📖 Documentation [android] [ios] 💬 WeChat (微信) 😄 Discord 📑 FAQ
DUIX, short for “Dialogue User Interface System”, is an AI-powered digital human interaction platform created by Silicon-based Intelligence. By open-sourcing the capabilities of digital human interaction, developers can easily integrate large-scale models, automatic speech recognition (ASR), and text-to-speech (TTS) capabilities, enabling real-time interaction with digital humans. It supports one-click deployment on multiple platforms including Android and iOS, allowing each developer to effortlessly create intelligent and personalized digital human agents, and apply them in various industries.
- Low deployment cost: No need for customers to provide technical teams for cooperation, supports low-cost rapid deployment on various terminals and large screens.
- Low network dependence: Suitable for virtual assistant self-service in scenarios such as subways, banks, and government affairs.
- Diverse functions: Can meet the diverse needs of video, media, customer service, finance, radio, and television in multiple industries according to customer requirements.
- Provides customized AI anchors and intelligent customer service for multi-scene image rental.
- Exclusive image customization: Supports the customization of exclusive virtual assistant images, with options for low-cost or deep image generation.
- Broadcast content customization: Supports the customization of exclusive broadcast content, applied in training, broadcasting, and other scenarios.
- Real-time interactive Q&A: Supports real-time dialogue, can also customize exclusive Q&A databases to meet consulting inquiries, voice chat, virtual companionship, and vertical scene customer service Q&A.
duix-android: android demo
duix-ios: ios demo
For android, refer to README_en.md
For ios, refer to GJLocalDigitalSDK_en.md
We offer a selection of digital human models for download and use. We will update the local model packages from time to time so that you can download and utilize the latest models. Below is the list of currently available local model packages:
 |
 |
 |
|---|---|---|
| Eric Download |
Zi Xuan Download |
Ming Xuan Download |
 |
 |
 |
 |
 |
|---|---|---|---|---|
| Sophie Download |
Mu Rong Xiao Download |
Cold Flame Download |
Amelia Download |
Zhao Ya Download |
 |
 |
 |
 |
 |
|---|---|---|---|---|
| Yi Yao Download |
Xin Yan Download |
Xiao Xuan Download |
Si Yao Download |
Shi Ya Download |
 |
|---|
| Dear Sister Download |
Please download the model files according to your needs. We recommend that you carefully read the accompanying documentation after downloading to ensure the correct installation and use of the models.
https://apps.apple.com/us/app/duix-your-ai-companion/id6451088879
Answer: Digital human customization is supported, and it is based on videos. You can customize a unique digital human for use in the SDK. However, customized digital humans are a paid service; you can contact the customer service email for more information.
Answer: To customize a silicon-based digital human, you need to shoot a 3-5 minute real-life video with the person speaking on camera. The specific poses and content of the speech can be determined based on the actual usage scenarios. For example, if it is used to produce legal consultation videos, you can choose to appear in formal attire and speak about legal topics, creating a consistent overall situation.
Answer: We offer digital human customization配套 with the SDK open-source interface, priced at 9800 yuan/set, including image + voice. For more customization needs, please contact the customer service email.
Answer: Currently, the open-source version mainly provides public models, and customization is available if needed
Answer: The training service currently only supports deployment on our internal servers for calls
Answer: The streaming driver is currently under改造 optimization and is not yet supported;
Answer: The callback methods for the start and end of the broadcast can be found in the SDK documentation
Answer: The digital human action API is not yet supported
Answer: Instructions for replacing downloaded files can be found in the code
- 3.0.4: Fixed an issue where some devices’ default gl float low precision caused the avatar to not display properly.
- 3.0.3: Optimized local rendering
- We have drawn on wenet for audio features.