Enlarge a small portrait.
Add plausible hair, fabric, and skin texture while using low creativity to reduce changes to facial identity.960x540 source, upscaled 1x to 960x536
Magic Image Refiner
Control how much new texture it invents while keeping the result close to your source.
From this.
To this.
960x540 source, upscaled 1x to 960x536
Before and after, on real images
Improve a property exterior.
Reconstruct brick, foliage, and roof texture in a small listing photo while preserving the overall building layout.960x540 source, upscaled 1x to 960x536
Upscale AI artwork.
Build richer foliage, clothing, and painted textures into a low-resolution image without changing its main composition.960x540 source, upscaled 1x to 960x536
Clean up a product shot.
Improve edges and surface texture in a compressed supplier image while keeping creativity low to limit product changes.960x540 source, upscaled 1x to 960x536
Refine an old family scan.
Reduce distracting scan defects and reconstruct clothing detail cautiously; the result is not an archival or evidentiary record.960x540 source, upscaled 1x to 960x536
Balance new detail with resemblance
Magic Image Refiner is diffusion-based, so it synthesizes plausible texture rather than recovering missing pixels. Creativity controls the strength of reconstruction, while resemblance helps constrain the result to the source.
- •Creativity range: 0 to 1
- •Default creativity: 0.25
- •Resemblance range: 0 to 1
- •Default resemblance: 0.75
- •Lower creativity is usually more conservative
Target a 2048px short edge
Choose original-size processing or a 1024px or 2048px short-edge target. The long edge follows the image shape with internal rounding, so this is not a fixed 2x or 4x enlargement.
- •Up to a 2048px short-edge target
- •Original-size refinement available
- •Maximum input: 1024x1024
- •Wide and square images supported
- •Download as PNG or JPG
Reconstruct texture and rough edges
Tiled conditioning helps the model add coherent local detail across fabric, foliage, masonry, and digital artwork. Compression blocks and noise can still be reinterpreted as new texture, especially at higher creativity.
- •Adds plausible fine texture
- •Works across large image regions
- •Can soften visible compression artifacts
- •Useful for foliage, fabric and masonry
- •High creativity may introduce artifacts
Use conservative settings for faces and text
There is no separate face-restoration or text-preservation control. Low creativity and high resemblance can limit drift, but names, logos, facial features, and small lettering must be checked carefully.
- •Inspect eyes, teeth and skin at 100%
- •Check every letter after processing
- •Keep creativity low for product logos
- •Higher resemblance can reduce visual drift
- •Not suitable for forensic interpretation
How it works
1
Upload your image
Choose the low-resolution or damaged image you want to refine. The uploaded image must fit within the 1024x1024 maximum input size.
2
Choose size and settings
Pick original, 1024px, or 2048px short-edge processing. Adjust creativity, resemblance, and HDR, starting near the conservative defaults when faces, text, or products must remain recognizable.
3
Inspect and download
Compare the result with the source at full size, paying close attention to faces, text, logos, and repeated patterns. Download the accepted result as a PNG or JPG.
Pricing for Magic Image Refiner
Runs on credits — no per-model surcharges, no surprise billing.
40credits
per generation
40 credits per image
Frequently asked questions
What is Magic Image Refiner?▾
Magic Image Refiner is a diffusion-based image enhancement model designed to add detail and enlarge existing images. It is suited to low-resolution photos, digital artwork, textured scenes, and compressed images where plausible reconstruction is acceptable.
How does Magic Image Refiner work?▾
The model resizes the source, divides its visual information into manageable regions, and uses tiled conditioning to guide a diffusion-based refinement pass. Creativity controls how strongly it changes the image, while resemblance encourages it to follow the original structure.
Does Magic Image Refiner invent detail that was not there?▾
Yes. It reconstructs plausible detail based on patterns learned by the model; it cannot recover information that the camera or original file never captured. This can improve fabric, foliage, hair, and masonry, but invented changes may be unacceptable in faces, documents, logos, evidence, or exact product records.
How much can Magic Image Refiner enlarge an image?▾
You can process at the original size or target a 1024px or 2048px short edge. The long edge is calculated from the image shape with internal rounding, so the model does not use a single fixed 2x or 4x scale for every file.
What is the maximum input size for Magic Image Refiner?▾
The maximum accepted input is 1024x1024 pixels. Wide images such as 1024x576 fit within that limit, but neither input dimension can exceed 1024 pixels.
Does Magic Image Refiner work on faces and keep text readable?▾
It can make faces appear sharper, but it may alter eyes, teeth, skin texture, or identity-defining features. Small text and logos can also become misspelled or reshaped. Use low creativity, higher resemblance, and inspect the result closely before publishing.
What output formats does Magic Image Refiner support?▾
Finished images can be downloaded as PNG or JPG. PNG is useful when you want to avoid additional JPEG compression, while JPG produces a smaller file.
Can I upscale AI-generated images with Magic Image Refiner?▾
Yes. AI-generated artwork is a strong use case because the model can add plausible texture to soft foliage, clothing, architecture, and painted surfaces. Check hands, faces, lettering, and repeated patterns because refinement can amplify or reinterpret existing generation errors.
How much does Magic Image Refiner cost?▾
Each refinement uses credits, with usage affected by the selected output size and processing requirements. Review the displayed credit estimate before starting the job.
How does Magic Image Refiner compare with other AI upscalers?▾
Magic Image Refiner emphasizes controllable, generative detail reconstruction rather than strictly faithful pixel enlargement. It is useful when added texture is desirable, while a more conservative upscaler may be preferable for documents, logos, evidence, or products that must remain exact.