← Back to all projects

TEN Framework

A real-time multimodal conversational stack — voice, telephony, avatars — under Apache terms with conditions

OfficialSource-available
Stars
11.2k
Forks
1.4k
Open issues
255
Last commit
30 Oct 2026

What is TEN Framework?

It covers the parts of a spoken agent that sit outside the model and usually get assembled by hand: detecting when someone is speaking, deciding when a turn has ended, separating speakers, connecting to the phone network, driving a lip-synced avatar, even reaching an embedded device. Read the licence before building on it. Apache 2.0 is qualified by additional conditions from Agora that prohibit hosting the framework on end-user devices, mobile terminals included, and prohibit deploying it in a way that competes with Agora's own offerings.

What can you do with TEN Framework?

  • Start from a working voice assistant — Example agents run locally under Docker Compose, so the first working spoken conversation is a matter of supplying keys and starting containers rather than an integration project.
  • Get turn-taking right — Voice activity detection and turn detection are separate components of the ecosystem rather than an afterthought, and they are where the perceived quality of a spoken agent is usually decided.
  • Answer and place phone calls — A SIP example connects the agent to the telephone network, which is a different problem from a browser microphone and the one that most business use cases actually need.
  • Give the agent a face — A lip-synced avatar example drives Live2D characters, alongside separate examples for speaker diarisation and live transcription.
  • Reach an embedded device — An integration guide covers running against ESP32 hardware, so the target can be a small physical device rather than a workstation or a browser tab.

Before you choose TEN Framework

  • The Apache grant carries additional conditions: you may not host the framework on end-user devices including mobile terminals, nor deploy it in a way that competes with Agora's offerings.
  • The example agents expect an Agora application ID and certificate alongside the model, speech-recognition and text-to-speech keys, so the quick start depends on a commercial account rather than open components alone.

Star history

21 Aug to 30 Oct · +145

11.1k11.2k

Frequently asked questions

Is TEN Framework free for commercial use?

TEN Framework is published under Apache-2.0 with additional conditions, a source-available licence rather than OSI-approved open source — read its terms before relying on it commercially.

How can TEN Framework be deployed?

TEN Framework is available as Self-hosted.