What becomes possible when live avatars are cheap?

Live avatars are usually shown as impressive demos.

But the more interesting question is what becomes possible when they are cheap and reusable enough to become part of real products.

That is the focus behind PhotoCall: one portrait photo becomes a reusable live avatar identity that can appear across different conversations, roles, scenes, and product flows.

The core strengths are:

• one-photo avatar creation

• reproducible character identity

• very low runtime video cost

• no GPU needed during conversations

• shareable links and simple embedding

• useful for both B2C products and builder workflows

The approach is intentionally different from full real-time video generation. Most of the video work happens once during preprocessing, and the conversation reuses avatar states at runtime.

That keeps the economics attractive, but it also creates trade-offs: avatar creation is not instant yet, and lip sync is still an area I’m actively improving.

The bet is simple: if realistic avatars become cheap enough to use repeatedly, they stop being just demos and start becoming a product layer.

Curious where people would use this first: onboarding, tutors, support, interview practice, product demos, companions — or something completely different?