P-Video-2-Pro

P-Video-2-Pro is a cinematic video generation model developed by Pruna AI, built upon the MiniMax H3 architecture. It is designed to generate short-form video clips ranging from 5 to 15 seconds in duration with native generated audio. The model supports text-to-video and first/last-frame conditioning, giving users precise control over the beginning and ending visual states of a sequence. Operating primarily at resolutions such as 480p and 768p, P-Video-2-Pro can be scaled up to higher resolutions using complementary upscaling mechanisms while preserving a standard 24 fps frame rate. It features adjustable generation recipes—including speed, quality, and cost modes—allowing developers and creators to balance inference speed and visual fidelity according to their specific requirements. Additionally, the model incorporates independent prompt upsamplers to expand concise inputs and enhance adherence to the user's intent.