P-Image-Upscale is an image-to-image upscaling model developed by Pruna AI. It is engineered to deliver fast, high-fidelity image enhancements with sub-second generation times. The model excels at sharpening details and improving realism without altering the core compositional layout of the input image. This makes it highly suited as a final polishing step in creative pipelines, especially after text-to-image generation or image editing workflows.
The model supports multiple upscaling workflows to cater to different resolution requirements. In its default Target Mode, users specify a destination resolution in megapixels, with processing supporting outputs up to 128 megapixels. Alternatively, Factor Mode allows users to multiply the dimensions of each side of the image by a designated scaling factor, typically between 1x and 8x.
P-Image-Upscale offers two optional enhancement toggles to adjust the style and texture of the output. The detail enhancement feature (enhance_details) focuses on sharpening fine micro-textures and edges, which can increase contrast compared to the original image. The realism enhancement option (enhance_realism) improves the organic textures and visual quality, which is particularly effective for refining artificial-looking segments in AI-generated graphics.
The model is optimized for marketing assets, retail visualizations, VFX concept art, and text-heavy promotional creatives. It is designed primarily to restore and elevate existing details rather than redrawing components or correcting weak source compositions.