Meta has introduced a new AI model that can understand different languages during conversations and generate text on the fly. Learn what Meta's new AI is and how it works.
Meta Muse Voice Transcribe
Now, AI won't just be limited to converting voice to text, but will also be able to understand different languages during conversations. On September 1st, Meta introduced a new speech-to-text model called Muse Voice Transcribe, which can convert conversations to text on the fly. Remarkably, it can handle multiple languages within a single conversation.
According to Meta, this model supports over 70 languages, including five major Indian languages: Hindi, Tamil, Telugu, Malayalam, and Kannada. So, let's explore what Meta's new Muse Voice Transcribe is and how it works.
What is Meta's new Muse Voice Transcribe?
On September 1st, Meta launched a new AI speech-to-text model called Muse Voice Transcribe. It can convert people's conversations into text by listening to them simultaneously. It can even understand different languages spoken within the same conversation. According to Meta, Muse Voice Transcribe supports over 70 languages, including five Indian languages: Hindi, Tamil, Telugu, Malayalam, and Kannada.
What is special about Muse Voice Transcribe?
A key feature of Muse Voice Transcribe is its code-switching support. If a person switches from English to an Indian language or from one language to another during a conversation, the model is able to understand this. This doesn't require separate models or additional processing. It's believed this feature could be particularly useful in a diverse language-using country like India, where people often speak a mix of English and regional languages during conversations.
How does Muse Voice Transcribe work?
This model offers streaming transcription. This means you don't have to wait for the entire recording to finish. The text is generated as the person speaks. According to Meta, it can recognize the voices of individual speakers in recordings with more than 20 people and can handle recordings longer than an hour.
Accurate transcription with fast response
Muse Voice Transcribe determines how long to listen for each word. It responds quickly to simple words, while it may take a little longer to recognize difficult ones. According to Meta, as of September 1, 2026, this model ranks first on Artificial Analysis's streaming speech-to-text leaderboard.
How much does Muse Voice Transcribe cost?
Muse Voice Transcribe can be used through Meta's Model API. According to Meta, it is also being used for dictation in Meta AI for Mac and Muse Code. The API costs $3 for 1,000 audio minutes, which Meta says equates to approximately $0.18 per hour. In Indian rupees, this translates to approximately ₹285 per 1,000 audio minutes and approximately ₹16 per hour.
