The innovative Speech AI from Meta can comprehend an impressive assortment of over 4,000 languages.
In Brief
This new MMS initiative by Meta boasts the remarkable feature of recognizing a staggering collection of languages exceeding 4,000.
Aligned with its commitment to advancing AI transparency, Meta is making its code and models publicly available to support the preservation of linguistic heritage.
Through this initiative, Meta aims to contribute, even in a small way, to the preservation of the rich tapestry of languages found around the world.
Meta has created This AI language model refreshingly diverges from the likes of ChatGPT, focusing on open-source development. MMS project The MMS project not only intends to safeguard linguistic variety but also encourages further research, capable of recognizing over 4,000 spoken dialects and generating vocal output in upwards of 1,100 different tongues. Meta has... publicly released The organization has released its models and code today to further its ambitious objectives.
"Our goal is to make our developments and code accessible to inspire others within the research community to enhance our work,\" stated Meta. \"Through this undertaking, we aspire to safeguard the incredible linguistic diversity our world offers.\"

Training speech recognition and text-to-speech systems without pre-existing transcription labels can be quite challenging. These labels are crucial for... machine learning These tools aim to accurately classify and interpret data. Nevertheless, for many languages facing extinction in the near future, \"this data simply does not exist,\" as outlined by Meta.
In a novel approach, Meta utilized audio recordings from religious texts to gather necessary data. The company explained, \"We turned to translations of well-known religious texts like the Bible, which have been extensively explored for language translation studies across various languages, due to their abundant translations.\" They sourced audio files of different languages being read from publicly available translations. This way, researchers at Meta were able to capture audio data from over 4,000 languages. model .
At first glance, this strategy might appear to create a bias towards Christian perspectives within the AI model. Yet, before dismissing this notion, it’s important to consider the methodology: The researchers utilized a connectionist CTC temporal classification approach (also referred to as sequence-to-sequence or sequence-type models), which surprisingly demands less computational power than many traditional methods for speech association. Meta asserts that this framework did not lead to a male bias, despite the majority of the recordings being made by male voices. large language models Meta employed wav2vec 2.0, a unique model focused on self-supervised learning of speech representations, to enhance data usability. This self-supervised technique they derived from unlabeled input produced outstanding results. According to Meta, multilingual speech models achieved remarkable performance, surpassing existing frameworks by covering tenfold more languages.
While Meta’s new speech-to-text models are not infallible, they may occasionally produce inaccuracies in translating words or phrases, which could lead to potentially offensive or incorrect interpretations, as the company noted. The responsible evolution of AI technology hinges on collaboration within the AI field. alignment model By making MMS available for open-source exploration, Meta envisions curbing the decline of language use. In this ideal landscape, technologies such as assistive devices, TTS, and even immersive tech like virtual and augmented reality can empower individuals to learn and communicate in their native tongues. The company expressed, \"We dream of a future where technology serves to invigorate language fidelity, enabling people to interact and harness information in their chosen dialects.\" Whisper The financial report for Q1 2023 brought unexpected news: despite recent restructuring moves, the company pleasantly surprised investors with a noteworthy upswing in sales, leading to a 12% rise in shares this Wednesday.
A recent study revealed that 4,000 participants engaged in AI-guided psychotherapy without their knowledge.
Transforming disruption into opportunity: How Stage Meta’s Teleport Plaque is set to reshape the Metaverse inhabitants' experiences.
- Recently, Meta has announced The FTC attempts to impede Meta’s VR aspirations via legal action.
Read more related articles:
- Please be aware that the content shared on this page should not be construed as legal, tax, investment, financial, or any other type of advice. Ensure you only invest what you can afford to lose and consider seeking independent financial counsel if uncertain. For more information, we recommend checking the terms and conditions alongside the help and support resources provided by the issuer or advertiser. MetaversePost strives to deliver precise and impartial reporting; however, fluctuations in market conditions may occur without notice.
- Damir leads the team as product manager and editor at Metaverse Post, focusing on areas such as AI/ML, AGI, LLMs, Metaverse, and Web3 topics. His contributions engage a substantial audience of over a million users monthly. With a decade’s worth of expertise in SEO and digital marketing, he is a recognized name in the industry, with mentions in outlets like Mashable, Wired, Cointelegraph, The New Yorker, Inside.com, Entrepreneur, BeInCrypto, and more. As a digital nomad, he travels across the UAE, Turkey, Russia, and the CIS regions. Damir holds a bachelor’s degree in physics, which he believes equips him with essential critical thinking abilities to thrive in the ever-evolving online landscape.
- Blum commemorates the one-year milestone with accolades for ‘Best GameFi App’ and ‘Best Trading App’ at Blockchain Forum 2025.
Disclaimer
In line with the Trust Project guidelines Addressing DeFi fragmentation: An insight into how Omniston enhances liquidity on the TON network.