Consistent with human anatomy

Detail Model • Type of model: ESRGAN variant specifically trained for anime/manga datasets. • Scale: 4× (example: 512×512 → 2048×2048). • Characteristics: • Focus on line art: lines become clearer, not blurred. • Shading is more defined, especially in anime cel-shading. • Colours are cleaner, reducing noise or artefacts. • Results are more "stylised" compared to photo-realistic models. ✨ Advantages • Anime/manga scan: Cleans lines and text, suitable for digital remaster. • Digital illustration: Maintains original style while increasing resolution. • Cartoon-style comics: Darker black lines, more consistent shading. • Fanart / digital artwork: Provides an upscale result that stays true to the anime style. ⚠️ Disadvantages • Less suitable for human photos → results may look artificial. • Light oversharpening on low-res images with thin lines. • Not natural for landscapes or realistic objects.

Sharp but still natural upscale results, suitable for professional use (photography, catalogue, fashion, documentation). Model Details • Model type: Variant of Remacri, trained with a high-resolution photo dataset. • Scale: 4× (example: 512×512 → 2048×2048). • Characteristics: • Maintains realism while adding sharpness. • Colours and textures are more consistent compared to UltraSharp. • Not too "smooth" like RealESRGAN_x4plus, but also not too "crispy" like UltraSharp. • The result is balanced: sharp, clear, but still natural. ✨ Advantages • Face photos: Skin, eye, and hair details are clearer without looking artificial. • Product/fashion: Fabric textures, materials, and objects stand out more. • Architecture: Building lines, wall textures, and small details are sharper. • Catalog/commercial: Suitable for images that need to look professional and polished. ⚠️ Disadvantages • Less suitable for anime/artwork → results may look too "photo-realistic" and lose the original style. • Light oversharpening can occur on very small low-res images. • Not as strong as UltraSharp for game textures or line art.