Skip to content
ReviewRamble

The weekly brief

AI, reviewed, in your inbox.

  • New reviews of tools, models, MCP servers and repos, with the verdict up front
  • Re-checked prices and ratings when something changes
  • New analysis pieces as they are published

Models, agents, MCP servers and the tools actually worth your time.

One email a week. Unsubscribe with the link in any brief. How we use your email

P-Video-2-Pro

Pruna's tuned MiniMax H3 sells speed by the second: three price modes, five resolutions and a mid-table arena score.

Reviewed by ReviewRamble editors · Updated

Visit P-Video-2-Pro Pay per use

Share this page:XLinkedInFacebookRedditWhatsAppEmail

Our rating

3.8/5

P-Video-2-Pro is a sensible, well-documented H3 for pipelines, and nothing about its output makes it the one to beat.

What is P-Video-2-Pro?

P-Video-2-Pro is a video generation model from Pruna AI, the Munich model-compression company, released in September 2026. Pruna's documentation describes it as built on MiniMax H3 and positions it as the quality tier of its P-Video family. One endpoint takes a text prompt, a first-frame image, or a first and last frame, and returns a 5 to 15 second clip at 24 fps with generated dialogue, music and sound effects. You cannot supply your own soundtrack.

It generates at 480p or 768p; 1080p, 2K and 4K are the 768p output upscaled by a separate Pruna model, at an extra $0.02, $0.04 or $0.08 a second. A mode setting chooses between cost, speed (the default) and quality recipes, and an optional prompt upsampler rewrites prompts at no extra charge. Seven aspect ratios are offered, from 16:9 to 1:1.

The safety filter is off by default, though sexual content involving minors and self-harm instructions are always blocked. The API allows 250 requests a minute and bills on the finished clip's length.

Is P-Video-2-Pro worth it?

P-Video-2-Pro is the most transparent of the MiniMax H3 derivatives and among the least distinctive. Pruna publishes a full rate card, a 250-requests-a-minute limit and exact billing rules, which is more than most rivals manage. On Artificial Analysis it ranks 13th for text-to-video at 1128 Elo and 10th for image-to-video at 1119, both below H3 itself.

The pricing is flexible. At 768p a second costs $0.025 in cost mode, $0.035 in speed mode and $0.075 in quality mode, and Artificial Analysis's figure of $5.70 a minute matches the 1080p quality rate of $0.095 a second. That is above HeyGen Video 1's $2.40 a minute, which scores 19 points higher on text-to-video.

The catch: anything above 768p is an upscale, not a native render, and you pay $0.02 to $0.08 a second extra for it. The safety filter is off unless you turn it on, which is convenient for a studio and a liability for a consumer app that has not read the docs. You also cannot feed it your own audio; for that Pruna points to its separate P-Video 2.

Our take: P-Video-2-Pro is the H3 to put in a pipeline when you want to tune each request for price and know exactly what the bill will be. On quality alone, HeyGen Video 1 and MiniMax's own H3 are ahead for less.

Where it earns its place

  • A backend engineer at a stock-footage marketplace

    Generate thousands of short preview clips overnight at the lowest cost

    Batch requests in cost mode at 480p, $0.01 a second, within the 250-a-minute limit, then rerun the clips buyers pick in quality mode at 768p for $0.075 a second.

  • A two-person games studio making trailer drafts

    Animate key art from a first and last frame, with sound

    Send the start and end frames with a prompt describing action and audio; a 10-second 768p clip in speed mode costs $0.35. Upscale only the final cut to 1080p.

Before you commit

Setup
Create a Pruna API key, upload frames with POST /v1/files, then POST /v1/predictions with the p-video-2-pro model; add Try-Sync: true for a synchronous call.
Integrations
Pruna REST API; resold on Runware, Segmind, Wiro, Fliki, each::labs and Scenario; p-video-upscale for 1080p to 4K.
Data and privacy
Pruna's retention and training terms for prompts and frames were not stated on the model documentation we read.
Cost at scale
100 minutes a month at 1080p in quality mode, $0.095 a second, is $570; the same in cost mode at 768p is $150.
Limits
5 to 15 seconds; native 480p and 768p only; 24 fps; 250 requests a minute; no audio input; safety filter off by default.

What’s good

  • Full public rate card with three cost modes per resolution
  • 250 requests a minute and billing on finished clip length
  • Widely resold, from Runware to Scenario

What’s not

  • Below MiniMax H3 and HeyGen Video 1 on Artificial Analysis
  • 1080p, 2K and 4K are upscales of 768p output
  • Safety filter off by default; no audio input

Best for

Developers who want an H3-based video API with predictable per-request pricing and published limits.

P-Video-2-Pro alternatives

P-Video-2-Pro: common questions

Is P-Video-2-Pro worth it?
P-Video-2-Pro is a sensible, well-documented H3 for pipelines, and nothing about its output makes it the one to beat. We rate it 3.8 out of 5.
How much does P-Video-2-Pro cost?
Pruna API per output second (cost / speed / quality mode): 480p $0.01 / $0.02 / $0.04; 768p $0.025 / $0.035 / $0.075; 1080p $0.045 / $0.055 / $0.095; 2K $0.065 / $0.075 / $0.115; 4K $0.105 / $0.115 / $0.155.
What are the downsides of P-Video-2-Pro?
Below MiniMax H3 and HeyGen Video 1 on Artificial Analysis; 1080p, 2K and 4K are upscales of 768p output; safety filter off by default; no audio input.
Who is P-Video-2-Pro best for?
Developers who want an H3-based video API with predictable per-request pricing and published limits.
How much does P-Video-2-Pro cost for a team or in production?
100 minutes a month at 1080p in quality mode, $0.095 a second, is $570; the same in cost mode at 768p is $150.
Is P-Video-2-Pro safe for company data?
Pruna's retention and training terms for prompts and frames were not stated on the model documentation we read.
What are the best P-Video-2-Pro alternatives?
The closest alternatives we have reviewed are HeyGen Video 1, MiniMax H3 (Hailuo 3.0), Pika H3 Flex, Boreal-H3 and LTX-2.5.

Answers checked Nov 21, 2026.