Clarity Upscaler: Generative Detail Enhancement
Clarity Upscaler is an open-source diffusion-based upscaler that combines a tiled Stable Diffusion pipeline with ControlNet tile guidance. Instead of merely interpolating pixels, it re-synthesizes texture, edges and fine structure at the target resolution, producing results that look like a higher-resolution photograph. It is the primary engine for youcut.ai's Image Upscaler (2x/4x/8x) and the sharpening stage of the Photo Enhancer.
Strengths
- • Adds convincing texture (skin, fabric, foliage)
- • Tunable creativity vs. fidelity
- • Handles heavy JPEG artifacts
- • Scales to 8K via tiling
Limitations
- • Slower than feed-forward upscalers (10–30 s at 4x)
- • High creativity settings can alter fine details
- • Small text may be re-drawn incorrectly
Best for
- • Photos destined for print
- • Low-quality web images and screenshots
- • AI-generated images needing more resolution
| Specs | |
|---|---|
| Task | Generative super-resolution |
| Scales | 2x, 4x, 8x (tiled) |
| Latency | ~5 s (2x) to 30 s (8x) |
| Provider key | fal-ai/clarity-upscaler |
Try Clarity Upscaler in
Compare
Frequently asked questions
Does Clarity Upscaler change my image?+
At youcut.ai's default fidelity setting, composition and identity are preserved while texture is re-synthesized. Turn on face enhancement only for portraits to avoid unnecessary changes.
How does it compare to Topaz Gigapixel?+
Both add detail; Clarity is diffusion-based and better at reconstructing texture from very low-quality sources, while Topaz runs locally and requires a paid desktop license and a GPU.
What is the maximum output?+
8192 px on the long edge through tiled processing.
Why is 8x slower?+
The image is processed in overlapping tiles; 8x produces 64 times the pixels of the original, so many more tiles are diffused.
When does youcut.ai use AuraSR instead?+
For fast previews and for illustrations where exact line preservation matters more than texture synthesis.