JoyAI Image Edit: The Essentials

Last updated: May 6, 2026

Covers model_joyai-image-edit

asset image

JoyAI Image Edit on Scenario is an instruction-first image editor. You upload one picture, describe the change in everyday language, and download an edited still that keeps context when you ask for swaps in lighting, props, framing, or small scene rearrangements. It is built for creative iteration, not pixel surgery with a brush.


How JoyAI Image Edit works on Scenario

Start from a clear base image: good light, readable subject edges, and a single main idea per run. Open the model, attach the photo in Image, then write Instructions as if you were briefing another artist. Run the job, review the still, and either tighten the instruction or adjust Guidance and Inference Steps before you spend another credit.

Negative Prompt is optional but useful when a problem keeps returning, for example “no extra text” or “no duplicate faces.” Seeds are optional but powerful when you are polishing copy in the prompt while keeping the rest of the randomness stable.

Write instructions as outcomes, not as UI commands. Say “replace the soda can with a water bottle” instead of pasting tool names. That habit reduces odd artifacts.


Examples

Thumbnail creator: expression swap on a YouTube thumbnail

A content creator uploads a finished YouTube thumbnail featuring a young girl holding a carrot with a neutral, worried expression. Instructions read "change the girl's facial expression to ecstatic, radiant with joy. Preserve the pose, body proportions, hairstyle, and overall appearance. Maintain the original style of the image. Preserve lighting, shadows, and details." Negative Prompt lists "no new props, no changes to the phone screen graphics, no changes to the text overlays." Guidance stays near default, Inference Steps at thirty. Good output shifts only the face while the carrot, the phone with the pig illustrations, and all text badges remain untouched.

Prompt:

Keep the person in the image unchanged, but change their facial expression to match the following instruction: Ecstatic, radiant with joy. Preserve the pose, body proportions, hairstyle, and overall appearance. Maintain the original style of the image, stylized or photorealistic. Preserve lighting, shadows, and details.

Interior design: remove furniture and inpaint the gap

A home staging photographer uploads a living room shot with a cream sofa, a knit throw, plants, and a small black-legged side table beside the couch. Instructions read "remove this coffee table and everything on it from the image. Inpaint the area realistically using nearby visual information, ensuring consistent lighting, perspective, and texture. The final image should look natural and unchanged aside from the removal." No Negative Prompt is needed. If the inpainted floor patch shows a seam, they raise Inference Steps by five and rerun with the same Seed.

Prompt:

Remove this coffee table and everything on it from the image. Inpaint the area realistically using nearby visual information, ensuring consistent lighting, perspective, and texture. The final image should look natural and unchanged aside from the removal.

Portrait retouching: mood lighting pass on an existing photo

A photographer uploads a daylight portrait of a young woman in a beige knit sweater seated in front of a bookshelf with colorful books and a window. Instructions say "the scene is at night, illuminated by warm, flickering candlelight, casting soft shadows, with darkness visible outside the window. Keep the subject's face, clothing, and bookshelf geometry exactly as they are." They raise Inference Steps into the high thirties for a cleaner grade, then reuse Seed while nudging Guidance up by half a point until the candle warmth feels cinematic, not orange.

Prompt:

The scene is at night, illuminated by warm, flickering candlelight, casting soft shadows, with darkness visible outside the window. Keep the subject's face, clothing, and bookshelf geometry exactly as they are.

Animation: isolate background by removing foreground characters

An animator uploads a cartoon illustration of a monkey surfing a giant banana in a tropical jungle, surrounded by parrots and a toucan. Instructions read "remove only the foreground elements from the image to isolate the background. Analyze the scene and remove elements that appear in the foreground only: the monkey, the banana, the birds in flight. Keep everything else intact, including the jungle trees, leaves, flowers, and atmospheric depth. Fill removed areas with contextually appropriate background content that matches the surrounding environment. Output a clean, seamless background plate with no visible artifacts." If residual banana peel lingers near the bottom edge, they add "remove all banana fragments from the lower third" and rerun with the same Seed.

Prompt:

Remove only the foreground elements from the image to isolate the background. Analyze the scene and remove elements that appear in the foreground only: people and characters in the foreground, animals and creatures in the foreground, vehicles in the foreground, objects and props placed in the foreground, any other elements occupying the front layer of the scene. Keep everything else intact, including architecture and structures, landscape and terrain, sky and atmospheric elements, static environmental details, background and midground textures, any elements that are part of the scene's depth beyond the foreground. Fill removed areas with contextually appropriate background content that matches the surrounding environment. Output a clean, seamless background plate with no visible artifacts.

Use Cases

  • Games and real-time art: Iterate skins, props, and lighting on character stills without rebuilding the whole render pipeline.

  • Marketing and growth: Refresh hero shots for seasonal campaigns while keeping product silhouette and label legibility.

  • Film and episodic: Explore grade and framing notes on reference plates before committing a full comp.

  • Education and documentation: Update titles and simple diagram accents on exported slides.

  • E-commerce and catalogs: Swap backgrounds or colorways on standard packshots at volume.

  • Social content: Produce alternate crops or vibe passes from one master still for channel variants.