# AI Generative Fill — Replace & Add Objects in Photos ZEnhancer’s AI generative fill is an online image-editing tool that replaces or adds elements inside a painted mask using descriptive text prompts. Upload a JPG, PNG, or WebP photo, paint the area you want to change, describe what to add or preserve, and download your edited result. ## How to Use AI Generative Fill Online Create a generative fill edit in three steps, from upload through export. ### Upload an Image and Mark the Area Upload one clear image, then check orientation. Mark an image area using brush controls; adjust size for your intended object and its contact shadow. If you choose another source image, draw a new selection so each marked area matches it. If an image exceeds a limit, resize a copy for upload. Keep your source image for later detail checks. Check both image limits: file bytes measure upload size; pixel count measures image dimensions. ### Describe What Should Replace the Selection Type precise descriptive wording into prompt fields. Name replacement items, intended textures, colors, contextual illumination. Specify background elements that must stay stable alongside new objects. Clear text instructions help our generative fill tool render coherent details aligned with camera perspective. Avoid ambiguous adjectives; focus on tangible object attributes. ### Generate, Compare and Download Check your quote and eligibility before generating. Compare each generative fill result with its source image. Inspect edges, scale and lighting. Check unselected areas too. Export the recorded product image as PNG or JPG. ## AI Generative Fill Tool Controls and Costs Use AI generative fill tool controls to refine an image before another edit. ### Edit the Selection with Simple Brush Controls Refine each image selection with this AI generative fill tool before submitting. - Mark an image area precisely: Paint adds it; Erase removes it. - Reverse brush mistakes: Undo steps back; Redo restores a selection change. ### See the Cost Before Starting an Edit Sign in with a verified email to use GPT Image 2. The AI generative fill tool shows the current credit cost and checks your balance before you submit. If a connection drops, check the current edit before starting another. - Check price before submitting: Quote shows cost; Eligibility explains access. - Keep an edit traceable: Status shows progress; Recovery resumes checking. ### Keep the Source and Review the Result A 30-minute local draft stores your source image with its painted selection and prompt. Server retention remains unverified. - Compare image changes: Original retains source view; Compare reveals differences. - Save an image copy: PNG/JPG export reuses the completed or recorded result; downloading does not start another generation. ## Generative Fill Results and Quality Checks Inspect your original image alongside each image result. A generative fill edit needs checks in both selected and unselected areas. ### Review the Original, Selection and Result Review five details in each generative fill result against its source image. A selection is guidance. Any image detail may change; check a full image view before saving. - Request: Does the object match your description? - Edges: Look for halos, smears or abrupt seams. - Scale: Check size, perspective and surface contact. - Light: Review shadow direction, brightness and material reflections. - Context: Inspect faces, lettering and background lines beyond your selection. ### Check Edges, Scale and Background Changes Inspect image edges for halos, smears or abrupt changes in focus. Compare scale against nearby objects. Match display sizes before judging a resized image, then zoom in to inspect its edges. Recorded product image: a complete pedestal is visible, with a faint tonal seam. Each practice example includes a source image with a painted selection. Both results are pending verification. ## Image Generative Fill for Everyday Edits Use generative fill within an existing photo. Compare a recorded product image or practice with an outdoor or room image. The two practice results are pending verification. ### Replace a Product Prop in an Image Start with the recorded product example: a bottle beside a wooden display block. Its loaded prompt asks for a low round white marble pedestal, matching the soft left light and contact shadow while keeping the amber bottle, blank label, tabletop and background unchanged. The recorded result is a reference, not a new generation. Check label lettering, glass reflections, pedestal scale and the join with the tabletop; an unselected bottle or background detail may still change. ### Add a Detail to an Existing Photo Choose an outdoor image with an empty corner on a picnic table. Mark a small area that leaves room for the new object and its shadow. Try: “Add one blue ceramic mug on the table, with a handle facing right and soft morning light.” Review the generative fill result at full view, then inspect the mug closely. Does its size fit nearby objects? Does the shadow touch the table? Compare surrounding wood grain and background detail with your source image before saving an edited copy. ### Refresh One Part of a Room Scene Use a room image with an open floor area beside a window. Select space for a plant and pot, including where the pot would meet the floor. Try: “Add a small rubber plant in a terracotta pot, with daylight from the window.” After generative fill, check leaf size, pot perspective and the contact shadow. Compare floorboards, window edges and furniture with your original image. If the addition looks too large, revise your prompt or selection before another edit; the first result is a candidate to inspect. ## Write Better Prompts and Refine Your Selection For an online AI generative fill tool, a useful prompt names the replacement and the details you want to keep. Selection size supplies another part of the instruction. ### Name the Replacement and Its Key Details Name an object, color, material and position for your image. For example: “a small blue ceramic vase on a table, lit from the left.” Include details you want to retain. An AI generative fill prompt expresses your intention. Compare the output with your source image to see what actually changed. ### Fix Oversized Masks and Unclear Boundaries Leave some surrounding image context visible. If an edge looks wrong, adjust your