Is P-Video-2-Pro-Avatar state of the art?Pending benchmarks.Announced Oct 22, 2026; no independent results yet.
Pruna AI's talking-avatar model, built on MiniMax H3. It turns one portrait and a text script or uploaded audio into a lip-synced talking video, 1 second to 3 minutes long at 24 fps, in 480p or 768p. Scripts are read by one of 30 voices in 10 languages. Available through the Pruna API and Playground, and on Replicate.
The details so far
- Lab
- Pruna AI
- Announced
- Oct 22, 2026
- API
- Live · p-video-2-pro-avatar
- Open weights
- Closed
- Based on
- MiniMax H3 (lab post)
- Input
- One portrait (jpg, png or webp) plus a text script or uploaded audio; uploaded audio sets the timing
- Output
- 24 fps MP4, 1 second to 3 minutes, 480p or 768p; 1080p, 2K and 4K are upscales of the 768p video
- Voices
- 30 voices across 10 languages
- API price
- $0.02 per second of video at 480p, $0.04 at 768p (Pruna docs)
- Availability
- Pruna API and Playground; Replicate (prunaai/p-video-2-pro-avatar)
Tags marked unsure are not stated on the lab's own pages.
What would let us call it
Verdicts use only independent boards we already track. Two consecutive snapshots are required before any crown.
- Publicly announced with a source
- Usable by the public (API or weights)
- Listed on an independent board we track
- Enough results for a rank in a category
- Holds across two snapshots (crowns only)
What the lab claims
Self-reported. Shown for context; not used in any verdict.
Lab claim: about 0.97 s of generation per second of output video at 480p and about 2.2 s at 768p.
Lab claim: realistic talking video with natural expressions and accurate lip sync.
Sources
LABp-video-2-pro-avatar on the Pruna developer docsdocs.api.pruna.ai/guides/models/p-video-2-pro-avatar · Oct 22, 2026LABPruna AI announcement on Xx.com/PrunaAI/status/2113300788713832616 · Oct 22, 2026REPOprunaai/p-video-2-pro-avatar on Replicatereplicate.com/prunaai/p-video-2-pro-avatar · Oct 22, 2026Tracking continues until an independent board lists it.