- tech360.tv
Meta Releases AI Model for Translating Speech in Dozens of Languages
Updated: Jan 8
[Edited] Meta Platforms, the parent company of Facebook, has unveiled an AI model capable of translating and transcribing speech in multiple languages.
The SeamlessM4T model, released by Meta on Tuesday, has the potential to facilitate real-time communication across language barriers.
According to a blog post by the company, the SeamlessM4T model can support translations between text and speech in nearly 100 languages. It also offers full speech-to-speech translation for 35 languages, combining functionalities that were previously available only in separate models.
CEO Mark Zuckerberg envisions these tools enabling interactions between users from around the world in the metaverse, a collection of interconnected virtual worlds that Meta is heavily investing in for the company's future.
Meta has made the model available to the public for non-commercial use, as stated in the blog post.
This year, the social media giant has released several AI models, most of which are free. One notable release is the Llama language model, which poses a significant challenge to proprietary models offered by Microsoft-backed OpenAI and Google.
Zuckerberg believes that an open AI ecosystem works to Meta's advantage, as the company benefits more from crowd-sourcing the creation of consumer-facing tools for its social platforms rather than charging for access to the models.
However, Meta faces legal questions similar to those encountered by the rest of the industry regarding the training data used to develop its models.
In July, comedian Sarah Silverman and two other authors filed copyright infringement lawsuits against both Meta and OpenAI, accusing the companies of using their books as training data without permission.
Regarding the SeamlessM4T model, Meta researchers stated in a research paper that they collected audio training data from 4 million hours of "raw audio originating from a publicly available repository of crawled web data." However, they did not specify which repository was used.
The text data used in the model was sourced from datasets created last year, which extracted content from Wikipedia and associated websites, according to the research paper.
The release of the SeamlessM4T model by Meta showcases the company's commitment to advancing AI technology and its potential to bridge language barriers. By providing access to this AI model, Meta aims to foster global communication and enhance user experiences within its platforms.
Meta Platforms has released the SeamlessM4T AI model for translating and transcribing speech in multiple languages.
The model supports translations between text and speech in nearly 100 languages and offers full speech-to-speech translation for 35 languages.
CEO Mark Zuckerberg sees these tools as crucial for facilitating interactions in the metaverse.
Meta has made the model available for non-commercial use.
Source: REUTERS