Skip to content

MILS

Caption images, audio and video without training

Visit official site

What MILS is

MILS is described as an open source project from Facebook Research that shows LLMs handling visual and audio tasks without training.

What you can do with it

  • Generate descriptions for images
  • Generate descriptions for audio
  • Generate descriptions for video

Getting started

  • Create the conda environment from the environment file
  • Download the image, audio and video datasets
  • Set the paths and run the matching script

Pros & cons

✓ What we like

  • Free and open source with install guides
  • Multi GPU parallel processing is supported

! What to watch out for

  • Setup needs datasets and a conda environment
  • Paths are set in a paths configuration file

FAQ

Who released it?

The material names Facebook Research as the publisher.

What datasets are named?

The use cases name MS-COCO, Clotho and MSR-VTT.

Last reviewed: 2026-11-15

More LLM API platform tools

View all →

How we review