Skip to content

LLM DeepASR incremental integration

sagatake edited this page Nov 8, 2024 · 4 revisions

A Full demo configuration is available at Greta - Mistral incremental - NVBG - MM - DeepGram(ASR) - Cereproc.yaml

365225992-b6427472-77f9-4f15-a342-21c2de07a9e6

This demo allows us to communicate with Mistral incrementally (sentence by sentence) through Greta; the ASR module automatically listens and sends the request to the Mistral module (see below). If you want to use functionality of automatic gesture generation, please modify the configuration refering here

ASR integration

  • A DeepASR module can be linked to an LLM module, in this case, the transcribed speech is automatically sent to the LLM module and processed by it
  • A DeepASR module can be linked to a Feedback module; in this case, if automatic speech is enabled, then the DeepASR module will automatically listen until Great starts talking and start listening again when Great is done talking.
  • The Feedback module can be replaced by any turn-tacking module

Getting started with Greta

Greta Architecture

Quick start

Advanced

Functionalities

Core functionality

Auxiliary functionalities

Preview functionality

Nothing to show here

Previous functionality (possibly it still works, but not supported anymore)

Clone this wiki locally