English quick take
This page gives you three tested prompt recipes, a local filter-and-grain editor, and a provider routing table so you can run the same look in the model you already pay for.
Girigo AI Studio 10 min read
AI Horror Photo Editor
Start with a photo you own, keep the face recognizable, and build a cinematic horror portrait without handing your image to an unknown upload box.
Quick answer
Local look lab
Preview the mood before you spend a credit
Upload a photo you own. The canvas, grade, grain, and caption stay in this browser.
No uploadNo accountPNG export
Choose a photo to start. Nothing leaves this browser.
Prompt studio
Start from a structured brief, not a blank box
Pick a look, set the direction, and copy a prompt that names what must stay fixed.
Direction: cinematic portrait; preserve identity; one person.
Use the uploaded photo as the primary reference. Preserve the subject's recognizable facial identity, facial proportions, skin tone, approximate age, and natural anatomy. Transform the subject into an original gothic Halloween fashion portrait. Change the hairstyle into long violet braids with delicate silver hair accessories. Replace the clothing with an elegant black-and-purple velvet outfit with subtle lace and metallic details. Place the subject in a moonlit Victorian garden with soft fog, glowing lanterns, and dark autumn foliage. Use photorealistic editorial photography, realistic fabric texture, natural skin detail, and cinematic soft lighting. Keep the same number of people as the uploaded photo. Do not alter facial identity, add extra limbs, duplicate fingers, distort anatomy, or add text, logos, or watermarks.
Three useful outcomes
What you can make on this page
Portrait
A haunted portrait that still reads as you
The prompt names what must stay fixed before it names the costume.
Open the studio
Local edit
A free preview before you spend credits
Upload locally, tune grain, vignette, and caption, then export a PNG.
Open the studio
Routing
The same look across five providers
Compare a budget route, a fast route, and a character-focused route.
Open the studioPrompt lab
Three copy-ready prompt recipes
Gothic portrait
Midnight Braids
Best first test for a cheap identity-preserving edit.
Use the uploaded photo as the primary reference. Preserve the subject's recognizable facial identity, facial proportions, skin tone, approximate age, and natural anatomy. Transform the subject into an original gothic Halloween fashion portrait. Change the hairstyle into long violet braids with delicate silver hair accessories. Replace the clothing with an elegant black-and-purple velvet outfit with subtle lace and metallic details. Place the subject in a moonlit Victorian garden with soft fog, glowing lanterns, and dark autumn foliage. Use photorealistic editorial photography, realistic fabric texture, natural skin detail, and cinematic soft lighting. Keep the same number of people as the uploaded photo. Do not alter facial identity, add extra limbs, duplicate fingers, distort anatomy, or add text, logos, or watermarks.
Avoid: identity drift, plastic skin, extra fingers, extra limbs, text, logo, watermark
Quiet horror
Moonlit Widow
Use when you want atmosphere instead of costume spectacle.
Edit the uploaded portrait into a restrained Victorian ghost story still. Keep the subject's face, age, skin texture, body proportions, and pose recognizable. Add a sheer black veil, a high-collar charcoal dress, and a single silver brooch. Move the scene into an abandoned glass greenhouse at night, with wet stone, ferns, and a cold moonbeam cutting through the roof. Use soft volumetric light, realistic fabric, subtle film grain, and a desaturated blue-green palette. The mood should be melancholy, not gory. Keep one person only. Avoid a new face, changed ancestry, extra people, deformed hands, floating objects, captions, and watermarks.
Avoid: gore, new face, changed ethnicity, extra people, deformed hands, captions
Analog horror
Static Saint
Good for testing how a model handles texture without losing identity.
Use the uploaded photo as the identity reference. Keep the subject recognizable while turning the portrait into an analog-horror broadcast still. Dress the subject in an ivory ceremonial coat with a cracked porcelain halo behind the head. Add restrained VHS scanlines, a faint timecode in the lower corner, and a cold fluorescent light from above. The scene is a small chapel with peeling paint and one red exit sign. Keep the face intact and the anatomy natural. Do not add a second face, a monster mouth, extra arms, readable brand text, or a watermark.
Avoid: monster face, second face, extra arms, brand text, watermark, heavy blur
WaveSpeed, SeeAPI, fal, Replicate, Kie
Which model should run this prompt?
| Provider | Best for | Typical route | Docs |
|---|---|---|---|
| WaveSpeedBudget-first | Low-cost image edits and image-to-video drafts | Seedream edit, FLUX Kontext, Wan image-to-video | Docs |
| falSpeed-first | Fast hosted queues and strong model coverage | FLUX, Kling, and other hosted media models | Docs |
| ReplicateExperiment-first | Broad catalogue for experiments and batch comparisons | Community and first-party image/video models | Docs |
| SeeAPIRouting layer | One API surface for several model families and fallback routes | Image and video inference endpoints | Docs |
| KieCharacter-first | Nano Banana image edits and Kling video routes | Nano Banana, Kling video | Docs |
Step by step
A workflow that protects identity
- Upload a photo you have the right to edit. A front-facing photo with even light is easiest.
- Pick one preset and read the negative constraints before you change anything.
- Use the local editor to test crop, contrast, grain, and caption placement for free.
- Copy the prompt into your chosen provider and keep the reference image at full resolution.
- Compare the result against the original face before you upscale or share it.
Failure modes
Where most attempts fail
- Do not ask the model to make a different person; identity preservation is the whole job.
- Do not use a low-resolution screenshot as the reference and then blame the model for soft eyes.
- Do not stack three horror styles in one prompt; pick one lighting idea and one wardrobe idea.
- Do not publish a portrait of someone else without consent, even if the output looks stylized.
- Do not add text in the image model. Add captions with the local editor or a design tool.
Editorial brief
AI Horror Photo Editor: the longer brief
Most horror portrait prompts fail for a boring reason: they spend all their words on the costume and none on the face. The model then invents a new person. This page flips the order. First it locks identity, skin texture, age, pose, and the number of people. Then it adds wardrobe, location, and light. That order is why the same prompt can work across different image-edit models.
The local editor is the cheap part of the workflow. Use it to test a crop, a colder grade, a vignette, and a short caption before you spend a single credit. When the composition feels right, copy the prompt into a provider and compare the generated face against the original at 100 percent zoom. If the eyes, jawline, or skin texture drift, stop and fix the reference or the prompt before you generate more variations.
Deeper guide
Go deeper before you spend credits
Read the deeper guide
The prompt is only half the job. The field note walks through the failure modes, the identity check, and the model comparison behind this page.
Open the field noteFAQ
Questions this page should answer fast
Is this AI horror photo editor free?
The local editor on this page is free and runs in your browser. Model generation costs depend on the provider you choose; check the provider dashboard before you run a batch.
Does the page upload my photo?
No. The local canvas editor reads the file in your browser. The page does not send the image to a server. If you copy a prompt into a provider, that provider's own privacy terms apply.
Which model is best for a haunted selfie?
Start with an image-edit model that accepts a reference image, such as a Seedream edit route on WaveSpeed, a FLUX Kontext route, or a character-focused route through Kie. Compare one cheap run and one higher-quality run before committing to a batch.
Can I use the result commercially?
That depends on the model license, the provider terms, and whether you have rights to the source photo. Keep a record of the model and license, and do not use a real person's likeness without permission.
Checked sources