How to Use Inpainting in AI Image Tools

Master AI inpainting to edit, replace, and enhance specific parts of images. Step-by-step guide for DALL-E, Midjourney, and Stable Diffusion tools.

1. Select your inpainting tool. Choose an AI platform that supports inpainting. DALL-E 2 offers the simplest interface through ChatGPT Plus or OpenAI's website. Midjourney requires the --vary (region) parameter. Stable Diffusion works through interfaces like Automatic1111 or ComfyUI with dedicated inpainting models.
2. Upload and prepare your base image. Import your source image to the platform. Ensure the image meets size requirements—DALL-E accepts up to 4MB, Midjourney needs images uploaded to Discord, and Stable Diffusion works best with 512x512 or 1024x1024 pixel images. The image should have clear, well-defined areas you want to modify.
3. Create your selection mask. Mark the area you want to modify. In DALL-E, use the eraser tool to brush over the region. Midjourney requires uploading a separate black and white mask image where white areas indicate what to change. In Stable Diffusion, paint directly on the image or upload a mask file where white pixels mark the inpainting area.
4. Write your inpainting prompt. Describe what should appear in the masked area. Be specific about objects, colors, lighting, and style. For DALL-E, type directly into the prompt box. Midjourney users add their prompt after the image and mask URLs. Stable Diffusion accepts prompts in the positive prompt field while the negative prompt excludes unwanted elements.
5. Adjust inpainting settings. Configure generation parameters for optimal results. In DALL-E, choose between 'replace' and 'add' modes depending on whether you're substituting or adding elements. Stable Diffusion users should set denoising strength between 0.3-0.8, with lower values preserving more of the original image. Increase CFG scale to 7-12 for better prompt adherence.
6. Generate and refine results. Run the inpainting process and evaluate the output. Most tools generate multiple variations—select the best result or generate additional options if needed. Look for seamless blending between the inpainted area and original image, proper lighting consistency, and accurate representation of your prompt.
7. Post-process for final quality. Download your result and make final adjustments if necessary. Use traditional photo editing software like Photoshop or GIMP to blend edges, adjust colors, or fix small inconsistencies. Many AI tools compress outputs, so consider upscaling the final image using AI upscalers like Real-ESRGAN or Topaz Gigapixel.