DUIX is an AI-powered platform for real-time digital human interaction, targeting developers seeking to integrate advanced AI capabilities like LLMs, ASR, and TTS into applications. It offers a low-cost, low-network-dependence solution for creating intelligent digital human agents across various industries, with a focus on ease of deployment on mobile and terminal screens.

How It Works

The platform leverages a suite of AI technologies including speech recognition (ASR), speech synthesis (TTS), natural language understanding (NLP), AIGC, and Large Language Models (LLM) to enable digital humans that can "hear, see, speak, and understand." This multimodal approach facilitates intelligent human-computer interaction, aiming for a seamless and responsive user experience.

Quick Start & Requirements

• Installation: The project provides SDKs for Android and iOS. Specific installation commands are detailed in the respective documentation.
• Prerequisites: Requires mobile development environments for Android and iOS. Downloadable local digital human models are available.
• Resources: Links to documentation for Android and iOS SDKs are provided.

Highlighted Details

• Supports one-click deployment on Android and iOS platforms.
• Offers a selection of downloadable digital human models for various use cases.
• Customization of digital avatars based on user videos is available as a paid service.
• Free avatar models are available for commercial use, with a license agreement required for enterprises exceeding specific user/revenue thresholds.

Maintenance & Community

• Community support is available via WeChat and Discord.
• Updates and demonstrations can be followed on Twitter.
• Contact email: james@duix.com.

Licensing & Compatibility

• Free avatar models are available for commercial use, with a commercial license agreement required for larger enterprises. Specific terms for commercial use should be clarified with the provider.
• Customization services are paid.

Limitations & Caveats

The API for controlling digital human actions is not currently supported, and streaming data for broadcasted WAV files is under modification. Callback methods for broadcast start/end are available as per SDK documentation.