# Advanced Inpainting: Adding New Elements Interactively A common and powerful use case for Fater's inpainting capabilities is adding entirely new elements (like people, objects, or features) into an existing scene, making them appear as if they naturally belong and interact with their surroundings. This requires a different masking and prompting strategy than simply modifying or removing an existing object. This guide explains the correct workflow for adding a new element that needs to interact spatially with existing elements in your image, such as placing a person on a sofa. ### The Common Pitfall: Masking the Existing Object When trying to add a new element (e.g., "a woman") onto an existing object (e.g., "a sofa"), a common incorrect instinct is: 1. Mask the *entire existing object* (the sofa) using an AI segmentation tool with "the sofa" as the prompt. 2. Prompt the AI with "a woman on a sofa" to fill that sofa-shaped mask. **Why this approach is flawed:** * **Spatial Constraint:** The AI will try to fit the *entire concept* of "a woman on a sofa" strictly *within the 2D boundaries of the original sofa's mask*. A person sitting naturally will have parts of their body (head, shoulders, legs) extend *beyond* the 2D footprint of the sofa. The mask becomes an overly restrictive container. * **Loss of Context for the New Element:** The AI doesn't see the original sofa pixels as direct context *for the woman*; it only sees the area *around* the sofa mask. It's forced to reinvent the sofa *and* place the woman within that reinvented sofa's limited masked space. * **Unnatural Results:** This often leads to the new element appearing squashed, unnaturally contorted to fit the mask, or disproportionate. ### The Correct Workflow for Adding New Elements To successfully add a new element that interacts with an existing one, follow these steps: **1. Define the Space for the *****New***** Element (Masking)** * **What to Mask:** Your mask should define the area where the *new element itself* will appear and occupy space. For a "woman on a sofa," this mask should cover the silhouette of the woman, including where her body overlaps the sofa and where parts like her head or legs would extend into the surrounding space. * **The Existing Object Remains Unmasked:** Crucially, the existing object (the sofa) should *not* be part of this primary mask. It should remain visible *around and underneath* the area you're masking for the new element. * **Tools for Masking:** * **AI Suggestion:** Use tools like "Prompt to Mask" if you want the AI to help identify a general area for the new element, but be prepared to refine it. For instance, you might ask the Agent to suggest a region. * **Manual Masking:** Use the **Mask Brush**, **Polygonal Lasso**, or **Mask Rectangle** tools on the Inpaint Toolbar to draw the mask for the new element. * **Important:** The mask should be for the *woman*, not the sofa she sits on. *(Imagine a diagram here: A sofa is visible. A semi-transparent mask is drawn over part of the sofa and extends above and slightly to the sides, representing the volume a sitting person would occupy.)* **2. Ensure Sufficient Context (Generation Area & Margin)** * Once your mask for the new element is defined: * Use the **"Resize Canvas to Masked Area"** action (Crop icon on the Inpaint Toolbar or its equivalent). * **Crucially, when using this action, ensure a generous margin is included around the mask.** Many AI models require a margin factor of `1.5` to `2.0` (meaning the Generation Area is 1.5 to 2 times wider and taller than the mask itself, centered on the mask). * **Why the Margin is Vital:** This expands the Generation Area Bounding Box significantly beyond the immediate mask. It ensures the AI "sees" enough of the surrounding environment (the sofa, the room, lighting conditions) to: * Understand the perspective and scale. * Generate realistic lighting and shadows on the new element, consistent with the scene. * Blend the new element seamlessly with the existing objects. * Without an adequate margin, the inpainted element might have harsh edges or inconsistent lighting. **3. Select an Appropriate Inpainting AI Model** * Choose an AI model from the **"Edit"** category in the Model Selector (Left Sidebar). * Models like **`Ultra [Edit/Basic]`**, **`High [Edit/Basic]`**, or **`Lite [Edit/Basic]`** are generally suitable for this task. Refer to the AI Model Directory for details on each model. **4. Craft Your Inpainting Prompt** * Your text prompt (in the Floating Prompt Area) should describe the **new element you want to appear within the masked area.** * **Example:** If your mask defines the space for the woman, your prompt should be something like: * "A woman reading a book, sitting comfortably." * "A person relaxing on the sofa, looking towards the window, photorealistic." * The AI will use the unmasked sofa (visible as context) to make the generated person appear to be interacting with it. **5. Configure Key Model Parameters** * **`Destroy Fill Area Behind Mask` (or equivalent "Denoising Strength"):** * When adding a completely new element, you typically want this set to **`Yes` (or a strength of `1.0`)**. This tells the AI to fully replace any existing pixels under the mask with the newly generated content. * Adjust other parameters like `Steps`, `Guidance`, `Seed`, `Style Type` as needed. Check the Parameter Glossary for more information. **6. Initiate Generation and Iterate** * Click the **Generate** button (✨) in the Floating Prompt Area. * Review the results in the Generation Task List. * If the first attempt isn't perfect: * Refine your mask. * Adjust your prompt. * Tweak model parameters (especially `Seed` if you liked the composition but want to change details). * Ensure your margin for the Generation Area was sufficient. #### Key Takeaways * **Mask the *****New***** Element's Space:** Define the volume the new object will occupy. * **Context is King:** The existing interacting object (e.g., sofa) must be visible and unmasked *around* your new element's mask. * **Use "Resize Canvas to Masked Area" with a significant margin (equivalent to a factor of 1.5-2.0).** This is non-negotiable for good results. * **Prompt for the *****New***** Element:** Describe what you want to see *within* the mask. * **Use Inpainting Models:** Models from the "Edit" category are designed for this. By following this workflow, you can leverage Fater's AI to add new elements into your scenes with greater realism and believable interaction. --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://fater.gitbook.io/fater-ai-docs/generating-images-with-ai/advanced-inpainting-adding-new-elements-interactively.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.