flashlight/wav2letter

Facebook AI Research's Automatic Speech Recognition Toolkit

What it solves

wav2letter++ is designed to provide a high-performance framework for end-to-end automatic speech recognition (ASR), enabling the reproduction of various research papers and the use of pre-trained models for converting speech to text.

How it works

The project provides a set of recipes and pre-trained models that integrate with the Flashlight machine learning library. It supports various ASR architectures, including streaming ConvNets, sequence-to-sequence models with time-depth separable convolutions, and lexicon-free speech recognition.

Who it’s for

It is primarily intended for researchers and developers working on speech recognition, specifically those looking to reproduce academic results or implement state-of-the-art ASR architectures.

Highlights

  • Supports multiple research-backed ASR architectures including ConvNets and Seq2Seq.
  • Provides recipes for semi-supervised learning and self-training for speech recognition.
  • Includes pre-trained models for immediate use.
  • Integrated with the Flashlight library for high-performance execution.

Related

  • Project
    openspeech-team/openspeech

    A framework for building end-to-end automatic speech recognition systems, providing reference implementations of 20+ ASR models and recipes for multiple languages.

  • Project
    espnet/espnet

    An end-to-end speech processing toolkit that provides a unified framework for ASR, TTS, speech translation, and enhancement using PyTorch.

  • Dispatch
    Fine-Tuning Wav2Vec2-BERT for Low-Resource ASR

    Hugging Face demonstrates how to fine-tune Meta's Wav2Vec2-BERT model for Automatic Speech Recognition (ASR) in low-resource languages, achieving performance comparable to Whisper-large-v3 while being significantly faster and more resource-efficient.

  • Project
    QuentinFuxa/WhisperLiveKit

    An ultra-low-latency, self-hosted speech-to-text pipeline that enables real-time transcription and translation using advanced simultaneous speech research and multiple model backends.

  • Project
    AudarAI/Audar-ASR-V1

    A family of Arabic-first generative speech recognition models that provide state-of-the-art transcription for dialectal Arabic and code-switched speech.