One year after its open-source release, the TEN Framework has gained traction as a foundational tool for developers building real-time voice-based AI systems. Initially launched in 2024, the framework supports the development of voice agents capable of processing speech, generating responses, and understanding context in real time.
Supported by Agora and a fast-growing global developer community, TEN was created to address the technical challenges of building low-latency, multimodal AI applications. Over the past year, it has been adopted in diverse projects ranging from AI companions and language translators to customer support bots and interactive learning tools.
Key Advantages of TEN Framework
TEN Framework provides a production-ready framework for real-time, low-latency, multimodal AI.
- Real-time and production-ready: Built for stable, full-duplex audio streaming with millisecond-level latency.
- Extensible and vendor-neutral: Integrate easily with any customized LLM/STT/TTS service.
- Developer-friendly: Build with Python, Node.js, C++, or Go — or use the visual TMAN Designer for drag-and-drop development.
- Multimodal support: Combine voice, vision, and contextual signals for richer, more human-like interactions.
With TEN, developers can transition from prototype to production more efficiently, without being tied to a specific vendor or managing complex infrastructure behind the scenes.
Expanding the Voice AI Toolkit
Earlier this year, TEN introduced two open-source components that have become essential building blocks for the real-time conversational AI stack: TEN VAD and TEN Turn Detection (TTD).
TEN VAD — Lightning-Fast Voice Detection, Backed by Industry Experts
TEN VAD is a lightweight, high-performance Voice Activity Detection (VAD) system that minimizes latency and boosts transcription accuracy in real-time voice agent applications. Developers worldwide are already seeing tangible results.
TEN Turn Detection — Making Conversations Feel Natural and Effective
TEN Turn Detection (TTD) is TEN’s context-aware dialogue manager that helps AI engage in human-like, flowing conversation.
With over 98% accuracy in detecting natural turn-taking — knowing when to pause, listen, or speak — TTD enables smooth, interruptible dialogue and more natural user experiences.
By precisely identifying user intent and conversational flow, developers can build human-like voice agents that respond fluidly, even in fast-paced or emotionally nuanced scenarios.
From Tools to Ecosystem: Accelerate Conversational AI Development with Templates
To accelerate innovation, TEN launched more than 10 open-source voice agent templates, allowing developers to deploy agents — from voice AI assistants to real-time transcription or SIP integration — in minutes.
A Global, Developer-Driven Community
Over the past year, the TEN team organized developer meetups in San Francisco, Tokyo, Paris, Beijing, and Kyoto, bringing together hundreds of participants to exchange technical ideas and feedback. Insights from these events have contributed to improvements across the framework.
TEN also held its first online hackathon, which drew developers from multiple countries experimenting with new voice-agent applications such as real-time translation and conversational companions.
Defining the Next Era of Conversational AI
TEN Framework will continue to evolve as an open-source conversational AI framework built for real-time, scenario-driven voice interactions. It will further deepen collaboration with developers worldwide, enabling every idea to be rapidly transformed into production-ready applications.
Together with the global developer community, TEN Framework is shaping an open, supportive, and continuously evolving ecosystem for intelligent voice interaction.
Join the TEN Community
Explore the ecosystem, contribute, and start building the future of conversational AI today.
- Website: theten.ai
- GitHub: github.com/ten-framework/ten-framework
About TEN Framework
TEN Framework is an open-source ecosystem that empowers developers and enterprises to build, deploy and scale real-time multimodal conversational AI agents. With support for voice, vision, text and immersive avatar interactions, TEN delivers ultra-low latency, cross-platform performance and flexible extensibility. As part of its community-driven framework, TEN enables integrations across edge and cloud, multiple languages (C++, Go, Python and more), and plug-and-play extensions for rapid AI agent development. Visit theten.ai or explore the code on GitHub at github.com/ten-framework/ten-framework.
Media Contact
June Wang
developer@theten.ai
