Skip to content

LSLM

A dialogue model that listens while speaking

Visit official site

What LSLM is

The model is described as a dialogue system that improves natural interaction through full-duplex modelling. It listens while speaking and can be interrupted and answered in real time.

What you can do with it

  • Hold a full-duplex conversation
  • Generate speech with decoder-only TTS
  • Take live audio through a streaming encoder

Getting started

  • Integrate the model into a dialogue system
  • Configure fusion and interaction settings
  • Train for a specific dialogue scenario
  • Test duplex ability under noise

Pros & cons

✓ What we like

  • Streaming SSL encoder handles live audio input
  • Integration into existing dialogue systems is described as low impact

! What to watch out for

  • Fusion strategy and interaction settings must be configured
  • The model is trained for specific dialogue scenarios and instructions

FAQ

What does FDM stand for?

Full-duplex modelling, listening while speaking.

What fusion strategies are named?

Early fusion, middle fusion and late fusion.

Last reviewed: 2026-11-15

More AI audio tools tools

View all →

How we review