Image Inpainting
How to use Image Inpainting in Qolaba — masking areas with the brush tool, writing replacement prompts, and getting clean, consistent inpainting results
Image Inpainting lets you selectively replace or modify a specific area of an existing image — using a brush to mask the region and a text prompt to define what replaces it. Everything outside the masked area remains exactly as it was.
What Inpainting Does
When you mask an area and enter a prompt, the model replaces that region with newly generated content that matches the context and style of the surrounding image. This allows precise, targeted edits without touching any other part of the image.
Common use cases:
Removing unwanted objects or people from a scene
Replacing a product background in a specific zone
Changing clothing, accessories, or color on a subject
Fixing generation artifacts or anatomical errors
Adding new elements to an existing composition
Swapping text or signage within an image
How to Use Inpainting
Step 1 — Open Image Editing → Inpainting from the workspace.
Step 2 — Upload the image you want to edit, or select one from your generation history.
Step 3 — Adjust the brush size to match the area you need to mask. Use a larger brush for broad areas and a smaller brush for precise detail work.
Step 4 — Paint over the area you want to replace. The masked region is highlighted — review it to make sure only the intended area is covered.
Step 5 — Enter a prompt in the prompt field describing what should replace the masked area.
Step 6 — Click Generate. The model fills the masked region with new content based on your prompt while preserving everything outside it.
Brush Controls
Brush Size
Increase or decrease the brush diameter for broad or precise masking
Undo
Step back through masking actions to correct mistakes
Redo
Reapply a previously undone masking action
View Original
Toggle between the original image and the edited output for comparison
Writing Effective Inpainting Prompts
Inpainting prompts work differently from generation prompts — they need to describe the replacement content in relation to the surrounding image context.
Be specific about the replacement:
Reference the surrounding context when relevant:
For removal without replacement:
Tip: The more specific your replacement prompt, the more accurately the model fills the masked area. Vague prompts like "something better" or "fix this" produce inconsistent results. Describe exactly what you want to appear.
Tips for Clean Inpainting Results
Mask precisely — over-masking areas you want to keep causes the model to alter them. Use a small brush for edges and detailed boundaries
Use a slightly larger mask than needed — masking just inside the boundary of an object sometimes leaves visible seams. Extend the mask slightly beyond the object edge for cleaner blending
Match context in your prompt — if the surrounding image has a specific lighting style or texture, reference it in your prompt so the replacement blends naturally
Use Undo to refine the mask — take time to get the mask right before generating. A clean mask produces a dramatically better result than a rushed one
Compare with View Original — always compare the output against the original before downloading to verify the edit looks natural
Last updated