Live-session photos and clips on the same character.
An interactive cam-style session runs on latency: a user asks for a pose or an action and the app needs a result back inside the session, not a coffee break later. realistic-sharp-v1’s ~9 second render and video-v1’s async webhook are built around that clock.
This page is for teams building cam-style or interactive companion apps, where a live session needs a fast image or short clip back from a character in real time. Xavira’s fastest photoreal model (realistic-sharp-v1) renders in about 9 seconds; realistic-v2 is available for a sharper result when a moment is worth the extra few seconds; video-v1 delivers longer clips via webhook or polling.
Which models
Pick based on speed vs. sharpness — every model shares the same character and credit system.
realistic-sharp-v1RecommendedSub-10-second renders keep an in-session request from stalling the interaction.
realistic-v2A sharper, slower option for a highlight moment worth the extra ~6 seconds and the second credit.
video-v1A 5 or 10 second clip for the "show me" moment — async by design, so the app polls or gets a webhook rather than holding a connection open.
Example prompts
Prose, not weighted tokens — Xavira adds the character’s traits and a quality prefix server-side.
“standing by the window, backlit by streetlight, playful expression, looking directly at the camera”
“sitting on the edge of the bed, phone propped up, casual loungewear, soft lamp light”
Example generations — SFW and 18+ (blurred)
Real output from the model(s) above. Adult examples are shown blurred; sign up to generate at full resolution.
SFW
realistic-sharp-v1
SFW
realistic-sharp-v1
SFW
realistic-v2
18+ preview
realistic-sharp-v1
18+ preview
realistic-v2
Ready to generate?
Free credits on signup, no card required. Same API, same character primitive, every use case above.