Choose a Reference Photo and Image Model
Use a sharp, front-facing or three-quarter selfie with visible eyes, a complete hairline, and even light. For a couple or group prompt, choose a source in which every face is large enough to read.
SelfyzAI's AI image workspace offers multiple image-model choices, including Nano Banana, GPT Image, and the SelfyzAI model. The labels in the live workspace are the current source of truth because availability may change. If you would rather start from a preset look, the AI effects library provides a template-led route.
Retro Flash and VHS Halloween Prompts
1. 1990s Halloween Sleepover
Prompt: Use the uploaded selfie as the identity reference. Preserve the subject's recognizable facial structure, skin tone, eye color, hairstyle, and age. Place her at a late-1990s Halloween sleepover, sitting on the floor beside a carved pumpkin and scattered candy wrappers, wearing black cat-inspired makeup, a velvet choker, and a simple black top. Use direct point-and-shoot flash, a dim bedroom background, orange string lights, candid posture, 4:3 framing, analog grain, soft focus, and slight color bleed. Keep the face unobstructed. No logos, no readable text, no extra fingers, and no copied character costume.
2. Haunted Video-Store Portrait
Prompt: Keep the person in the uploaded selfie recognizable, including face shape, skin tone, eyes, hair, and approximate age. Recreate the portrait inside a fictional late-night video store decorated for Halloween, with generic blank cassette cases, paper bats, fluorescent ceiling light, and a red neon glow from the window. Dress the subject in a dark denim jacket with a small pumpkin pin. Add VHS scan lines, tape noise, slight chromatic fringing, on-camera flash, and a candid waist-up composition. Do not generate brand names, movie titles, readable labels, copyrighted characters, extra hands, or distorted facial features.
3. Y2K Flash Costume Portrait
Prompt: Use the uploaded portrait as the identity anchor and retain the subject's facial proportions, skin tone, eye color, hairstyle, and age. Create a Y2K Halloween party photo with a metallic silver top, smoky eyeliner, tiny bat-shaped hair clips, and glossy black nails. Place her against a silver curtain with orange balloons and soft silhouettes of party guests. Use a harsh compact-camera flash, deep background shadows, red-eye-free natural eyes, slight digital noise, a tilted candid frame, and early-2000s color rendering. No text, no brand logos, no face-covering mask, no extra fingers, and no celebrity resemblance.
4. Retro Pumpkin Patch at Dusk
Prompt: Preserve the uploaded subject's recognizable identity, facial structure, skin tone, hairstyle, eye color, and age. Place her in a pumpkin patch at dusk wearing a rust-colored cardigan, dark skirt, and subtle witch-inspired hat. Add hand-painted wooden signs with no readable wording, hay bales, dry leaves, warm bulbs, and a blue evening sky. Photograph the scene like a faded 35mm family snapshot with direct fill flash, mild grain, slightly muted colors, and an imperfect off-center composition. Keep anatomy natural and the face sharp. No logos, no text, no extra limbs, and no fictional-character styling.
Gothic and Editorial Halloween Prompts
5. Candlelit Gothic Library
Prompt: Use the uploaded selfie as the identity reference and keep the person's face, skin tone, eyes, hair, and age recognizable. Place her in a dark old library wearing a tailored black velvet dress with a high neckline and an antique-style brooch. Add tall shelves, unmarked books, candles in safe holders, and a faint haze. Use warm candlelight on the face, cool shadow fill, a calm direct gaze, realistic skin texture, shallow depth of field, and medium-format fashion photography. No readable book titles, no blood, no mask, no extra fingers, no logos, and no copied character design.
6. Moonlit Ghost Bride
Prompt: Keep the uploaded subject recognizable and preserve her face shape, skin tone, eye color, hairstyle, and approximate age. Transform the setting into a moonlit garden wedding after midnight. Dress her in an aged ivory lace gown with a translucent veil behind the face, faded white roses, and restrained pale makeup. Add wet stone, drifting mist, bare branches, and soft moonlight with a gentle warm reflector on the face. Use elegant editorial photography, muted blue-gray tones, realistic fabric, and a vertical portrait. No wounds, no blood, no face obstruction, no text, and no extra limbs.
