August 29, 2026AIExplore

How to Use Adobe Firefly Generative Fill to Add or Replace Objects

Paint a precise selection in Adobe Firefly Generative Fill, describe only the fill region, refine the mask, and keep surrounding lighting consistent.

Adobe Firefly

Adobe Firefly Generative Fill adds or replaces content inside a brushed selection while keeping the rest of the frame. Official Adobe Generative Fill help walks through Edit image, Fill, brush selection, optional references, and a text prompt for the painted region. Adjust brush size and hardness. Use Select less or Select more to clean the mask. Credit use can vary by model. Check the indicator beside Fill. On paid Firefly plans, standard image features such as Generative Fill may have unlimited access that does not deduct credits. Confirm live plan details. Overview: /explore/firefly.

This guide covers add and replace workflows with precise masks and short fill prompts. Pair with Remove at /blog/how-to-use-adobe-firefly-generative-remove, Text to Image at /blog/how-to-use-adobe-firefly-for-text-to-image-generation, and Expand at /blog/how-to-use-adobe-firefly-generative-expand.

When Generative Fill is the right job

Use Fill when the composition is mostly right and only a region must change. Prefer Text to Image when the whole scene is wrong. Prefer Remove when the goal is erasure without a new object.

Step by step workflow

1. Open Edit image and choose Fill

Load a Firefly generation or an upload. Select Fill. Check the credit indicator for the active model before large edit batches.

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: Session card Source: kitchen product shot Tool: Generative Fill Model indicator: review before generate Goal: add vase on table. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "Session card Source: kitchen product shot Tool: Generative Fill Model indicator: review before generate Goal: add vase on table" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Inputs:
Use the actual source material required for this task, together with the intended audience, destination, format, and quality requirements. Specify relevant files, text, URLs, structured data, reference assets, dimensions, language, tone, timing, model or generation settings, credentials, field mappings, or integration values when they apply. Keep secrets out of the example and use the product's supported credential or configuration mechanism.

Workflow:
1. Start with one representative input that contains the important characteristics of the real workload.
2. Configure adobe-firefly specifically for the requested task and make the important settings explicit.
3. Run the first pass and inspect the result against the objective.
4. Correct only the settings or source material responsible for a failed requirement instead of changing everything at once.
5. Validate the intermediate result before passing it to the next step.
6. When the result meets the requirements, save the successful configuration so the same process can be repeated consistently.
7. If this is a multi-step workflow, verify each handoff and keep a clear fallback or review path for failures.

Requirements:
The example must use supported functionality only. Do not invent product features, pricing, limits, integrations, model names, API behavior, or unavailable settings. Preserve important source data and formatting. Validate required fields before processing, handle empty or invalid input explicitly, prevent accidental duplicate processing where relevant, and stop or route the item for review when a required step fails. For credentials, use the supported secure configuration rather than placing secrets directly in the content.

Expected output:
The final result should directly satisfy the "Session card Source: kitchen product shot Tool: Generative Fill Model indicator: review before generate Goal: add vase on table" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

2. Paint a precise mask

Brush only the change region. Use Select less or Select more after the first pass. Soft edges help blend; hard edges help protect nearby subjects.

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: Mask checklist [ ] Brush covers only the target [ ] Edges do not cut into faces or logos you must keep [ ] Select less or more applied if needed [ ] Hardness matched to edge type. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "Mask checklist [ ] Brush covers only the target [ ] Edges do not cut into faces or logos you must keep [ ] Select less or more applied if needed [ ] Hardness matched to edge type" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "Mask checklist [ ] Brush covers only the target [ ] Edges do not cut into faces or logos you must keep [ ] Select less or more applied if needed [ ] Hardness matched to edge type" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

3. Prompt the selection only

Describe what belongs inside the painted region. Name lighting continuity. Avoid rewriting the entire scene.

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: Fill prompt Object: ceramic bud vase with one white flower Continuity: match kitchen window light Locks: keep surrounding objects unchanged. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "Fill prompt Object: ceramic bud vase with one white flower Continuity: match kitchen window light Locks: keep surrounding objects unchanged" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "Fill prompt Object: ceramic bud vase with one white flower Continuity: match kitchen window light Locks: keep surrounding objects unchanged" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

Practical Generative Fill prompts

Add bud vase

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: [brush empty table] A small ceramic bud vase with one white flower. Match kitchen lighting. Keep surrounding objects unchanged.. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "[brush empty table] A small ceramic bud vase with one white flower. Match kitchen lighting. Keep surrounding objects unchanged." that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "[brush empty table] A small ceramic bud vase with one white flower. Match kitchen lighting. Keep surrounding objects unchanged." requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

Replace chair

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: [brush the chair] Replace with a mid century walnut chair, same camera angle, same window light, realistic wood grain. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "[brush the chair] Replace with a mid century walnut chair, same camera angle, same window light, realistic wood grain" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "[brush the chair] Replace with a mid century walnut chair, same camera angle, same window light, realistic wood grain" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

