Outpainting
Outpainting generates new, coherent image content beyond the original canvas edges, extending a picture outward.
Outpainting (also "uncropping") extends an image past its original borders, letting a diffusion model invent plausible surroundings that match the existing lighting, perspective, and style. The model treats the new area as a masked region conditioned on the visible pixels plus your prompt, then fills it so seams disappear. It is the complement of inpainting, which edits inside the frame. Outpainting is used to change aspect ratio (turn a square into a 16:9 banner), reveal more of a scene, or rebuild a tight crop into a full composition. Best results come from extending in modest steps rather than one huge jump.
When to use outpainting
- Reframing a portrait or product shot to a wider or taller aspect ratio.
- Adding headroom or background for text overlays and ad placements.
- Recovering a scene that was cropped too tightly.
When not to use outpainting
- When you need the new area to match a real, specific place — the model invents, it does not recall.
- Extremely large extensions in one pass; expand in stages to keep coherence.
Example
Input: Square 1:1 product photo + outpaint left/right to 16:9, prompt: "same studio backdrop, soft gradient" Output: A widescreen banner where the backdrop continues seamlessly on both sides of the product.
Common mistakes
- Outpainting a huge area at once, producing repeated or warped content.
- Leaving the prompt empty — describe what should be in the new region.
- Forgetting to blend the seam; a small overlap mask hides the join.
FAQ
What is outpainting?
Outpainting generates new, coherent image content beyond the original canvas edges, extending a picture outward.
When should I use outpainting?
Reframing a portrait or product shot to a wider or taller aspect ratio. Adding headroom or background for text overlays and ad placements. Recovering a scene that was cropped too tightly.
What are the most common mistakes with outpainting?
Outpainting a huge area at once, producing repeated or warped content. Leaving the prompt empty — describe what should be in the new region. Forgetting to blend the seam; a small overlap mask hides the join.
Related terms
- Denoising strength — Denoising strength controls how much an image-to-image run reshapes the source image, from 0 (untouched) to 1 (source ignored).
- Seed — A seed is an integer that initializes the random number generator inside an image, video, or audio model, making generation reproducible.
- CFG scale (classifier-free guidance) — CFG scale controls how strongly a diffusion image model follows its text prompt — higher values stick closer to the prompt, lower values explore more.
- Negative prompt — A negative prompt is text that tells an image, video, or audio generator what to avoid producing — the opposite of the main prompt.
- Diffusion model — A diffusion model is a generative neural network that creates images, video, or audio by iteratively denoising random noise toward a learned target distribution.
Sources
Last updated: 2026-06-02. Raw markdown: https://promtable.com/glossary/outpainting.md.