duixcom/Duix-Mobile

🚀 The best real-time interactive AI avatar(digital human) with on-premise deployment and <1.5 s latency.

What it solves

Duix Mobile provides a way for developers to create real-time, interactive AI avatars on mobile and embedded devices. It removes the dependency on cloud servers for the avatar's visual and interactive behavior, enabling low-latency, private, and lightweight interactions that feel more human-like.

How it works

It is an open-source SDK that can be deployed across iOS, Android, and other embedded screens. Developers integrate their own or third-party Large Language Models (LLM), Automatic Speech Recognition (ASR), and Text-to-Speech (TTS) services into the SDK to power the conversation. The system supports lip synchronization, streaming audio with barge-in capabilities, and local execution of core functions to ensure stability in poor network conditions.

Who it’s for

Mobile and embedded device developers building applications for intelligent customer service, virtual doctors, virtual lawyers, virtual companions, and virtual tutors.

Highlights

  • Ultra-Low Latency: Response times under 120ms on Snapdragon 8 Gen 2 SoC.
  • On-Device Deployment: Operates locally to reduce resource consumption and increase privacy.
  • Cross-Platform: One-click deployment for Android and iOS.
  • Human-like Interaction: Supports natural facial expressions, emotional cues, and streaming audio with interruption support.
  • Modular Design: Allows for fast customization of industry-specific solutions.

Related

  • Project
    CopilotKit/channels-sdk

    An SDK that connects AI agents to Slack, Microsoft Teams, and Discord, enabling native interactive UI and human-in-the-loop approvals.

  • Project
    alan-ai/alan-sdk-web

    An SDK for embedding an intelligent layer into web applications that enables the real-time generation of business logic and UI components.

  • Project
    livekit/python-sdks

    A Python SDK for integrating real-time video, audio, and data features into applications, enabling the creation of multi-modal AI and live streaming tools.

  • Project
    tetherto/qvac

    A cross-platform SDK and model provider for running LLMs, speech, vision, and generative AI locally on any device with built-in P2P inference delegation.

  • Project
    tuya/TuyaOpen

    A cross-platform C/C++ SDK and framework for building AI-agent hardware that integrates embedded devices with multimodal AI and leading LLMs via Tuya Cloud.