image selection to include a relevant shadow, then try a revised prompt. Wider masks can change more of a scene; narrower ones can leave part of an unwanted object. Judge each image result rather than assuming a mask guarantees a clean edge. ## Compare Online AI Generative Fill Tool Options Compare an online AI generative fill tool by the work you can complete, access conditions and export limits. These observations were recorded on November 17, 2026. ### Compare Workflow, Access and Download | Platform | Brush & Prompt Workflow | Access & Account Gate | Export Options | | --- | --- | --- | --- | | ZEnhancer | Paint the area to edit, write a prompt and generate with GPT Image 2, or open the recorded product example. | Verified email and sufficient credits required for a new edit. The tool shows the quote before submission. | JPG and PNG export of the recorded 1536×1024 result. | | Pixelcut | Brush + prompt tested | Preview available in this test; HD login prompt | JPG downloaded; HD not downloaded | | Pixlr | Selection + prompt tested | Account prompt before generation in this test | Unknown: not tested | | EzEnhancer | Brush + prompt; one result tested | 4 trial credits used in this test | PNG downloaded | ### Read File Limits and Test Conditions Pixelcut and EzEnhancer each yielded one downloaded file. Pixlr stopped at a sign-in prompt. No shared benchmark measured comparative speed or quality. Internal model details remain unverified. | Platform | Input used or proposed limit | Output observed | Cost in this test | Speed | | --- | --- | --- | --- | --- | | ZEnhancer | Recorded 1536×1024 example; public upload up to 10 MiB / 40 MP | JPG and PNG export of the recorded 1536×1024 output | Opening the recorded example uses no credits | Not measured | | Pixelcut | Built-in product image | 720×1080 JPEG (733,271 bytes); HD 1280×1920 sign-in | Free preview observed; HD requires login | Not measured | | Pixlr | 800×600 PNG | Unknown (untested due to gate) | Not tested | Not measured | | EzEnhancer | 800×600 PNG | 1408×1056 PNG (1,015,496 bytes) | 4 credits; 10→6 | Not measured | ## Generative Fill Questions and Answers These generative fill answers cover input, access, result checks and recovery. ### What Is Generative Fill and How Does Inpainting Work? Generative fill uses an image selection and a text description to guide added or replacement content. This is a form of inpainting. The selection identifies your intended image change; it does not guarantee that surrounding image details stay identical. ### How Do I Start and Which Image Formats Can I Upload? Generative fill accepts one JPG, PNG or WebP image up to 10 MiB and 40 megapixels. Before generation, source image and selection share a working grid capped at 2048 px on its longest edge. ### What Does Generative Fill Cost and Is an Account Required? Generative fill uses GPT Image 2. Sign in with a verified email and check the credit quote in the editor before each edit. You can open, compare and download the recorded example without using credits. ### How Should I Paint the Selection and Write the Prompt? For an AI generative fill edit, cover the object and relevant contact shadow without selecting unrelated details. Name your requested replacement, material, color and position. Instructions to retain a face or label describe your intention; inspect each image result to see whether it was followed. ### Can Details Outside the Selection Change? Yes. Generative fill can change details outside the selected area, not just pixels touching its edge. Compare your full source image with each image result, especially faces, text and geometric lines. Use a more precise editing method when exact preservation is essential. ### What Happens If an Edit Fails or I Refresh the Page? After an interruption, check your current generative fill request before starting another. Refresh can restore a local draft within 30 minutes. Server retention is unverified. Check whether credits were returned after a failed request. ### How Can I Compare and Download the Edited Image? Compare each generative fill result with its source image. Download that recorded product image as PNG or JPG at 1536×1024. Retrying a download reuses your result. ## Product facts and verification status - Generative Fill uses GPT Image 2 with a native selection mask. - Original input: one JPG, PNG or WebP photo up to 10 MiB and 40 megapixels. The editor normalizes the image and selection to a shared working grid up to 2048 pixels on the longest side. - The painted selection becomes a PNG mask with transparent selected pixels. The mask and normalized source must have identical pixel dimensions. - Results are digital images at the actual returned dimensions and format. Inspect the full image before saving. Source resolution and unchanged pixels outside the mask are not guaranteed. - One edit with GPT Image 2 costs 4 credits. Sign in with a verified email and review the quote before submitting. - An accepted request uses a separate recovery checkpoint. Refresh recovery checks the existing request status and does not submit a second generation. - Example input photos are AI-created test inputs. A result is identified as a real native-mask output only when its source, mask, prompt and provider run are recorded together. - The mask guides the edit and does not guarantee exact pixel boundaries. Inspect objects, lighting, edges and unselected details in the full result. ## Still awaiting evidence - The exact server retention period for inputs and results remains unverified. The local draft window and recovery checkpoint do not establish a server deletion deadline. - Provider consistency, comparative quality and speed, exact server retention and output watermark behavior remain unverified unless a specific recorded test states otherwise. ## Scope and limits Use this tool to add or replace details inside the existing image frame. Canvas expansion, background removal and clarity enhancement are separate tasks. ## Reference documents - llms.txt: https://enhancerfree.com/generative-fill/llms.txt - llms-full.txt: https://enhancerfree.com/generative-fill/llms-full.txt - index.md: https://enhancerfree.com/generative-fill/index.md - tool guide: https://enhancerfree.com/generative-fill.md - pricing: https://enhancerfree.com/generative-fill/pricing.md - policies: https://enhancerfree.com/generative-fill/policies.md