Replace the background of a portrait with an original haunted manor garden, matched shadows, fog, and a ready TripoSplat Halloween photo workflow online.
A practical background edit for photos that are already working. Faces, clothing, and poses stay fixed while a new dusk-lit manor garden is fitted behind them.
The prompt asks for matched perspective, shadows, and light direction, not just a flat background. Weathered stone, iron gates, candlelit windows, and silver mist complete the scene.
Copy the exact recipe or open the editor with it already filled in. The instructions separate identity, costume, scene, light, and failure cases so a retry changes one variable at a time.
Use the uploaded photo as the exact identity and pose reference. Preserve every person's recognizable face, body proportions, clothing fit, and natural anatomy. Replace only the background with an original haunted manor garden at dusk: weathered stone steps, black iron gates, twisted bare trees, warm candlelit windows, and low silver mist. Match the new environment's perspective, shadows, and light direction to the source photo so the subject looks naturally placed. Keep the subjects sharp and photorealistic with realistic edge detail. Keep the same number of people as the uploaded photo. Avoid changing faces or clothing, extra people, floating objects, mismatched perspective, text, logos, and watermarks.
Use the supplied photo as the exact first frame. Create a five-second atmospheric background animation without changing the people. Preserve every face, body, clothing, pose, and subject position exactly. Bare branches move slightly, mist drifts across the stone steps, candle flames flicker in the windows, and a soft light travels across the garden. Keep the camera almost still with a very slow push-in. Do not replace the subjects, alter faces or costumes, add people, introduce text, or add watermarks.
Prompts stay in English because the image and video models follow English instructions most reliably. You can edit them after the editor opens.
Open the look that matches the subject you have: one person, a pet, a couple, or a scene.
Use a sharp, well-lit photo you have the right to edit. The prompt protects identity before it changes the styling.
Run GPT Image 2 in the image editor, then copy the same result into the video workspace for a short animation.
The public workspaces keep provider selection behind one generation contract. KIE image routes are live today; the same job can move through fal, Replicate, WaveSpeed, or SeeAPI when those routes are configured.
Low-cost image and image-to-video routes when this provider is configured.
Relay route for supported image and video models when configured.
Primary 3D and motion route; configured media jobs can use it as a fallback.
Queued image and video fallback for supported models when configured.
Active image route for GPT Image 2, Nano Banana 2, Seedream, Flux, and related models.
Provider availability and price follow the runtime configuration. The page never promises a route that the current deployment cannot execute.
Start with one look, then create a matching portrait, pet edit, or short video with the same visual language.
A dark fairytale portrait built from this season’s mascot-horror energy, without recreating a film character. One face stays recognizable while the costume, mane, horn, forest, and light become a complete editorial scene.
An editorial hair-and-wardrobe transformation for anyone who wants the braided-cosplay trend without a salon appointment. The face and pose stay familiar while the styling becomes a complete gothic fashion story.
A glowing fantasy portrait for people who want the cosmic-princess trend with original fashion direction. The model keeps the face and body believable while building a complete observatory scene.
Replace the background of a portrait with an original haunted manor garden, matched shadows, fog, and a ready TripoSplat Halloween photo workflow online.
The instructions preserve every face, body proportion, clothing fit, and pose. The requested edit is the background only.
A believable background has to share the source photo’s camera height, shadow direction, and light color. The prompt asks the model to match those before adding mood.
