Add, replace, or blend visual elements into an existing image with a text prompt and optional reference image. APOB AI’s Chat to Edit helps preserve the original subject while generating new content that matches the scene.
Step 1: Upload the base image
Open APOB AI, choose Image, select Edit image, and open Chat to edit. Upload the landscape you want to change in the Base image area.
Start with the clearest source available. Decide which subject, position, silhouette, horizon, lighting, and foreground details must remain unchanged before adding anything new.
Step 2: Add the fireworks as a reference
Use the optional reference-image slot to upload the fireworks photo. The second image should guide the burst shapes, colors, glow, smoke, and atmosphere—not replace the base landscape.
Choose a reference with a clear visual idea and tell the model exactly which parts to borrow and which parts to ignore.
Step 3: Describe how the images should blend
Write a precise edit description that separates the protected content from the new generated content. For this example, use the following prompt:
Prompt:
Use Image 1 as the base composition and Image 2 only as the visual reference for the fireworks.
Blend a spectacular fireworks display into the sky of Image 1. Preserve the solitary leafless tree slightly right of center, including its exact shape, branches, trunk, position, scale, and silhouette. Keep the rolling green hill, distant mountains on the left, original perspective, and calm landscape composition unchanged.
Gradually transition the warm orange sunset sky into a natural deep-blue twilight sky while retaining a subtle golden glow near the horizon. Add multiple layered red, gold, blue, and white fireworks inspired by Image 2, with realistic radial bursts, sparkling trails, soft smoke, and atmospheric depth. Place the fireworks behind and around the tree without covering its main silhouette or obscuring the branches.
Make the fireworks cast a subtle, physically realistic glow onto the tree edges, grass, and surrounding sky. Match the landscape’s photographic lighting, depth, color balance, and realistic texture so the final result looks like one naturally captured photograph.
Do not add the crowd, people, buildings, stage lights, or other foreground elements from Image 2. Do not crop, stretch, duplicate, distort, or reposition the tree. Do not change the hill’s shape. Remove the small white number or screenshot artifact in the bottom-right corner. No text, no logos, no unnatural objects, high-resolution photorealistic result.
Step 4: Review and refine the generated fill
Compare the result with the base image at full size. Check that the tree, hill, horizon, perspective, and branch silhouette are still intact, then inspect the fireworks for natural depth, smoke, edge light, and believable color spill.
If the first result is not right, revise one instruction at a time and generate again—for example, reduce the number of fireworks, preserve the tree more strictly, or keep more of the original sunset near the horizon.
After the composite is approved, continue in APOB AI with Face Swap, Edit Clothing, Upscale, or Remove Background when the project needs another focused edit.
Add new visual elements without rebuilding the scene
Introduce props, effects, decor, products, or background details while keeping the useful parts of the source image in place.
Guide the result with text and a reference image
Use language to define the edit and a second image to communicate a specific object, color palette, lighting mood, or visual treatment.
Protect the composition that already works
Describe the subject, pose, silhouette, camera angle, perspective, lighting, shadows, and objects that must stay unchanged.
Refine the image through natural-language feedback
If the first generation is too bright, crowded, soft, or inconsistent, tighten one instruction and generate another version from the same source.
Product and ecommerce scene updates
Add seasonal props, replace an unwanted object, or test a new setting while keeping the product as the protected focal point. Canva’s official AI Replace overview similarly describes adding or replacing objects while preserving the scene’s look, lighting, and shadows.
Marketing and social creative variants
Turn one approved source into multiple campaign concepts by changing selected details instead of reshooting the entire scene. Adobe’s Generative Fill documentation covers adding, removing, and replacing image content with prompts.
Reference-guided composites
Use a second image to communicate the object, mood, palette, or visual treatment you want to bring into the base composition. Adobe documents reference-image guidance for an object or full scene in its Reference Image guide.
Photo cleanup and visual repair
Remove a distraction, reconstruct a small damaged area, or complete an interrupted background. Adobe’s Generative Fill tutorial explains adding, expanding, and removing content while matching scene properties such as lighting and perspective.
Responsible synthetic editing
When an edit materially changes the meaning of an image, keep the original, review the final result, and disclose the change where context requires it. Content Credentials provides an open provenance approach for communicating how digital content was created or edited.
Fast concept iteration
Test a new object, atmosphere, background detail, or campaign direction without reconstructing every pixel manually.
Text plus reference control
Combine a written brief with a visual reference when words alone do not fully communicate the desired look.
Preserves useful surrounding context
A focused edit can keep the approved subject and composition while generating only the area that needs to change.
