Voicebox, a new AI-powered speech tool from Meta that it refers to as a “eraser” for audio editing

Voicebox is an AI model for speech-generation tasks, such as “editing, sampling, and stylizing” audio samples, that Meta just unveiled. It can create premium audio files and edit previously recorded recordings to cut out undesirable parts while preserving the overall look and feel of the footage. The AI tool can produce voice in six different languages and is multilingual.

According to Meta, in the future, AI solutions like Voicebox may make it possible for those who are visually impaired to hear written communications from other people in their own voices. Additionally, it might provide virtual assistants lifelike voices and aid in the editing and production of audio tracks for online content. Additional jobs that the new AI tool can do include the following:

Generating text-to-speech in the same manner: The most recent model can analyse a brief two-second audio sample and produce text-to-speech output in the same style.

Eraser for audio editing: It may restore that segment of the audio note by replacing misspelt words or filtering out noise, avoiding the need to re-record the voice. According to Meta in the Newsroom article, “you can, for instance, crop a speech segment that was interrupted by a dog barking and tell Voicebox to re-generate that segment – like a rubber for audio editing.”

Cross-language speech generation: After assessing a user’s speech in English, French, German, Spanish, Polish, or Portuguese, Voicebox may create “a reading of the text” (from the audio input) in the same manner. To put it simply, it may enable anyone to produce audio output in their voice in a foreign language.

Voicebox creates speech that is “more representative” of how people really speak in the six languages mentioned above since it is built on a vast quantity of data.

It’s important to note that Voicebox’s AI also learns from the untranscribed audio it gets as input. Voicebox is not available for public usage because to the dangers and potential abuse that come with it. Additionally, the code has not yet been made public.

Related Articles

Leave a Reply

Your email address will not be published.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

Back to top button