I moved my live avatar runtime to the backend — this finally makes it feel SaaS-capable
Quick build update on PhotoCall.
I’m building a live avatar system that turns a single photo into a reusable photorealistic avatar.
The original idea was simple: do the expensive work upfront during preprocessing, so runtime usage stays cheap.
But I had one big practical problem:
Even if the avatar looked good, the frontend/runtime experience still depended too much on the user’s device.
On stronger machines it worked well.
On weaker devices, browser-side video handling could get choppy or unstable.
That is a bad foundation for a SaaS product.
So I moved the live avatar runtime fully backend-side.
This feels like a much bigger milestone than adding another feature, because it makes the system much closer to something other products could actually use.
The direction now is:
plug-and-play widget for other products
later API integration for builders who need more control
It’s still not “perfect lip sync at any cost”.
The tradeoff is intentional: photorealistic, reusable, cheap to run, and reliable enough to embed into real products.
Curious how other builders would think about the first integration model:
Would you start with a simple plug-and-play widget, or go directly toward an API?
