4 to 32 megapixels, or a 2K–8K target
Most versions take a megapixel target; the SeedVR2 pair take a resolution target instead. The setting you see is the one that version actually uses.
- 4 MP
- 8 MP
- 16 MP
- 32 MP
- 2K
- 4K
Loading service status…
Convert images from low-resolution to high-resolution
Enlarging an image is arithmetic; restoring one is not. The upscaler adds detail, cleans noise and repairs faces rather than interpolating what is already there — which is why the gains are dramatic on a small or degraded source and modest on something already sharp. Eight versions cover the range from a cheap, natural clean-up to a diffusion restoration that targets 8K.
Most versions take a megapixel target; the SeedVR2 pair take a resolution target instead. The setting you see is the one that version actually uses.
The `safe_face` versions restore the rest of the frame while leaving the face alone, so a likeness survives the pass.
Creativity decides how freely the model may add detail that was not in the source. Low keeps it honest; high reconstructs.
The SeedVR2 versions rebuild detail rather than sharpening it, which is what old and heavily compressed sources need.
Generate multiple images or videos at once. Save time with efficient workflows.
The cheapest version starts at a single credit. The exact figure is shown before you run it.
Upload the image you want enlarged or restored.
Pick a version and a target size. Face-safe variants are marked.
Download at full resolution, with no watermark.
Resizing stretches the pixels you have. The upscaler reconstructs detail the source never recorded — texture, edges, facial features — using a model trained on what those things look like at higher resolution.
Natural Clarity supports 4, 8 or 16 MP; Ultra Clarity supports 16, 24 or 32 MP; the Basic and Premium modes support 4 or 8 MP; the SeedVR2 versions target 2K, 4K or 8K instead of a megapixel figure.
To a degree, yes — that is what makes it better than interpolation, and it is why the creativity control exists. Keep it low when accuracy matters, and use a face-safe version when a likeness has to survive.
The diffusion restoration versions are the ones for that. They rebuild rather than sharpen, which is the right behaviour for a scan with grain, compression damage or fading.
No, Seduca does not add watermarks to generated content. All images and videos you create are delivered without watermarks, ready for immediate use in your projects.