Skip to main content
Welcome to the SALT framework documentation. SALT is the core toolkit for building, training, and deploying African language models.

​
Why SALT?

SALT streamlines the process of working with low-resource languages by providing:
  • Standardized Datasets: Easy auditing and loading of text and speech data.
  • Pre-trained Models: State-of-the-art checkpoints for NMT, ASR, and TTS.
  • Training Pipelines: Tools to fine-tune models on your own data.

​
Explore the Framework

Quick Start

Get up and running with SALT in minutes.

Installation

Step-by-step setup guide.

Datasets

Explore the text and speech catalogs.

Training

Train your own custom models.

​
Core Components

The documentation is organized by the main pillars of the project:
  • Models: Machine translation, speech recognition, and synthesis.
  • Pipelines: End-to-end workflows for data processing and model training.
  • Diarization: Advanced speaker identification tools.