P-Video Avatar
Pruna AI · AI videoP-Video Avatar is Pruna AI's dedicated talking-head model: animate a portrait image with either a speech audio clip or a text script, and get natural lip sync and facial expression in an MP4.
It is optimized for speed and cost among avatar/lipsync options while preserving identity from a clear front-facing still. Hosted APIs expose 720p and 1080p; output length follows the driving audio (or generated speech from a script).
On VidMachine, AI Influencer scenes use image-to-video at 720p via Replicate with each scene's starting frame and speech audio at 5 credits per second.
Key features and benefits
Portrait plus speech audio
Drive lip sync from a clean portrait and a dialogue track. When audio is provided, it takes priority over built-in text-to-speech script options.
Optional visual direction
A video prompt can nudge how the person appears or gestures while speaking—performance cues without replacing the audio timing.
Talking-head focus
Built for presenter-style shots where the face is the focal point. Best results come from clear, front-facing portraits with minimal occlusion.
720p or 1080p
Provider pricing scales with resolution. VidMachine locks generation to 720p for consistent cost and pipeline behavior.
Technical specifications
ProviderPruna AI (P-Video Avatar)
Primary inputsPortrait image (jpg, jpeg, png, webp); speech audio URI; optional video_prompt
Resolutions720p or 1080p (VidMachine: 720p only)
DurationFollows the supplied audio track (or generated speech when using voice_script on the provider API)
Use cases and applications
AI Influencer talking-head segments: each scene's start frame plus sliced narrator speech for lip-synced delivery.
Virtual presenters, localized spokesperson clips, and social avatars where a single plate defines look and audio defines timing.
Why this model
Choose P-Video Avatar when your brief is talking-head delivery from a still plus speech—not general open-world scene generation.
For broader scene motion or text-only video, use P-Video or other cinematic models in AI Video projects.
How VidMachine uses it
VidMachine generates AI Influencer scene clips via Replicate (prunaai/p-video-avatar) at 720p using each scene's starting frame, motion/video prompt, and speech audio. Output length follows the audio.
Billing is 5 credits per second of output video at 720p. The model is available only on influencer_video_model_priority (not for standard AI Video segments).
What you should know
What resolution does VidMachine use?
Always 720p. The Replicate API also supports 1080p, but VidMachine locks resolution to 720p for consistent cost and pipeline behavior.
Can I use P-Video Avatar for regular AI Video projects?
No. It requires scene speech audio and is listed only under influencer video model priority. Use P-Video for general AI Video scenes.
How is P-Video Avatar billed?
5 credits per second of generated scene video duration at 720p.