ten-framework
Open-source framework for conversational voice AI agents
What is ten-framework?
Open-source framework for conversational voice AI agents
Overview of ten-framework
What it is
TEN is an open-source framework for real-time multimodal conversational AI. It is designed for building conversational voice AI agents. The TEN ecosystem includes the TEN Framework, Agent Examples, VAD, Turn Detection, and Portal.
Key features
- Real-time multimodal conversational AI framework
- Supports both RTC and WebSocket connections
- Extensible with Memory, VAD, Turn Detection, and other extensions
- Includes agent examples such as Multi-Purpose Voice Assistant, Doodler, Speaker Diarization, Lip Sync Avatars, SIP Call, and Transcription
- Can run on Espressif ESP32-S3 Korvo V3 development board
- Provides TMAN Designer for customizing agent examples
- Supports deployment with Docker and other cloud services (Vercel, Netlify, etc.)
Use cases
- Building low-latency, high-quality real-time voice assistants
- Creating interactive doodle boards that turn spoken prompts into sketches
- Real-time speaker diarization and labeling
- Lip-sync avatars with multiple vendor support
- SIP-based phone calls powered by TEN
- Audio transcription tools
- Integrating LLM-powered communication with hardware (ESP32-S3)
Getting started
See the repository README for detailed prerequisites and setup instructions. The quick start involves cloning the repo, setting environment variables (Agora App ID, OpenAI API key, Deepgram ASR, ElevenLabs TTS), and running Docker Compose.
cd ai_agents
cp ./.env.example ./.env
docker compose up -d
docker exec -it ten_agent_dev bash
cd agents/examples/voice-assistant
task install
task run
License and pricing
The license is not stated (NOASSERTION). The project is listed as open-source with pricing type "open_source".
Based on the project's public README and repository metadata.