7. Crimson Vampire Salon
Prompt: Use the uploaded selfie as the exact identity reference and retain recognizable facial proportions, skin tone, eye color, hair, and age. Create a high-fashion vampire-inspired portrait in a dark crimson salon. Style the subject in a black satin suit with a sculptural collar, subtle burgundy lipstick, and a single vintage ring. Add red curtains, dark wood, candle reflections, and soft atmospheric haze. Use dramatic side lighting, crisp eyes, realistic skin, controlled shadows, and a polished magazine editorial composition. Suggest the theme through styling only: no blood, no oversized fangs, no logos, no text, and no existing character costume.
8. Dark Botanical Sorceress
Prompt: Preserve the identity of the person in the uploaded portrait, including facial structure, skin tone, eyes, hairstyle, and age. Place her in a greenhouse at night wearing a deep green layered gown, a slim metallic headpiece, and botanical jewelry. Surround her with dark leaves, pale moths, condensation on glass, and small points of moonlight. Use a three-quarter pose, soft green rim light, warm light on the face, realistic plant detail, cinematic 50mm photography, and a rich black-green palette. No readable symbols, no face paint that changes identity, no extra fingers, no duplicate jewelry, and no copyrighted fantasy design.
Costume, Couple, and Group Halloween Prompts
9. Skeleton Queen Portrait
Prompt: Use the uploaded portrait as the identity reference. Preserve the subject's face shape, skin tone, eye color, hairstyle, and age beneath elegant skeleton-inspired makeup. Apply a symmetrical black-and-silver bone motif around the eyes and cheekbones without replacing the human face. Dress her in a black gown with a silver crown made of abstract branches. Place her against a smoky charcoal studio backdrop with cool rim light and a soft frontal beauty light. Use sharp editorial photography and a centered composition. No exposed gore, no text, no extra teeth, no extra fingers, and no copied character design.
10. Witch and Alchemist Couple
Prompt: Use the uploaded couple photo as the identity reference and preserve each person's distinct facial structure, skin tone, eye color, hairstyle, age, and position. Style one subject as a forest witch in layered dark-green fabric and the other as an alchemist in a charcoal coat with small glass bottles. Place them at a wooden worktable with herbs, candles, abstract diagrams, and curling fog. Use warm practical lights, cool window moonlight, realistic hands, balanced focus on both faces, and cinematic portrait photography. Do not merge faces, swap clothing, duplicate props, add readable writing, or imitate protected characters.
11. Elegant Haunted Masquerade Couple
Prompt: Preserve both people from the uploaded photo as separate, recognizable identities. Place them at an elegant haunted masquerade in a candlelit ballroom, wearing black and burgundy formal outfits with small decorative masks held in their hands rather than covering their faces. Add chandeliers, faded floral arrangements, soft fog near the floor, and distant dancing silhouettes. Use a composed full-length portrait, warm highlights, cool shadows, realistic fabric, and subtle film grain. Keep both faces clear and hands natural. No logos, no readable text, no extra people in the foreground, and no copied costume designs.
12. Friendly Monster Group Photo
Prompt: Use the uploaded group photo as the identity map. Preserve every person's distinct face, skin tone, hairstyle, approximate age, and left-to-right position. Give each person a different original Halloween styling: mossy forest creature, moonlit witch, elegant vampire, pumpkin magician, and silver ghost, adjusted to the number of people present. Keep the mood playful rather than frightening. Place the group in a decorated photo booth with direct flash, orange and purple lights, and paper decorations. Maintain correct anatomy and separate costumes. No face merging, no extra people, no text, no logos, and no protected characters.
When you want an entire guide devoted to one-person costume transformations, How to Create AI Halloween Costumes includes eight focused looks. For the quicker template-led social format, AI Halloween Trend: How to Make Yours explains the upload-and-generate route.
Decorations and Invitation Prompts
13. Gothic Dinner-Party Decoration
Prompt: Create a realistic Halloween dinner table in a dark dining room. Use black linen, tarnished silver candlesticks, deep red flowers, small white pumpkins, antique-style plates without logos, and a centerpiece of twisted branches. Add warm candlelight, soft window moonlight, gentle haze, and a symmetrical editorial composition. Leave a clean area on the right for typography to be added later. Rich black, burgundy, ivory, and muted gold palette. No people, no readable text, no brand marks, no unsafe open flames near fabric, and no copied franchise imagery.