Add product prop

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: [brush blank desk corner] Add a closed matte notebook and a black pen, aligned to the desk edge, same soft overhead light. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "[brush blank desk corner] Add a closed matte notebook and a black pen, aligned to the desk edge, same soft overhead light" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "[brush blank desk corner] Add a closed matte notebook and a black pen, aligned to the desk edge, same soft overhead light" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

Replace background patch

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: [brush busy poster area] Soft neutral wall texture matching the room color temperature, no text, no patterns. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "[brush busy poster area] Soft neutral wall texture matching the room color temperature, no text, no patterns" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "[brush busy poster area] Soft neutral wall texture matching the room color temperature, no text, no patterns" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

Add plant

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: [brush empty floor] A potted monstera in a terracotta pot, matching floor reflections, natural daylight. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "[brush empty floor] A potted monstera in a terracotta pot, matching floor reflections, natural daylight" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "[brush empty floor] A potted monstera in a terracotta pot, matching floor reflections, natural daylight" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

Replace packaging

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: [brush bottle label area] Blank cream label with soft paper texture, no readable text, keep bottle shape and lighting. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "[brush bottle label area] Blank cream label with soft paper texture, no readable text, keep bottle shape and lighting" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "[brush bottle label area] Blank cream label with soft paper texture, no readable text, keep bottle shape and lighting" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

Add sky detail

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: [brush clear sky region] Soft afternoon clouds matching existing sun direction, subtle haze, no aircraft. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "[brush clear sky region] Soft afternoon clouds matching existing sun direction, subtle haze, no aircraft" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "[brush clear sky region] Soft afternoon clouds matching existing sun direction, subtle haze, no aircraft" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

Replace food garnish

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: [brush plate garnish] Fresh basil leaves and lemon zest, appetizing color, same overhead light. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "[brush plate garnish] Fresh basil leaves and lemon zest, appetizing color, same overhead light" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "[brush plate garnish] Fresh basil leaves and lemon zest, appetizing color, same overhead light" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

Add reflection

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: [brush table surface near mug] Subtle soft reflection of the mug, matching wood grain and window light. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "[brush table surface near mug] Subtle soft reflection of the mug, matching wood grain and window light" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "[brush table surface near mug] Subtle soft reflection of the mug, matching wood grain and window light" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

Replace shoes

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: [brush shoes] White leather sneakers with soft scuffs, same stance and shadow direction. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "[brush shoes] White leather sneakers with soft scuffs, same stance and shadow direction" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "[brush shoes] White leather sneakers with soft scuffs, same stance and shadow direction" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

More copyable examples

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: Add hand prop [brush empty hand] Holding a takeaway coffee cup with blank sleeve, no logos, match skin tone lighting. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "Add hand prop [brush empty hand] Holding a takeaway coffee cup with blank sleeve, no logos, match skin tone lighting" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "Add hand prop [brush empty hand] Holding a takeaway coffee cup with blank sleeve, no logos, match skin tone lighting" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: Replace pillow [brush sofa pillow] Linen throw pillow in muted sand, soft folds, same sofa perspective. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "Replace pillow [brush sofa pillow] Linen throw pillow in muted sand, soft folds, same sofa perspective" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "Replace pillow [brush sofa pillow] Linen throw pillow in muted sand, soft folds, same sofa perspective" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: Add path stones [brush path gap] Continuation of cobblestones matching wetness and color, no new objects. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "Add path stones [brush path gap] Continuation of cobblestones matching wetness and color, no new objects" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "Add path stones [brush path gap] Continuation of cobblestones matching wetness and color, no new objects" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: Replace lamp [brush lamp] Minimal brass floor lamp, warm bulb glow, same room exposure. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "Replace lamp [brush lamp] Minimal brass floor lamp, warm bulb glow, same room exposure" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "Replace lamp [brush lamp] Minimal brass floor lamp, warm bulb glow, same room exposure" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: Add window plant [brush sill] Small succulent in a white ceramic pot, soft backlight, keep sill edge clean. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "Add window plant [brush sill] Small succulent in a white ceramic pot, soft backlight, keep sill edge clean" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "Add window plant [brush sill] Small succulent in a white ceramic pot, soft backlight, keep sill edge clean" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: Replace car color [brush car body] Deep forest green paint with realistic specular highlights, keep shape and reflections. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "Replace car color [brush car body] Deep forest green paint with realistic specular highlights, keep shape and reflections" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "Replace car color [brush car body] Deep forest green paint with realistic specular highlights, keep shape and reflections" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: Add mist [brush midground air] Soft morning mist matching the forest palette, no new subjects. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "Add mist [brush midground air] Soft morning mist matching the forest palette, no new subjects" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "Add mist [brush midground air] Soft morning mist matching the forest palette, no new subjects" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

Scenario:
A realistic adobe-firefly workflow is being prepared for Use Adobe Firefly for Text to Image Generation using adobe-firefly. The starting requirement is: Replace skyline haze [brush distant city] Softer atmospheric haze, keep building silhouettes, naturalistic color. The work should be tested on a representative input first so the result can be reviewed before the same process is applied to the full project or production workload.

