About
P Video 2 Pro Reference is Pruna AI's reference-to-video model, the second release on the P Video 2 Pro line. It is built on MiniMax H3 and takes a prompt together with the things the shot has to stay true to: up to nine reference images, three reference videos and three reference audio files in one request. The output is a finished 5 to 15 second scene at 480p or 768p, 24 fps, with generated audio.
References are guidance, not a frame to animate. The prompt decides what is on screen and the references keep it consistent: a character's face across cuts, a product's shape, a location's light, the motion of a clip, the voice on a track. For a frame-exact opening or closing shot, pass a first or last frame image.
What to expect
- •Reference images for characters, products, locations and style, up to nine
- •Reference videos of 2 to 15 seconds for motion, framing and style, up to three
- •Reference audio to drive dialogue, music or pacing, up to three files
- •Soft first and last frame keyframes
- •Generated audio in every output
- •Two generation recipes,
speed(default) andcost - •Clips of 5 to 15 seconds at 480p or 768p, 24 fps, eight aspect ratios including adaptive
- •An explicit seed for reproducible reruns
Pricing
Every component is billed: the output video per second, and each reference on top of it. Per second of output: 480p $0.025 (cost) / $0.0375 (speed), 768p $0.04375 (cost) / $0.075 (speed). Per second of reference video: 480p $0.04375 (cost) / $0.0625 (speed), 768p $0.1125. Per reference image: $0.0125 (cost) / $0.01875 (speed). Per second of reference audio: $0.001875 (cost) / $0.004375 (speed). A reference clip is counted only up to the length of the output.
Specifications
| Base model | MiniMax H3 |
| Inputs | prompt + at least one image or video reference; optional audio references and keyframes |
| Reference images | up to 9 (keyframes included), 20 MB each |
| Reference videos | up to 3, 2 to 15 s each, 15 s combined, up to 1080p 60 fps, 100 MB each |
| Reference audio | up to 3, 15 s combined, 50 MB each |
| Resolutions | 480p, 768p (default 768p) |
| Generation mode | speed (default), cost |
| Duration | 5 to 15 seconds (default 5) |
| Frame rate | 24 fps |
| Audio | generated with the video |
| Aspect ratios | adaptive, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 1:1 |
| Prompt upsampler | off, turbo, max (default turbo) |
| Seed | supported |
What it is not
- •It is not image-to-video: a reference image guides the shot, it does not become
the first frame. Use
first_frame_imagefor that, or P Video 2 Pro. - •It does not edit existing footage. That is P Video Edit.
- •It has no quality recipe. The 768p speed recipe is the top tier.
- •A prompt alone is not accepted: at least one image or video reference is required.
Availability
P Video 2 Pro Reference goes live on Thursday 15 October 2026.
