Super resolution uses AI to upscale images while adding realistic detail. How it works: Neural networks learn mapping from low-res to high-res images, predicting plausible high-frequency details (textures, edges, fine features) that aren't in the original. Key architectures: ESRGAN (Enhanced Super-Resolution GAN) pioneered realistic upscaling, Real-ESRGAN handles real-world degradation (blur, noise, compression), SwinIR uses transformer attention for better quality. Use cases: Upscale old photos/videos, enhance surveillance footage, improve game textures, prepare images for large prints. Limitations: Cannot recover information that wasn't captured - AI hallucinates plausible details. Faces and text can distort. 2x upscaling most reliable, 4x+ increasingly fabricated. Popular tools: Topaz Gigapixel AI (commercial, excellent quality), Real-ESRGAN (open source), Waifu2x (anime-optimized), Upscayl (free GUI). Tips: Clean source images before upscaling, use face-specific models for portraits, multiple smaller upscale passes sometimes beat single large jump.