duix-avatar

Duix.Avatar is a free and open-source AI avatar project developed by Duix.com. It is a fully offline video synthesis tool designed for Windows and Ubuntu systems that clones a user's appearance and voice to create digital human videos driven by text and voice input. Core features include: - Precise appearance and voice cloning using AI algorithms to capture facial features and reproduce voice characteristics. - Text and voice-driven virtual avatars, using natural language processing to convert text to speech and drive avatar expressions and actions. - Efficient video synthesis with synchronized audio-video lip-syncing. - Multi-language support for eight languages: English, Japanese, Korean, Chinese, French, German, Arabic, and Spanish. Key advantages include fully offline operation for privacy protection, a user-friendly interface, and support for importing multiple models. The project uses Docker-based deployment with three main Docker images (fun-asr, fish-speech-ziming, and duix.avatar). Hardware requirements include an NVIDIA graphics card (recommended RTX 4070), 32GB RAM, and over 100GB of free disk space. It supports both Windows 10 and Ubuntu 22.04 installations. The project provides open APIs for model training, audio synthesis, and video synthesis, accessible locally through HTTP endpoints. It also offers a separate commercial API service for business users who need enterprise-level SLA assurance. The project acknowledges using fun-asr for ASR and fish-speech-ziming for TTS. It supports global free commercial use, with enterprises exceeding 100,000 users or $10 million annual revenue needing a commercial license agreement.