NAFNet-deblur Image to Image
Input
Hint: Drag and drop image files from your computer, images from web pages, paste from clipboard (Ctrl/Cmd+V), or provide a URL. Accepted file types: jpg, jpeg, png, webp, gif, avif
Customize your input with more control.
Logs
NAFNet-deblur | [image-to-image]
NAFNet delivers specialized image restoration at $0.0225 per megapixel through its nonlinear, activation-free architecture. Trading complexity for focused restoration accuracy, it processes degraded images without the overhead of traditional activation functions. Ideal for photographers and content teams dealing with motion blur, camera shake, or low-light noise in existing imagery.
Use Cases: Photo Restoration | Motion Blur Removal | Low-Light Enhancement
Performance
NAFNet delivers specialized restoration capabilities at a competitive price point for image enhancement workflows, with per-megapixel pricing that scales efficiently for batch processing.
| Metric | Result | Context |
|---|---|---|
| Input Format | Single image (JPG, PNG, WebP, GIF, AVIF) | URL-based input via API |
| Output Resolution | Matches input dimensions | Preserves original aspect ratio and size |
| Cost per Megapixel | $0.0225 | 44 megapixels per $1.00 on fal |
| Seed Control | Optional integer parameter | Reproducible restoration results |
| Related Endpoints | NAFNet-denoise | Denoise variant for noise-specific restoration |
Restoration Without Activation Function Overhead
NAFNet's architecture eliminates traditional nonlinear activation functions (ReLU, GELU, etc.) that typically add computational cost to neural networks. Instead, it uses simple operations like multiplication, addition, and normalization to achieve restoration quality comparable to heavier architectures.
What this means for you:
-
Focused restoration pipeline: Processes blurry or noisy images through a streamlined network designed specifically for degradation removal, not general-purpose image manipulation
-
Deterministic output control: Optional seed parameter ensures identical restoration results across multiple runs of the same image, critical for batch processing or A/B testing restoration settings
-
Format flexibility: Accepts five common image formats (JPG, PNG, WebP, GIF, AVIF) via URL input, eliminating pre-processing format conversion steps
-
Resolution preservation: Outputs match input dimensions exactly, maintaining original composition without forced upscaling or downscaling artifacts
Technical Specifications
| Spec | Details |
|---|---|
| Architecture | NAFNet |
| Input Formats | JPG, JPEG, PNG, WebP, GIF, AVIF (URL-based) |
| Output Formats | PNG with preserved dimensions |
| Restoration Types | Deblur, denoise (via separate endpoint) |
| License | Commercial use enabled |
API Documentation | Quickstart Guide | Enterprise Pricing
How It Stacks Up
NAFNet-denoise ($0.0225/MP) – NAFNet-deblur targets motion blur and focus issues, while NAFNet-denoise addresses sensor noise and grain at identical pricing. Both share the same activation-free architecture but apply different restoration algorithms. Choose deblur for camera shake or subject movement, denoise for ISO noise or compression artifacts.
FASHN Virtual Try-On V1.5 ($0.05/image) – NAFNet-deblur restores existing degraded images at 2.2x lower cost per megapixel equivalent. FASHN generates new composite images by placing garments on models, serving fashion e-commerce workflows where synthesis matters more than restoration.