AI Models & Enhancement Presets

PixelForge includes four fine-tuned enhancement presets tailored to different categories of images.

1. General Photo AI (`Real-ESRGAN x4plus`)

Primary Use Case: Real-world photography, landscapes, architecture, nature, and realistic textures.
Under the Hood: Powered by the full Real-ESRGAN x4plus deep convolutional neural network. Reconstructs realistic details while suppressing JPEG artifacts and high-frequency noise.
Recommended Scale: 2× or 4×.

2. Anime & Illustration (`Real-ESRGAN x4plus Anime 6B`)

Primary Use Case: Digital artwork, anime screenshots, manga, sketches, vector graphics, and 2D illustrations.
Under the Hood: Uses a 6-block compact generator optimized specifically for line art crispness, flat color preservation, and halo prevention.
Recommended Scale: 2× or 4×.

3. Portrait & Face Enhance (`AI + Detail`)

Primary Use Case: Portraits, headshots, family photos, and pictures with human subjects.
Under the Hood: Real-ESRGAN neural super-resolution coupled with a skin-preserving micro-contrast filter that enhances eye detail and hair strands while preserving natural skin tones.

4. Fast Lanczos3 HD (`Instant CPU Resampling`)

Primary Use Case: High-speed batch resizing, clean graphic scaling, or environments where no GPU engine is installed.
Under the Hood: Multi-stage Lanczos-3 interpolation kernel with adaptive unsharp masking. Runs entirely on CPU with zero neural model download required.
Speed: Instantaneous (50–150ms per image).