Skip to content

百聆

An open-source voice assistant you can run without a GPU

Visit official site

What 百聆 is

The app is described as an open-source assistant for natural spoken dialogue, built on ASR, VAD, LLM and TTS. It is presented as running without a GPU on edge devices and low-resource environments, with support for interrupting a reply.

What you can do with it

  • Hold a spoken conversation and interrupt the assistant mid-reply
  • Deploy the assistant locally without a GPU
  • Replace or upgrade the ASR, VAD, LLM and TTS modules on their own
  • Recall user preferences and past conversations across turns

Application scenarios

  • Ask about the weather in Hangzhou by voice and get the forecast back
  • Generate IELTS speaking questions and a practice dialogue
  • Set a timed reminder to drink water at 8 every morning

Getting started

  • Clone the repository from the project GitHub page
  • Install the dependencies with pip install -r requirements.txt
  • Edit config/config.yaml to configure the ASR and LLM settings
  • Start the optional back end by running python server.py inside server

Pros & cons

✓ What we like

  • Modular parts can be replaced or upgraded as needed
  • Voice requests can trigger external tools and task reminders

! What to watch out for

  • Local setup requires installing dependencies such as requirements.txt
  • ASR and LLM settings must be configured in config/config.yaml

FAQ

Does it need a GPU?

The material states the assistant runs locally without a GPU and suits low-resource settings.

What modules make up the assistant?

ASR, VAD, LLM and TTS are listed as independent modules that can be replaced.

Last reviewed: 2026-11-15

More AI audio tools tools

View all →

How we review