Two Upscaling Models for Photos and AI-Generated Images

Wonder 3.5 targets photos, while Bloom 2 focuses on AI-generated images. Both upscaling models are available in Topaz Image Web and Mobile.

Wonder 3.5 and Bloom 2 are joining Topaz Image Web and the mobile app. The first supports a broad range of photographs and general images, while the second is specifically designed for AI-generated visuals, illustrations, and digital renders.

Wonder 3.5 builds on Wonder 3 with improved precision, realism, and detail recovery. Topaz highlights fewer repetitive patterns, better handling of visual noise, and more consistent results on heavily compressed or low-resolution sources. Text, tables, and structured graphics are also expected to be rendered more accurately.

Bloom 2 focuses more directly on flaws commonly found in AI-generated images. The model aims to enrich materials and textures while reducing invented details, anatomical distortions, and the artificial appearance of certain faces. Several creativity levels allow users to move from subtle corrections to more substantial transformations, with options for face recovery and color preservation.

Both models can also be used in workflows integrating ComfyUI, fal, Oxen AI, Krea, and Pixelcut. The desktop version of Wonder 3.5 is not yet available. Topaz says it plans to continue refining the model based on feedback from the web service before adding it to Topaz Photo, with no release timeline provided.