Omnilingual ASR
Photo credit: Meta

Meta’s Fundamental AI Research (FAIR) team is introducing Omnilingual ASR, a suite of models that provide automatic speech recognition for over 1,600 languages, including 500 low-resource languages that have never been transcribed by AI before.

The company says most current ASR systems focus on a limited set of high-resource languages, which exacerbates the digital divide. This new system is a significant step toward delivering a truly universal transcription system.

Omnilingual ASR introduces an “LLM-ASR” model that uses an LLM-style transformer decoder. This system achieves state-of-the-art performance, with character error rates below 10 per cent for 78 per cent of the languages it covers.

New languages with minimal data

A key feature of the new framework is its ability to learn new languages with minimal data. Meta says this shifts the paradigm for adding languages, as users can provide just a “handful” of paired audio-text samples to get usable transcription quality. This in-context learning capability removes the need for large-scale training data or access to high-end compute.

Alongside the models, Meta is open-sourcing Omnilingual wav2vec 2.0, a new 7B parameter self-supervised speech representation model, to be used for other speech-related tasks. The company is also releasing the Omnilingual ASR Corpus, a collection of transcribed speech in 350 underserved languages, curated in collaboration with global partners, including Mozilla Foundation’s Common Voice.

The models are being released under a permissive Apache 2.0 license in a range of sizes, from lightweight 300M versions for on-device use to the 7B models that offer top-tier accuracy.

Leave a Reply

Your email address will not be published. Required fields are marked *

You May Also Like

Massive AI study uncovers the secret GLP-1 side effects hidden on Reddit

Millions of patients are flocking to GLP-1 weight loss injections, but artificial…

Alarming new US survey shows half of patients rely on AI for medical choices

Across the United States, a dangerous new trend is emerging. Millions of…

One in four Americans now consult AI chatbots for medical advice

Millions of desperate patients are quietly abandoning the waiting room for a…

Global gambling firms rush to adopt AI despite severe lack of safety controls

The global gambling industry is racing to integrate artificial intelligence into its…

Why digital tears and online outrage fail to win modern political arguments

Scrolling through your social media feed today often feels like navigating a…

Tracking how war and energy policies dimmed night lights of Europe

While human civilisation is glowing brighter than ever before, the lights across…