Outpainting is the inverse of cropping: instead of trimming the image, you expand the canvas and let an AI model invent what would have been there. Used for fixing awkward crops in old photos, turning vertical phone shots into desktop wallpapers, or extending a square AI generation into a 16:9 banner. The terms extender, expander, and outpainting all refer to the same workflow.
When to use it
Use an extender when an image needs more space on one or more sides. Common cases: a portrait shot needs to become a landscape banner; a phone screenshot needs to fit a desktop layout; an old photo got cropped awkwardly. All major AI image platforms now support some form of outpainting — quality depends on the model and how much you're trying to extend.
How to use it (step by step)
- Pick the right tool for your image. For AI-generated images: Midjourney Zoom Out (uses the same model that made the image — best stylistic continuity). For real photos: Photoshop Generative Expand (Adobe Firefly, requires Photoshop CC subscription but the feature itself is free with the subscription). For full control: Stable Diffusion outpainting in ComfyUI.
- Extend in modest increments. Trying to triple the canvas in one step usually produces obvious seams or repeating patterns. Extend by 20-50% at a time, then accept the result before extending further. Each pass gives the model new context to build on.
- Match the lighting and perspective. The trickiest part of outpainting is keeping shadows and vanishing points consistent. If the original has strong directional light from the left, your prompt should include 'continuing the warm sunlight from the left edge' to nudge the model. Without this hint, models often invert lighting unintentionally.
- Clean up seams. Even the best outpainters sometimes leave a faint seam where the new content meets the old. Run a quick inpainting pass over that area to blend it. Photoshop's Healing Brush works for subtle cases.
Common mistakes to avoid
- Outpainting low-resolution sources — the extender will produce sharper detail in the new area than in the original, creating an obvious quality gradient. Upscale first.
- Forgetting that outpainted regions don't necessarily contain real objects (they're hallucinated). Don't use outpainted areas in documentary photography or legal contexts.
- Asking for too much new content at once — 'extend the sky and add a flock of birds and put a mountain on the right' overwhelms most models. Extend first, then add new objects with separate inpainting calls.
- Confusing extending with upscaling. Extending makes the canvas LARGER (more area). Upscaling makes the canvas HIGHER RESOLUTION (more pixels per inch). They solve different problems.
Frequently asked questions
What's the difference between AI image extender and AI image expander?
Nothing — they're synonyms. Both refer to outpainting: making an image's canvas larger by having AI invent new pixels at the edges. Some tools use one name, some use the other.
Is there a free AI image extender online?
Yes. Midjourney's Zoom Out is free during your trial credits. Photoshop Generative Expand is included with any Photoshop subscription (free trial available). ClipDrop offers a free outpainting tool with a daily limit. Stable Diffusion is completely free if you run it locally or use a free hosted UI.
Can I extend a real photograph?
Yes. Adobe Firefly (which powers Photoshop's Generative Expand) is specifically trained on commercial-license imagery and handles photographs well. Nano Banana 2 also extends photos cleanly via conversational instructions.
How much can AI extend an image?
Most models work best when extending by 20-50% per pass. Going beyond 100% (doubling the canvas) in a single step often produces visible seams or implausible content. The trick is iterative extension.
Does extending change the original image?
No. The original pixels stay exactly as they were. The new region is generated and stitched alongside. You can always crop back to the original if you don't like the result.