Magnific adds P-Image, a 2-second image model from Ideogram and Pruna
By Steven Van ·
The joint model generates 1K or 2K images across 13 aspect ratios in about 2 seconds.
Magnific added P-Image, a model built jointly by Ideogram and Pruna AI, and its headline feature is speed: images in just 2 seconds, across 13 different aspect ratios, at 1K or 2K resolution.
Why Pruna's involvement matters
Pruna AI is a model optimization specialist, its open-source framework compresses deep learning models through quantization, pruning, and compilation to make them faster and cheaper to run without materially degrading output quality. Pairing Ideogram's image generation with Pruna's compression expertise is a direct explanation for P-Image's 2-second generation time, that speed comes from real optimization work on the underlying model, not just a lighter-weight architecture trading away quality.
13 aspect ratios is a genuinely wide range
Most image models ship with a handful of standard ratios, square, portrait, landscape. Thirteen options covers a much broader set of real-world use cases without cropping after the fact, social posts, banner ads, product shots, print formats, and more, each generated natively at the right proportions rather than generated square and cropped down.
Speed as the differentiator
For a multi-model platform like Magnific, where users often generate many variations before landing on a final image, a 2-second model changes the actual workflow, iteration becomes closer to real-time exploration than a wait-and-review loop. That's a meaningful practical advantage for anyone doing rapid concept work or generating options at volume.
Sources: Magnific on X, Pruna AI.