Objective:
Create a production-ready result for "Replace skyline haze [brush distant city] Softer atmospheric haze, keep building silhouettes, naturalistic color" that directly supports the article use case. The output should be specific, reviewable, and reproducible rather than a generic demonstration. Keep the requested goal as the primary outcome and avoid adding unrelated features, assumptions, or unsupported capabilities.

Expected output:
The final result should directly satisfy the "Replace skyline haze [brush distant city] Softer atmospheric haze, keep building silhouettes, naturalistic color" requirement and be understandable without guessing what was supplied or configured. A successful run should produce the intended output in the requested format, with the important values and workflow decisions clear enough to reproduce the result. If the workflow can fail, the expected behavior should also make the failure visible and provide a clear next action instead of silently producing an incomplete result.

Tips

Paint smaller than you think, then expand the mask if needed
Describe the fill object, not the whole photograph
Add a reference under References when style drifts
Regenerate with a shorter prompt if results invent extra props
Confirm credit behavior for your model and plan
Finish critical edges in Photoshop when brand risk is high

Verification notes

Official Generative Fill help documents brush selection, Select less or Select more, References, and credit indicators by model. Treat plan unlimited claims as current Adobe plan text only. When credits, plan features, or model access differ from memory, open Adobe Firefly plans and the generative credits FAQ. Limits change, so do not promise client deliverables from outdated screenshots.

Keep a short creation log beside your Firefly history: prompt, aspect ratio, reference used, and the one field you will change next. That habit makes four option comparisons useful instead of random.

When credits, Free daily limits, or premium feature access look different than yesterday, open Adobe Firefly plans and your account generative credit indicator. Plan details change, so do not rely on memory for client promises.

Treat the first Generate as a sketch. Review all four options before rewriting the whole prompt. Small descriptor edits teach more than full rewrites.

If you collaborate, store winning prompts and reference files on a shared brand page. Reuse the skeleton and change subject or channel crop only.

For social clips and paid placements, decide aspect ratio before you generate. Asking for a new crop after the fact wastes time and may cost credits on premium paths.

Separate exploration from release prep. Explore freely, then re check Adobe commercial use wording, beta labels, and Content Credentials needs before monetizing.

Close each session by naming keepers and archiving rejects. A clean library saves time when a stakeholder asks for the mug shot from Tuesday.

Prefer clear visual language over vague vibes alone. Subject, light, materials, and one mood word outperform a pile of adjectives.

Move brand critical finishing into Photoshop or Illustrator when edges, type, or logo placement must be exact. Firefly accelerates exploration. Design tools finish production.

Confirm whether your current feature shows a beta label before you claim commercial clearance. Adobe commercial wording applies to features without the beta label per product materials.

Brush hardness matters as much as prompt wording. Soft brushes blend fabrics and skies. Firmer brushes protect product silhouettes and type areas you must not touch.

When Fill invents extra props, shorten the prompt to the single object and continuity line. Long scene essays encourage the model to redesign the room.

Reference images under Fill References help when brand materials must match a known finish. Keep the text prompt short so the reference can lead style.

Think of Generative Fill as a local rewrite, not a second full render. The surrounding pixels teach lighting and perspective. Your job is to give the model a clean mask and a short object brief so it can borrow that context honestly.

For replace jobs, name the outgoing object class in your notes even if the prompt only describes the incoming object. Knowing you swapped a chair for a stool helps reviewers compare continuity of scale and shadow length.

Fabric, glass, and chrome need continuity words. Ask for matching specular highlights or matching cloth folds. Without those cues, Fill can drop a correct object into incorrect light.

When brand packaging must stay untouched, mask away from labels entirely. Do not ask Fill to recreate logo lettering. Keep packaging plates locked and change only props, surfaces, or backgrounds around them.

Save before and after pairs for your brand library. Fill success is easier to teach with a mask screenshot and the short prompt than with a finished JPEG alone.

Common mistakes

Painting half the subject you meant to keep
Describing the entire scene instead of the fill region
Skipping Select less or Select more before generate
Stacking five new objects in one fill pass
Ignoring lighting mismatch at the mask edge
Assuming unlimited Fill without checking your plan

Related Adobe Firefly articles: /blog/how-to-use-adobe-firefly-generative-remove, /blog/how-to-use-adobe-firefly-for-text-to-image-generation, and /blog/how-to-use-adobe-firefly-generative-expand. Return to /explore/firefly for the broader guide set.