Pruna P Video 2 Pro

P Video 2 Pro — Cinematic Text & Image-to-Video with Native Audio

What is P Video 2 Pro?

P Video 2 Pro is Pruna AI's cinematic video generation model and the final release in the P-Video family, built on MiniMax H3 (Hailuo 3.0) and tuned by Pruna for a much faster generation pass than the base API. One endpoint takes three kinds of input — a text prompt alone, a first-frame image, or a first-and-last-frame pair — and returns a finished clip of 5 to 15 seconds at 480p or 768p, 24 fps, in seven aspect ratios.

Its signature is audio. Every output comes back with sound generated in the same pass as the picture — dialogue, ambience and effects — so you write the sound you want directly into the prompt instead of adding it in a separate step. That is the capability that most clearly separates the underlying model from silent text-to-video generators.

Key Features

• •Native generated audio in every clip — 32 kHz stereo, synced to the motion; write dialogue or sound effects into the prompt.
• •Three ways in, one endpoint — text-to-video, first-frame image-to-video, and a first-and-last-frame lock that steers where the shot ends up.
• •Three recipes — speed (the fast default), quality (more compute on the same prompt) and cost (speed's output at the lowest per-second rate, for high-volume runs).
• •Prompt upsampler at three levels — off, turbo, max — independent of the recipe.
• •5–15 seconds, 480p or 768p, 24 fps, in 16:9, 9:16, 4:3, 3:4, 3:2, 2:3 and 1:1. A reference image sets the canvas, so the aspect ratio follows the image when you supply one.
• •Reproducible with an explicit seed.

Best Use Cases

P Video 2 Pro is aimed at film-style beats where sound matters. In testing it rendered a forge scene with rhythmic hammer clangs, a taiko drummer whose still image animated into booming strikes, a rocket launch driven by first-and-last-frame images with a building engine roar, and a vertical quality-mode wok toss ready for social. That range maps to brand and product films with sound baked in, social-first vertical clips, game-world and pre-visualization shots, and dialogue two-shots — all without a separate audio mix.

Prompt Tips and Output Quality

Structure the prompt as subject, action, camera, then the sound. Keep one clear action and one clear sound event so the audio lands cleanly. Supplying a first-frame image switches the model to image-to-video and the aspect ratio follows that image; add a last-frame image to lock where the shot resolves. Use speed for drafts and quality for finals, 768p for delivery and 480p to iterate. Fix the seed to refine a shot across runs.

How to Use P Video 2 Pro

P Video 2 Pro turns a text prompt — or a first frame, or a first-and-last-frame pair — into a 5–15 second cinematic clip with audio generated in the same pass. The craft is in writing the sound alongside the picture and picking the recipe and resolution to match the job.

Getting the most out of it

Write the prompt as subject, action, camera, then sound, and keep it to one clear action with one clear sound event so the audio reads. To animate a still, pass it as the first-frame image; the aspect ratio then follows that image. To steer where a shot lands, add a last_frame_image from the same scene — the model fills the motion between the two frames. Reach for mode: quality on finals and stay on the faster speed default while you iterate.

Recommended Settings

• •Text-to-video (default): resolution: 768p, mode: speed, duration: 6, aspect_ratio: 16:9, prompt_upsampler: turbo, seed fixed for repeatable runs.
• •First-frame image-to-video: add image; drop aspect_ratio (the image sets it).
• •First-and-last-frame lock: add image and last_frame_image from the same scene in different states.
• •Social / finals: aspect_ratio: 9:16 and mode: quality; use 480p to halve cost while drafting.

Example prompts to try

Each example below ran end to end at the settings shown and came back with a synced audio track and motion that held through the clip.

• •Text-to-video, audio hero (default). Prompt: "A blacksmith in a dim workshop forge hammers a glowing orange blade on an iron anvil, bright sparks bursting on each strike, then lifts the blade and plunges it into a water barrel. Sound: the rhythmic clang of hammer on steel, the low roar of the forge, and a sharp hiss of steam as the steel is quenched." The clip returned rhythmic hammer clangs over a glowing blade and bursting sparks — keep the sound to one repeating event so it lands cleanly.
• •Quality recipe, vertical 9:16. Prompt: "A chef tosses vegetables in a wok over a roaring gas burner; the pan bursts into flame on the toss and the oil sizzles. Sound: the loud sizzle of oil, the whoosh of the flame, and the clatter of the wok." Quality mode at 9:16 read cleanly for social — switch to quality for finals and keep speed for drafts.

Reference-guided examples

P Video 2 Pro takes up to two reference frames: a first frame, and an optional last frame that steers the ending. Address them as the first-frame image and the last_frame_image.

• •First-frame image-to-video. Prompt: "The taiko drummer swings both sticks down and strikes the big drum, then again in a driving rhythm, his whole body behind each hit. Sound: deep booming taiko drum hits echoing through the wooden hall, sharp and rhythmic." A still of a raised-stick drummer animated into a full strike with booming hits — supply a first frame where the subject is poised to move so the motion has an obvious next beat.
• •First-and-last-frame lock. Prompt: "The rocket ignites and lifts off the pad, climbing on a bright exhaust plume as a rolling dust cloud spreads beneath it and the gantry falls away. Sound: a deep building engine roar and the crackle of burning fuel." With a first frame (rocket on the pad) and a last frame (rocket lifted off), the shot started and ended on those frames — give it two frames of one scene in different states to steer the arc.

Further shots the model is built for: "a barista steams milk, the wand hissing," "a drummer counts in and the band starts," or "a potter centres clay on a spinning wheel."

Does P Video 2 Pro generate audio?

Yes — native stereo audio is produced in the same pass as the video. Write the dialogue, ambience or effects into the prompt.

Can I upload my own audio track?

No. P Video 2 Pro takes no audio input; use P Video 2 to drive a clip from an existing track.

Does it do image-to-video?

Yes — supply a first-frame image, and optionally a last-frame image to steer the ending. The aspect ratio then follows the image.

What resolutions and lengths are supported?

480p or 768p, 5 to 15 seconds, 24 fps, in seven aspect ratios.

What is the difference between speed, quality and cost mode?

Speed is the fast default; quality spends more compute on the same prompt for a cleaner render; cost gives speed's output at the lowest per-second rate, for high-volume production.

Can it edit existing footage?

No — P Video 2 Pro only generates new scenes. Use P Video Edit for edits and P Video Avatar for talking-head avatars.