EdgePersona creates a completely offline localized intelligent digital human system

By /

EdgePersona is an open source project that aims to create a fully localized intelligent digital human system. The system is lightweight and efficient in design, has low hardware requirements, and is suitable for running on ordinary notebooks to ensure user privacy.

project information

The main features of EdgePersona include:

  • Fully localized operation: All data processing is done locally, without the need for a network connection, ensuring data privacy.
  • multi-modal interaction: Support voice dialogue, expression control and synchronous response to body movements to achieve natural human-computer interaction.
  • Low hardware requirements: Measured in real time on a notebook equipped with NVIDIA 3060 graphics card.
  • Deep character customization: Customize character, movements and voice styles through YAML profiles.
  • Multiple model support: Compatible with mainstream model formats such as Olama, HuggingFace, and GGUF.

system components

EdgePersona’s system architecture includes the following main modules:

  • Speech recognition (ASR): Convert user voice input into text.
  • Voice Activity Detection (VAD): Recognize the beginning and end of speech to improve recognition accuracy.
  • Big Language Model (LLM): Process text input and generate response content.
  • Text To Speech (TTS): Convert text responses to speech output.
  • memory module: Store and manage conversation history and realize context correlation.
  • robot control: Manage the expressions and movements of digital people to achieve a natural interactive experience.

installation and operation

environmental requirements：

  • NVIDIA graphics card (recommended VRAM ≥ 6GB)
  • Python 3.11.11 and above
  • Supports Windows, Linux and macOS (M-series chips require Metal acceleration)

installation steps：

  1. Clone warehouse:

    git clone https://github.com/zc-maker/EdgePersona.git

  2. Installation dependencies:

    pip install -r requirements.txt

  3. Run the main program:

    python main.py

application scenarios

EdgePersona works in a variety of scenarios, including:

  • Personal assistant: A smart assistant that runs locally to protect user privacy.
  • Education and training: Create interactive teaching roles to enhance the learning experience.
  • Entertainment and games: Develop virtual characters with personalized characteristics to enhance user engagement.

For more information and detailed documentation, please visit the project homepage:

Github：https://github.com/zc-maker/EdgePersona

Oil tubing:

Scroll to Top