P-Video is a video generation model made by Pruna AI. A video generation model is a neural network that produces video frames from a description. P-Video accepts text, image and audio input and produces clips of up to 15 seconds, with or without sound, at 720p or 1080p.
A program calls Pruna's prediction API with the model name p-video; the model is also offered through other inference providers. It can generate its own audio, including dialogue with lip sync, or use an audio file that the user supplies. A draft mode makes previews four times faster and cheaper, and prompt upsampling expands a short prompt. Pruna recommends prompts built from subject, action, scene, camera, lighting, style and audio. Related models in the same family make a portrait speak, swap characters or edit existing footage.
Pruna sells it pay per use, priced per second of output video, and it can be tried in the Playground of the Pruna user portal.