14. Cute Porch Decoration Concept
Prompt: Design a friendly Halloween front porch with carved pumpkins, paper ghosts, warm string lights, autumn leaves, a dark green door, and a small basket of wrapped candy with blank packaging. Use a cozy twilight atmosphere, realistic home photography, balanced warm and cool lighting, and a welcoming family-friendly mood. Keep the walkway clear and the decorations physically plausible. Leave negative space above the door for optional copy. No people, no readable text, no logos, no frightening gore, and no copyrighted characters.
15. Halloween Party Invitation Background
Prompt: Create a vertical Halloween party invitation background, 4:3 aspect ratio, with layered torn-paper textures, a black-to-deep-purple gradient, small orange moons, abstract bats, fine silver stars, and a decorative frame. Reserve a large uncluttered central area for event text to be added later in a design tool. Use crisp graphic edges, subtle print grain, and high contrast without generating any words. No readable text, no numbers, no logos, no trademarked symbols, and no recognizable fictional characters.
Bad Prompt Examples and How to Fix Them
Too vague: “Make me spooky and cinematic.” This does not define identity, scene, costume, light, or composition. Replace it with a concrete brief such as “preserve my face, waist-up black velvet portrait, candlelit library, warm key light, cool shadows, 50mm editorial photograph.”
Too many concepts: “Make me a witch, vampire, ghost bride, doll, and alien at a carnival in a castle.” Pick one primary subject and one setting. Save alternative concepts for separate generations.
Adapt the Prompts Across Image Models
The same creative brief can be used with Gemini-family image tools, Nano Banana, GPT Image, or SelfyzAI's own model, but the output may differ. Treat the prompt as a specification, not a fixed outcome.
If the model changes the face, shorten the style description and repeat the identity traits near the beginning. If a group prompt merges people, clarify left-to-right positions and remove small props. If the image looks flat, name the key and background lights.
Turn a Halloween Image into Video
Animate the selected still
After selecting a still, the image-to-video workflow can add motion. Ask for one or two actions: fog moving through the background, candle flames flickering, a slow camera push, a small glance, or fabric lifting in a light breeze. The broader AI video generator also provides access to current video routes and templates.
Use references or directed motion
For multiple identity or visual references, reference to video may better match the intended workflow. If body movement is the focus, review the available motion control route. Generated motion can introduce new face, hand, or fabric inconsistencies, so inspect the full sequence before exporting.
Real Limitations
Prompts do not lock an identity or ensure a specific composition. A clear reference and focused instruction can improve the odds, but model interpretation and random variation still matter.
Credits are used for image, video, and effect generation, and the amount varies by model, generation type, duration, and resolution. Check the displayed cost before generating. Free or unsubscribed exports include watermarks, while paid plans list watermark-free downloads; compare current plans and pricing before a final export.
Exact text, detailed hands, dense jewelry, and large groups remain common review points. Generate the visual separately from mission-critical typography. SelfyzAI's published purchase policy also states that all purchases are non-refundable, so confirm the plan and credit terms before paying.
Frequently asked questions
Do Halloween AI photo prompts work in Gemini?
They can provide a useful creative brief, but the output depends on the Gemini model, reference image, settings, and prompt interpretation. Start with identity, then define scene, light, camera style, and exclusions. Review the result rather than treating prompt wording as a fixed outcome.
Which Gemini model is best for Halloween photos?
Use the image-capable model currently available in the product you are using. Model names and access can change, so the live model picker is more reliable than an old tutorial. Test the same source and prompt before deciding which output suits your concept.
What is an easy beginner Halloween AI prompt?
Try: “Use my uploaded selfie as the identity reference. Preserve my face, skin tone, eyes, and hair. Create a waist-up Halloween portrait in a black velvet outfit beside glowing pumpkins at dusk, with realistic photography, warm face light, cool background light, natural hands, no text, and no logos.”
How do I keep my face in an AI Halloween selfie?
Use a clear portrait, put identity instructions first, keep the face unobstructed, and avoid extreme makeup or lighting in the first attempt. If the face changes, simplify the scene before adding more preservation phrases.
Can I use these prompts for men, couples, or groups?
Yes. Change the pronouns and costume details, and identify each subject by position. For groups, ask the model to preserve distinct faces and left-to-right placement, and keep props simple.
Conclusion
The most reusable Halloween AI prompts are structured visual briefs. Preserve identity first, choose one concept, describe a real light source and camera treatment, then remove only the failure modes that matter. Adapt the 15 starting points to the subject and evaluate each image before moving it into video or design.