Pruna's P-Video-2-Pro-Avatar ships on MiniMax H3 — and the docs' 4K is an upscale of 768p
neurocartel · reddit · 2026-10-23
Pruna AI released P-Video-2-Pro-Avatar on Oct 22, a talking-portrait model built on MiniMax H3: one portrait plus your audio or a text script yields lipsynced video from 1 second to 3 minutes at 24 fps.
- 30 voice options across 10 languages; controls include voice prompt for delivery, scene motion, seed, and prompt upsampling (on by default, with a vision model adding motion details)
- The catch: 1080p, 2K and 4K in the docs are upscales of a native 768p render — that's the number to compare against rivals
- Distributed via Pruna API and partners including Replicate, Cloudflare and WaveSpeed; a ComfyUI integration is coming
The poster questions how much lip and teeth detail a single portrait can really carry at 768p before upscaling starts inventing it.
More from Multimodal
- LTX2 19B Distilled Faceswap Model Trends on Hugging Face — linoyts · 2026-10-23
- Blender 5.2's new SDF Grid makes volumetric slicing trivial, in both directions — 5agado · 2026-10-23
- MOSS Image base model showcase: photoreal outputs, full reproducible settings — COMPLOGICGADH · 2026-10-23
- Kling AI officially launches Kling 4.0: 15 references per task, 30s clips, 10 keyframes — testingcatalog · 2026-10-23
- Run Lola Run director Tom Tykwer uses Seedance in hybrid AI filmmaking experiment — Uncanny_Harry · 2026-10-23
- "Still Worth Something" S01E01: a Pixar-style fully AI-animated comedy short — LPGroves · 2026-10-23