Facebook AI Research's C++ toolkit for building fast end-to-end automatic speech recognition systems.
An end-to-end speech-to-text engine from Facebook AI Research, now consolidated into Flashlight.
wav2letter is an open-source automatic speech recognition toolkit built in C++ by Facebook AI Research. It provides pre-trained models and recipes for transcription research. If you need an out-of-the-box executive assistant tool, this developer library requires engineering work to build into an application.
The toolkit translates spoken audio into text using end-to-end deep learning models. It includes recipes for reproducing published research benchmarks along with pre-trained acoustic models. Ongoing development has migrated into Flashlight's speech recognition module.
This toolkit suits speech recognition developers and C++ engineers building custom transcription pipelines. It does not suit non-technical productivity professionals seeking an instant dictation tool or meeting recorder.
Translates raw audio into text transcripts using deep learning architectures built in C++.
Includes scripts and pre-trained models to reproduce published speech recognition research results.
Offers native execution capabilities suited for direct integration into performance-focused applications.
Connects with the Flashlight application framework where ongoing engine development now resides.
The software is free to download and modify. The only cost comes from your own GPU hardware, server infrastructure, and development time.
877 open remote AI-training roles, updated November 16, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
No, it is a technical C++ engine for developers and requires custom code to handle audio input and user interfaces.
No, new development has moved to Flashlight's ASR application, though legacy tags remain accessible here.
Running inference or training requires Linux infrastructure with supported GPU hardware and C++ libraries.
See Zekai first in Google