Fits a broader image workflow
Continue with upscaling, background removal, face swap, or clothing edits after the generated fill is approved.
Generated content is not factual recovery
The model creates a plausible result; it does not recover the real object or scene that existed outside the source image.
Small text and fine details may distort
Labels, fingers, reflections, repeated patterns, thin edges, and tiny typography require close inspection.
Complex lighting and occlusion can need retries
A new object may need several generations before its scale, contact shadow, depth, and interaction with the scene feel natural.
Rights and provenance still need review
Use source and reference images you are allowed to edit, verify commercial usage requirements, and disclose meaningful synthetic changes when appropriate.
Enjoy more features
No Credit Card Required
Practical answers about prompts, reference images, protected details, results, and responsible use.
What is AI generative fill?
AI generative fill creates or replaces selected visual content inside an existing image. It uses a text prompt—and optionally a reference image—to generate pixels that fit the surrounding composition, lighting, perspective, and style.
How does APOB AI Generative Fill work?
In APOB AI, open Image, choose Edit image, and use Chat to edit. Upload a base image, optionally add a reference image, describe what to add or replace and what must remain unchanged, then generate and review the result.
Which APOB AI feature should I use for generative fill?
Use Chat to edit inside Image > Edit image. It accepts a base image, an optional additional image for visual guidance, and a natural-language edit description.
Can I add a new object to a photo?
Yes. Describe the object, location, size, angle, material, lighting, shadow, and relationship to the existing scene. Also name the subjects and areas that must not change.
Can I replace or remove an object?
Yes. Identify the object to replace or remove, then describe how the revealed area should be reconstructed. Review edges, reflections, contact shadows, and repeating textures after generation.
Can I use a reference image with generative fill?
Yes. Upload an additional image to guide a specific object, palette, texture, or visual mood. State clearly what the model should borrow from the reference and what it should ignore.
What is the difference between generative fill and AI Image Merger?
Generative fill focuses on adding, replacing, or reconstructing selected content within a base image. AI Image Merger is better when the main goal is combining broader content from two or more images into one composition.
What is the difference between generative fill and AI Image Expander?
Generative fill changes content inside the current composition. AI Image Expander creates new content beyond the original borders to uncrop a photo or change its aspect ratio.
Can generative fill preserve a person or product?
It can try to preserve the focal subject when the source is clear and the prompt explicitly protects identity, shape, position, color, lighting, and scale. Always inspect faces, hands, labels, logos, and product details before use.
What makes a good base image?
Use the sharpest source available with a clear subject, visible edges, consistent lighting, and enough surrounding context. Very small files, motion blur, dense patterns, and tiny text are harder to edit cleanly.
What should an AI generative fill prompt include?
Describe the new content, its location, scale, camera angle, material, lighting, color, depth, and shadow. Then list the subjects, geometry, perspective, text, logos, and other details that must remain unchanged.
Can I use AI generative fill for product photography?
Yes, for concepting and creative variants. Protect the real product’s shape, color, material, label, proportions, and scale, and verify that the final image still represents the product accurately.
Can I create social media images with generative fill?
Yes. Add decor, effects, environmental details, or clean space for later typography. Create each channel version from a high-quality source and check the final crop at the actual display size.
Can generative fill change a sky or background?
Yes. Describe the desired sky or background and specify which foreground objects, horizon lines, silhouettes, reflections, and lighting cues must be preserved. Complex hair and transparent edges may need another pass.
Does generative fill recover the real missing content?
No. It generates a plausible visual continuation or replacement based on the source and prompt. It does not recover factual pixels or objects that were never present in the uploaded file.
Why do generated fills sometimes show artifacts?
Ambiguous prompts, low-resolution sources, complex occlusion, text, hands, thin edges, reflections, and repeating patterns can cause distortions or visible seams. Narrow the edit and clarify protected details before trying again.
What should I do if the first result is not right?
Correct one issue at a time. Ask for fewer objects, stricter subject preservation, a more accurate horizon, softer light, cleaner edges, or a specific scale, then generate again and compare with the source.
Can I keep editing after generative fill?
Yes. After approving the composite, continue in APOB AI with Face Swap, Edit Clothing, Upscale, or Remove Background when those tools fit the project.
Can I use AI-generated fills commercially?
Commercial suitability depends on the source images, reference rights, subject permissions, brand requirements, and the current APOB AI terms. Verify those requirements and review the final image before publishing.
Is APOB AI Generative Fill free?
APOB AI may provide free credits or account-based access, while generations can consume credits depending on the tool and quality setting. Check the current credit indicator in the workspace before generating.
