Polyfaced

Targeting Q3 2026 — join the waitlist

AI Avatar Coming Soon

Generate consistent AI headshots and full-body avatars across styles, lighting and outfits — with identity stability across an entire batch, not 10 random faces.

Get notified when AI Avatar launches

Be the first to try it. No spam — one email when it ships.

What we're building

AI Avatar is the next surface landing on Polyfaced after the v1 launch. It's built on identity-aware diffusion — meaning unlike vanilla text-to-image, the model maintains the same face across multiple generations rather than producing similar-but-different people every run. This is the difference between a usable avatar product and a curiosity. You'll be able to upload a reference (or describe a fresh persona in text), then generate a batch of variations covering different outfits, lighting setups, camera angles, and emotional tones — all reading as the same person. Under the hood, AI Avatar will run on a model we're still evaluating across the leading providers (Higgsfield, Hailuo, an OpenAI-tuned variant, and a Stability open-weight option). We won't ship a placeholder, so the exact engine lands when we're confident it meets the quality bar.

Use cases we're building for

Social media profile sets

Generate a consistent set of profile photos for Twitter, LinkedIn, Instagram, and Discord — same face, different lighting and outfit. Refresh quarterly without rebooking a photo shoot.

Founder and team page photos

For solo founders and small teams who want a polished look on landing pages and pitch decks without committing to a full studio shoot. Generate variations until you find the one.

Game characters and avatars

Generate a hero, then iterate on outfits, expressions, and biome backgrounds for a complete character sheet. Drop the consistent set into Unity, Unreal, or 2D animation pipelines.

Story illustration and character design

For illustrators and authors who need to maintain visual continuity for a character across an entire book, comic, or series. One reference, dozens of consistent variations.

Roadmap detail

What's locked: identity-aware diffusion engine, batch generation, the same prompt box pattern from /studio/video and /studio/image. What's still being evaluated: which underlying model wins on quality-to-cost, what aspect ratios the v1 release supports, whether full-body or headshot-only ships first. Sign up for the waitlist and we'll email when the engine selection finalizes — and a second time when the surface goes live.

Meanwhile, try our shipped tools