P-Video-2: AI Video With Sound, Up to 20s (Guide & Price)

P-Video-2 is Pruna AI’s newer video model: it makes 1 to 20 second clips from text or an image, at 720p or 1080p, with dialogue, music and sound effects generated in the same pass. On VdoBloom it costs 2 credits per second at 720p and 4 credits per second at 1080p (about 40% less with Draft on), plus one extra frame per clip. A 5-second 720p clip is 11 credits, and a 4-second one, at 9 credits, fits inside the 10 free credits a new account starts with.

Disclosure: VdoBloom is our platform.

This guide covers what changed from the original P-Video, what every setting does on VdoBloom, what it really costs, how to prompt it for speech and sound, how the first-and-last-frame mode works, and when a different model with native audio is the better pick.

What P-Video-2 is, and what changed from P-Video

P-Video-2 was released on September 10, 2026 by Pruna AI and is served through Runware. Pruna positions it as a quality successor to P-Video rather than a speed upgrade. As of September 2026, Pruna’s P-Video-2 documentation lists the main gains as native-speech lip sync, sharper close-ups and “identity lock”. In practice that means:

• Sharper foreground subjects. Faces and objects close to the camera hold more detail.
• Stronger consistency. A person or product keeps the same look across the clip, and image-to-video stays closer to the photo you upload.
• Better lip sync for its own speech. When a character speaks a line the model generates, mouth movement lines up more closely with the audio.
• Clearer on-screen text. Signs, labels and short titles are more readable.
• Native audio. Dialogue, music and sound effects come out with the video. No separate audio step.
• Longer clips. Up to 20 seconds, where P-Video tops out at 10.

Both models sit next to each other in the VdoBloom model picker, so you can compare them on the same prompt. Here is how they differ on VdoBloom specifically:

P-Video is still slightly cheaper per second. Pick it for quick, simple shots where close-up detail and speech don’t matter much. Pick P-Video-2 when a face, a spoken line, readable text or a clip longer than 10 seconds matters. The P-Video 2 model page has the new model’s specs at a glance, and the P-Video model page has the original model’s full specs.

Settings and price on VdoBloom

Settings

P-Video-2 is available on two tabs: Text to Video and Image to Video. Open the model picker and choose P-Video 2 in the P-Video family.

• Duration: any whole number from 1 to 20 seconds, set with a slider.
• Resolution: 720p (standard) or 1080p (high quality).
• Draft (cheaper preview): a faster, lower-detail pass for testing a prompt. It works at either resolution.
• Aspect ratio (text to video only): 16:9, 9:16, 4:3, 3:4, 3:2, 2:3 or 1:1. On Image to Video there is no aspect setting, because the output follows the shape of your image.
• Audio: always on. Every clip comes with a soundtrack, and the price already includes it.
• Frame rate: 24 fps.
• Prompt: up to 2,000 characters.

How much P-Video-2 costs

Pricing is per second of video, plus one frame: every clip comes out one frame (1/24 of a second) longer than the length you pick, and that frame is billed too. The total is rounded up to a whole credit.

Three worked examples:

• A 5-second 720p clip is 11 credits (5.04 seconds × 2 = 10.08, rounded up).
• A 10-second 1080p clip is 41 credits (10.04 × 4 = 40.17, rounded up).
• A 5-second 720p Draft is 7 credits (5.04 × 1.2 = 6.05, rounded up). The maximum, a 20-second 1080p clip, is 81 credits.

The cost is shown before you generate. If a generation fails, the credits go back to your balance.

Speed, and when to use Draft vs standard

Be ready to wait. P-Video-2 is not the fastest model on VdoBloom. In VdoBloom’s integration runs on September 25, 2026, 2-second 720p Draft clips took about 4 to 5 minutes from submitting the job to getting the finished file. Longer or 1080p clips can take longer. You don’t need to keep the tab open. The video appears in your library when it is done.

Draft doesn’t make the wait short. Its real value is cost. Use it like this:

1. Iterate in Draft at 720p. Try your prompt at a short length (3 to 5 seconds) until the action, camera, voice and sound are right. Each 5-second attempt is 7 credits.
2. Lock the prompt. Change one thing at a time. If the voice is wrong, fix only the dialogue part of the prompt.
3. Render the final without Draft. Switch Draft off, pick 720p or 1080p and set the full length. Choose 1080p when faces or on-screen text need to hold up on a large screen, or when you plan to crop.

Draft output has less detail, so don’t judge fine texture or small text from it. Judge motion, timing, framing and audio.

How to prompt for dialogue and sound

Because audio is always on, the model will fill the soundtrack with something even if you don’t ask. Describing the sound you want gives you control over it.

Dialogue

• Put the spoken line in quotes and say who says it. For example: A woman in a yellow raincoat looks at the camera and says, “Okay, this is the last stop.”
• Match the line to the length. A 5-second clip fits one short sentence. A 20-second clip fits a few sentences with pauses. Too many words for the clip length leads to rushed speech or cut-off lines.
• Describe the voice. “Calm low voice”, “excited, slightly out of breath”, “whispering”.
• Keep one speaker per line. If two people talk, write them in order: He asks, “Ready?” She nods and says, “Ready.”
• Frame the speaker. Lip sync is easiest to see in a medium shot or close-up, facing the camera. That is also where P-Video-2’s sharper close-ups help most.

