RunPod has made Pruna AI’s P-Video-Edit model available through a pre-deployed public endpoint. Users can upload a source video lasting up to 15 seconds, describe one edit in text and optionally supply as many as four reference images to guide the result.
The hosted option removes the need to deploy a model or manage GPU infrastructure. Developers can first change parameters and upload media through RunPod’s browser playground. Applications can instead send a job to the endpoint’s asynchronous run route, keep the returned identifier and poll its status until the edited video URL is ready. RunPod recommends this submit-and-poll pattern because an edit can take several minutes, making one long-lived request unreliable.
The documented inputs include prompt enhancement, draft mode, audio preservation and a random seed. The endpoint does not work through OpenAI’s software-development kit, so Python clients call RunPod’s API directly. A sample request reported roughly 252 seconds of execution and a cost of $0.2268, but that single example is not a performance guarantee. The release provides quick access to the editor; output quality, timing and cost still need testing on each application’s own clips.