You have a great image. But the framing is too tight. The subject is cropped. Or you need a different aspect ratio for a new platform.
Cropping makes it smaller. You need it bigger with the scene extended naturally beyond the original edges.
Your image in. Extended composition out, seamlessly blended.
Run it now on Floyo!
Why Flux Fill Dev for Outpainting
Most outpainting tools generate obvious seams. The new content looks pasted on. Colors shift. Lighting doesn't match. The blend gives it away.
Flux Fill Dev uses DifferentialDiffusion, a node that blends the new generation into the existing image at a gradient rather than a hard edge. The extended content transitions smoothly into the original. Style, lighting, and detail carry across naturally.
seamless blending with no visible seams or edge artifacts
style, lighting, and color tone consistent across the extension
extend in any direction, left, right, top, bottom, or all sides
change aspect ratio without cropping or letterboxing
generates in about 28 seconds
How It Works
Upload your image. Set how many pixels to extend on each side using the ImagePadForOutpaint node. Write a prompt describing what the extended area should contain. Flux Fill Dev generates the new content and blends it seamlessly into the original using the DifferentialDiffusion approach.
The MaskPreview node lets you check the mask before generation runs, useful for confirming exactly which areas will be generated.
Key Inputs
Your Image
Any JPG or PNG. The original content stays untouched. Only the extended regions are generated.
Works well with:
portraits where you want to reveal more of the scene or environment
landscape and architectural images being reformatted for a new aspect ratio
product shots needing a wider background
illustrations and concept art being extended for a wider field of view
any image where the existing content has a readable style and lighting the model can continue
Prompt
Describe what should appear in the extended area. Match the existing scene context for seamless results.
Examples:
"continuation of the forest background, soft natural light, green foliage, same depth of field""clean white studio background, consistent soft lighting, no new objects""urban street scene continuing from existing environment, same lighting conditions, photorealistic""open sky and horizon, same golden hour lighting, wide landscape""continuation of marble surface, same texture and lighting, minimalist"
The more the prompt matches what's already in the image, the cleaner the blend.
Extension Amount (per side)
Set in pixels via the ImagePadForOutpaint node. Extend one side, multiple sides, or all four depending on the output format you need.
Common use cases:
Square to landscape: extend left and right
Portrait to square: extend top and bottom
Tight portrait crop: extend all sides to reveal more scene
Horizontal to vertical: extend top and bottom significantly
Feathering
Controls how gradually the mask blends into the existing image. Higher feathering = softer transition. Lower feathering = harder edge. Default works for most images, adjust if you're seeing the blend line.
What This Is Great For
Aspect ratio conversion: Change a square image to landscape, a portrait to square, or any format to another without cropping. The workflow extends the canvas rather than cutting it.
Portrait and subject reveals: A tightly cropped subject gets a full environment. Reveal the room, the landscape, or the scene context around the original crop.
Product photography: Extend a tight product shot to a wider background for a more spacious, editorial look. Consistent background style across the full extended canvas.
Social media reformatting: Take a hero image designed for one platform and extend it to fit another without rebuilding the composition.
Creative expansion: Extend illustrations, concept art, and generated images to reveal more of the world beyond the original frame.
What to Watch Out For
The prompt must match the existing scene. Describing content that conflicts with the original image's lighting, color, or style produces obvious inconsistencies at the blend point. Describe a continuation, not a new scene.
Very large extensions relative to the original image size give the model more area to invent with less visual context to anchor to. Extend in increments on complex images, two smaller passes produce cleaner results than one very large extension.
Highly detailed or textured backgrounds (dense foliage, complex patterns, detailed architecture) are harder to continue seamlessly than simple or gradual backgrounds. Clean studio backgrounds and open environments extend most reliably.
The MaskPreview node shows the mask before generation. Use it to confirm the extension area looks right before committing to the full run.