Sound effects and music

• Tie sounds to actions. “Her boots crunch on gravel”, “the kettle clicks off”.
• Describe the room. “Quiet office, air-conditioning hum”, “busy street, distant traffic”.
• Say what you don’t want. “No music, natural sound only” or “no voices” is often the difference between a usable clip and one with a random soundtrack.
• Name the music style if you want music. “Soft piano in the background, very quiet”.

A full example for a 6-second 9:16 clip: Close-up of a barista behind a counter, morning light. She slides a cup forward and says, “Oat flat white, extra hot.” Espresso machine hissing, low cafe chatter, no music. At 720p that is 13 credits, or 8 credits in Draft (6.04 × 1.2 = 7.25, rounded up).

For a broader look at how native audio works across models, see our guide to AI video generators with sound.

Image-to-video tips

On the Image to Video tab, your photo becomes the first frame and P-Video-2 animates from there. Its stronger input consistency means the result stays closer to your image, but the input still decides a lot.

• Crop first. The output follows the image’s shape. For a vertical video, upload a vertical image.
• Use a sharp, well-lit photo. Blur and heavy compression carry into the video.
• Don’t re-describe the whole image. The model can see it. Spend the prompt on motion, camera and sound: “slow push-in, he turns to the window and says, ‘It’s still raining.’ Rain on glass.”
• Keep motion plausible for the length. A big change of pose in 2 seconds looks rushed. Give it 5 or more.

Photos of real people

Only animate your own photo, or a photo of someone who has agreed to it. Never use images of minors. Don’t create intimate or sexual content of real people, and don’t put public figures into content meant to deceive. VdoBloom blocks content involving minors and non-consensual content, and prompts are screened before a generation starts.

First and last frame: morph between two images

On the Image to Video tab, P-Video-2 has an extra slot under your main image called Last frame (optional). When you add one, the video starts on your first image and ends on the second, and the model fills in the motion between them. In VdoBloom’s integration run the clip ended exactly on the supplied last image.

Good uses:

• Before and after. A messy room to a clean one, a sketch to a finished render, day to night.
• Product reveals. Closed box to open box, or a product from the front to the back.
• Controlled camera moves. A wide shot as the first frame and a close-up of the same scene as the last.
• Loops. Use the same image as both first and last frame for a clip that returns to where it started.

Steps:

1. Choose P-Video 2 on the Image to Video tab and upload the first image.
2. Add the last frame in the slot below it. Use the same aspect ratio as the first image, since the output follows the first image’s shape.
3. Describe the transition and the sound, for example “the camera slowly orbits as the lights switch on one by one, soft electrical hum”.
4. Give it enough time. Two very different images need 6 to 10 seconds to look natural instead of a quick blend.

The last frame is also a way to chain clips into a longer piece: end one clip on a frame and start the next clip from that same image. Our guide on making AI video longer covers that workflow in detail.

P-Video-2 vs other VdoBloom models with native audio

Several models on VdoBloom generate sound with the video. Prices below are for the settings shown, at the time of writing.

How to choose:

• Lowest cost per second with sound: P-Video-2. An 8-second 720p clip is 17 credits, versus 40 for an 8-second Veo 3.1 Fast clip.
• Longer than 20 seconds in one clip: Wan 3.0 or Seedance 2.5, both up to 30 seconds.
• You want to supply your own audio: Seedance 2 accepts reference audio clips. P-Video-2 on VdoBloom does not take audio input.
• Top-end motion and detail: Seedance 2 or 2.5, at several times the price.

A sensible approach is to find the idea with P-Video-2 in Draft, then decide whether the final needs a more expensive model at all.

Frequently asked questions

Is P-Video-2 free?

Not unlimited, but you can try it without paying. New VdoBloom accounts get 10 free credits, which covers one 4-second 720p clip (9 credits) or a 5-second 720p Draft (7 credits) with credits left over. After that it is 2 credits per second at 720p, plus the extra frame.

How long does P-Video-2 take to generate a video?

Expect minutes, not seconds. In VdoBloom’s integration runs, 2-second 720p Draft clips took about 4 to 5 minutes end to end, including time in the provider’s queue. Longer and 1080p clips can take more. You can leave the page; the video is saved to your library when it finishes.

Does P-Video-2 make videos with sound?

Yes. Every clip comes with audio: dialogue, music and sound effects generated with the picture. On VdoBloom audio is always on and included in the price. Describe the sound in your prompt, and write “no music” or “no voices” if you want a cleaner track.

What is the maximum length of a P-Video-2 video?

20 seconds per clip, at 720p or 1080p. For longer videos, chain clips using the last frame of one as the first frame of the next, or use Wan 3.0 or Seedance 2.5, which go up to 30 seconds.

Can I use my own audio with P-Video-2?

Not on VdoBloom yet. Pruna’s API supports audio-conditioned generation, but VdoBloom’s P-Video-2 only generates its own audio from your prompt. If you need a clip driven by your own sound, Seedance 2 accepts reference audio.

Why is P-Video-2 slow?

It generates the video and a full soundtrack together, and each job waits in the provider’s queue before it runs. Draft mode lowers the cost of testing more than the wait, so the fastest workflow is to write a careful prompt, test short Draft clips, and render the final only once.