Pre-Conference Session

Unleashing the Power of SeamlessM4T

Scaling Speech Translation for Enhanced Communication

04 October 2023, 13:00 to 17:00

Little America Hotel, Salt Lake City (USA)

About the Tutorial

Recent advancements in deep learning have revolutionized speech technology, enabling applications like voice assistants, language interpretation services, transcription services and many more. By breaking down language barriers and facilitating seamless communication, speech-to-speech and speech-to-text translation open new doors for global collaboration, improve accessibility, and enhance cross-cultural interactions. 

  • What will you learn?

    Building practical knowledge of how these models work

    Providing various strategies for effectively integrating them into your services 

    Optimization techniques to enhance data utilization and model performance, especially in low compute scenarios.

    Join us in this tutorial to stay up to date with the latest developments in speech technology and unlock the full potential of scaling speech translation, enabling effective communication in an interconnected world.

Sravya Popuri
Research Engineer
Meta AI
Changhan Wang
Research Engineer
Meta AI
Maha Elbayad
Research Scientist
Meta AI
Who should attend

This tutorial aims to empower language service providers (LSPs) and practitioners to tap into the cutting-edge advancements in speech translation models.



Session Title


1:00 pm

Setting the Stage:

Workshop Introduction and Overview

Sravya Popuri

1:15 pm

Exploring the Landscape of Speech Translation in localization : An Insight into seamlessM4T and its applications

- Landscape components: model builders, prompting environments, data pipelines, leaderboards.

- Ecosystems (OpenAI & Microsoft, Nvidia, Meta, Hugging Face, others).

- Open-source models to adapt.

Sravya Popuri

1:35 pm

A Behind-the-Scenes Look : Unveiling the training process of SeamlessM4T

- SeamlessAlign

- Deep dive on model architecture 

- Evaluation framework

Changhan Wang and Maha Elbayad

2:20 pm

Coffee Break

2:30 pm

Interactive session 

- Hands-on tutorial to run inference and finetune

- Q&A

Changhan Wang, Maha Elbayad and Sravya Popuri

4:45 pm

Closing remarks

Changhan Wang / Maha Elbayad

Registration & Location

The workshop will be held at the Little America Hotel in Salt Lake City, on Wednesday 4 October 2023 (the pre-conference day of the TAUS Annual Conference).

Get in touch to find out more about group discounts.

Note that Pre-conference Sessions are not included in the TAUS Annual Conference registration fee.

Pre-Conference Session

Unleashing the Power of SeamlessM4T

Registration for 1 person.

€ 380

Annual Conference

The TAUS Annual Conference is the place where key globalization and language stakeholders from the tech industry, global enterprises and their solution providers meet to benchmark strategies.

Pre-Conference Session

GenAI in Localization

An intensive introduction to generative AI for localization executives. Learn how to implement ChatGPT and its counterparts, and take the first step to transform your localization team into a center of excellence in language models. Organized by CustomMT.

Pre-Conference Session

Quality Estimation Workshop

A joint initiative between the main providers of Quality Estimation or Quality Prediction, TAUS and ModelFront.  Participants in this workshop will learn firsthand from technical leaders in these companies and their customers about the sustainable approach and real-world implementations.