real logos, a station window, a nebula outside, cinematic sci-fi light.
- Flower and moon. A botanical outfit and a silver moon outfit, a moonlit greenhouse, soft fog and warm lanterns.
3 background prompts
- Haunted mansion. Replace only the background with a distant mansion, bare trees, low fog and a cold moon.
- Pumpkin porch. Replace the background with a porch full of carved pumpkins, warm practical lights and autumn leaves.
- VHS room. Replace the background with a dark 1980s living room, a CRT glow, analog grain and a cold flash.
How to adapt a prompt for Gemini or ChatGPT
Paste the prompt into the chat and add:
Return only the final image prompt. Keep the identity sentence exactly as written. Do not add commentary.
Chat models often add helpful-sounding phrases that weaken the image prompt: “ultra-realistic,” “8K,” “masterpiece,” “trending on ArtStation.” Those words rarely help a modern image model. Ask for a clean prompt with concrete nouns and a clear light direction.
If Gemini rewrites the costume into something you did not ask for, remove the extra adjectives and keep the four-part structure. The model is better at planning than at knowing which details your image model can actually render.
Common failures and the one-line fix
- Face changes: repeat the identity sentence at the end. Shorten the scene.
- Costume too generic: add fabric and construction words: velvet, lace, wool, metallic trim, structured shoulders.
- Background looks painted: add a foreground object, a middle ground and a distant element. Depth is what makes a scene look real.
- Light is flat: name the key light, its direction and its color. “Warm lantern from camera left” is enough.
- Hands are wrong: hide the hands or simplify the pose. Do not ask for complex gestures.
- Image is too dark: raise the key light instead of adding “bright” or “well lit.”
A publishing checklist
- Review the face at 100 percent zoom.
- Check hands, jewelry and hair edges.
- Remove any text or logo the model invented.
- Confirm you have permission to use the source photo.
- Follow the platform's synthetic-media rules.
- Keep a note of the prompt that worked so you can reuse it next year.
Seasonal content is reusable. A prompt that worked in 2026 is an asset for 2027. Save the prompt, the model name and the settings, not just the image.
Frequently asked questions
Can I use these prompts in Gemini?
Yes. Paste the prompt and ask Gemini to return only the final prompt. Then paste that prompt into the image workspace.
Why do my prompts produce different results each time?
Image generation is probabilistic. Keep the identity sentence stable and change one variable at a time so you can compare results.
Should I write prompts in English?
English is the most reliable language for many image models. You can translate the result later; test before publishing.
Are these prompts free to copy?
Yes. Copy and adapt them for your own projects. Review the model provider's terms before commercial use.
Which prompt should I start with?
Start with the moonlit selfie. It has the fewest variables and gives you a clean baseline for judging identity retention.
The bottom line
Thirty prompts are useful, but the formula is more useful. Once you can write an identity lock, a costume sentence, a light sentence and an exclusion line, you can build a Halloween prompt for any direction. Use the Halloween AI prompts lab to edit and test one at a time, and move to the AI Halloween costume generator when you are ready to generate a full portrait.
Was this guide useful?
Halloween-KI-Prompts
Kopiere einen getesteten Prompt für Selfies, Kostüme, Haustiere, Paare, Hintergründe oder kurze Videos und ändere jeweils eine Variable.
Guide lesen