Current evidence summary
Pruna's P-Video-2-Pro is an H3-based text/image-to-video model with generated audio, first-frame and first+last-frame conditioning, 480P/768P output, 5–15 second clips and cost/speed/quality modes. It is not a simple higher-spec P-Video-2: Pro trades P-Video-2's 1080P/20-second/48fps/audio-input surface for a different conditioning and quality-speed stack. The source-backed use cases currently point to fast hosted t2v/i2v with native generated audio, first-frame and first+last-frame conditioning, teams comparing returned-duration and requested-duration provider billing, cost/speed/quality mode selection during the launch promotion.
External tests and provider examples
These are reports from the named publishers and providers; FrameSignal did not run these generations.
text to video music video
Reported metric: raw generation charge — 0.3797 USD
Runware publishes a 10-second 1344x768 quality-mode bronze-foundry example with a raw $0.3797 charge, output URL and about 1m9s end-to-end time.
Reported result: Provider displays about 1m9s end-to-end and exposes the generated output URL plus response cost.
Prompt: Bronze foundry art-pop music-video prompt; full prompt preserved at provider source.
Settings: duration seconds: 10 · resolution: 1344x768 · mode: quality · prompt upsampling: max · seed: 1061588653
Open source ↗text to video product demo
Reported metric: raw generation charge — 0.3797 USD
Runware publishes a second 10-second 1344x768 quality-mode self-rising flood-barrier example with a raw $0.3797 charge and about 1m23s end-to-end time.
Reported result: Provider displays about 1m23s end-to-end and publishes raw output.
Prompt: Self-rising flood-barrier demo prompt; full prompt preserved at provider source.
Settings: duration seconds: 10 · resolution: 1344x768 · mode: quality · prompt upsampling: max · seed: 1984884212
Open source ↗first frame image to video
Reported metric: raw generation charge — 0.0904 USD
Runware publishes a 5-second 768P speed-mode first-frame example with a raw $0.0904 charge and about 18 seconds end-to-end time.
Reported result: Provider displays about 18 seconds end-to-end and raw output.
Prompt: Tidal Accordion music-video prompt with supplied first frame; full prompt/input available at source.
Settings: duration seconds: 5 · resolution: 768P · mode: speed · seed: 1900679139
Open source ↗first last frame conditioning
Reported metric: raw generation charge — 0.5391 USD
Runware publishes a 15-second 768P quality-mode first+last-frame example with a raw $0.5391 charge and about 1m3s end-to-end time.
Reported result: Provider displays about 1m3s end-to-end and raw output.
Prompt: Cryogenic seed-vault B-roll prompt with first and last frames; full materials available at source.
Settings: duration seconds: 15 · resolution: 768P · mode: quality · prompt upsampling: turbo · seed: 1498766104
Open source ↗official provider example
Reported metric: total job duration — 10.8 seconds
Replicate exposes prompt, settings, output and a 10.8-second total job duration for an official P-Video-2-Pro example.
Reported result: Replicate exposes prompt/settings/output and total job duration; exact price was not captured.
Prompt: Example prompt and settings preserved on Replicate's official Pruna model page.
Settings: provider: Replicate · model: prunaai/p-video-2-pro
Open source ↗provider documented example
Reported metric: provider p50 runtime — 20 seconds
EachLabs documents an 8-second 768P quality example and a provider p50 runtime around 20 seconds.
Reported result: Provider documents a p50 runtime around 20 seconds and an example workflow.
Prompt: 8-second 768P quality product-film example; full prompt/settings available at provider page.
Settings: duration seconds: 8 · resolution: 768P · mode: quality
Open source ↗provider documented example
Reported metric: generation time — 49.98 seconds
Segmind documents a 6-second 768P speed example with approximately 49.98 seconds reported generation time.
Reported result: Provider reports approximately 49.98 seconds for the documented example.
Prompt: Blacksmith-generation example; exact prompt/settings preserved at provider page.
Settings: duration seconds: 6 · resolution: 768P · mode: speed
Open source ↗What official evidence suggests
Promising areas
- Documented text-to-video, image-to-video and first/last-frame conditioning
- Native/generated audio with 480P and 768P tiers
- Pruna exposes cost, speed and quality modes plus a current API route
- Provider examples preserve raw charges, seeds and end-to-end timing
Verify before paying
- Generalized V2V/editing and dedicated lip-sync remain unverified
- Maximum 768P, 15 seconds and 24fps are lower than some P-Video-2 surfaces
- Pruna bills returned duration while WaveSpeed bills requested duration
- No accepted independent numeric benchmark is available yet