Tavus Builds Avatar Video Infrastructure
Synthesia
Tavus is betting that avatar video becomes infrastructure, not destination software. Instead of selling a team a place to log in, write a script, and export a finished training or marketing video, it sells developers endpoints they can wire into their own products, so a CRM, support tool, or commerce app can generate avatar video inside an existing workflow. That shifts distribution from direct SaaS sales to platform embedding, and it makes Tavus look more like Twilio or Mux than Synthesia.
-
Synthesia has gone the opposite direction, bundling avatar generation with screen recording, script editing, translation, hosting, analytics, and publishing. That all in one workflow is why it scaled to about $100M ARR by March 2025 and to about $145.9M by September 2025, with 70% of revenue tied to enterprise deals.
-
Tavus monetizes on usage volume of its APIs, which fits products that want avatar video as a feature. A sales tool can auto generate one personalized intro per lead, or a support product can spin up a face on top of an answer flow, without asking users to leave the host app and learn a separate studio.
-
The broader market is splitting in two. One lane is application companies like Synthesia and HeyGen, which own the whole video workflow. The other is API providers like Tavus and D-ID, which supply avatar capability to larger software platforms that want to bundle AI video into an existing product suite.
As avatar generation gets cheaper and easier to copy, more value will move to whoever controls distribution and workflow. That favors platform companies embedding video natively, which is exactly the world Tavus is building for, while pushing Synthesia to keep moving up stack into the full system where enterprise teams create, manage, and publish business video.