Skip to main content

HitPaw Enhancement API

Access the world's most powerful AI media enhancement models designed for professional scale. Our technology provides the "last-mile" solution for improving the quality of images and videos, trusted by industry leaders to deliver industrial-grade results for everything from vintage archives to modern user-generated content.

Our models excel at handling specific degradation types—including noise, blur, compression artifacts, and low resolution—to produce high-fidelity outcomes.

1. Image Enhancement Models​

Our image models are categorized into two distinct classes to suit different needs: Standard (focusing on fidelity and accuracy) and Generative (focusing on creativity and reconstruction).

Super-Resolution & Upscale​

  • General Enhance: The go-to solution for 2x/4x upscaling in general scenarios. It balances noise reduction with detail generation, effectively handling the mixed-quality images found on social networks.
  • High Fidelity: Designed for high-quality source files. It upscales resolution while strictly preserving the original artistic intent and fine textures, avoiding the artificial "processed" look common in other upscalers.

Portrait Optimization​

  • Portrait Clear (Soft): A dual-purpose model that sharpens background details while applying a beautifying, soft-focus effect to facial features. Ideal for scenarios requiring a balance between clarity and aesthetics.
  • Portrait Natural (Textured): A V2 model built for realism. It prioritizes the recovery of high-fidelity skin textures (pores, grain), restoring faces with photographic accuracy rather than artificial smoothing.
  • Generative Portrait: Utilizes Diffusion technology to reconstruct faces in severely degraded or blurry photos, effectively generating plausible high-quality details where the original data is missing.

Restoration (Denoise & Sharpen)​

  • Sharp & Detail Denoise: Specialized 1x enhancement models that aggressively remove sensor noise and grain while pulling focus to the subject, ensuring crisp edges without introducing artifacts.

2. Video Enhancement Models​

Video restoration requires a specialized approach utilizing Temporal Stability. Our video models ensure that enhancements are fused consistently across multiple frames to prevent flickering, jitter, and artifacts.

Cinematic Upscaling​

  • Ultra HD Model: The industry standard for converting SD/HD footage to 4K. It uses deep convolution to excavate fine details and smooth out jagged edges (aliasing), resulting in a clean, natural high-definition picture.

Video Portrait Restoration​

  • Face Soft (Video): Automatically detects and tracks faces in motion, applying stable skin smoothing and feature enhancement. It uses facial prior knowledge to beautify subjects while maintaining identity.
  • Portrait Restore: A heavy-duty repair model for blurry or noisy faces in video. It fuses data from continuous frames to reconstruct facial features (eyes, mouth) that are otherwise lost in motion blur.

General Video Restoration​

  • General Restore: A versatile GAN-based model designed for complex inputs. It simultaneously de-noises, de-blurs, and removes compression artifacts, making it ideal for remastering old clips or fixing low-bitrate internet videos.
  • Generative Video: Powered by Stable Diffusion, this model reconstructs textures and details in extremely low-quality footage, bridging the gap between restoration and creation to salvage unusable content.

Core Advantages​

  • Quality: Industry-defining results fit for professional use cases, from Hollywood studios to AAA game development.
  • Fidelity: Unlike generic filters, our models are trained to preserve the original identity of subjects and the authentic texture of surfaces.
  • Versatility: With both Standard (accurate) and Generative (creative) options, you can choose the exact tool needed for restoration vs. reconstruction workflows.