Meta releases voice AI model Voicebox to help virtual assistants communicate with NPCs

2023-06-21 05:17:13 Author: Lao Wang

Recently, Meta launched an artificial intelligence voice model called Voicebox. Compared to models that typically focus on text and images, Voicebox creates voice messages for replies. According to reports, this model can accurately identify audio details and timbre in just 2 seconds of audio samples, and convert the text results into speech output. Currently, Voicebox supports English, French, German and Spanish. Voicebox can fill in the missing parts based on the content before and after the voice clip.

This technology can provide natural and realistic voice effects for virtual assistants or NPCs in the Metaverse. Voicebox can assist people with damaged vocal cords to achieve barrier-free functions to a certain extent. However, Voicebox is still in the research and development stage. Meta said that such artificial intelligence technology can be potentially harmful in terms of false forgery, so the company is working hard to find ways to effectively distinguish between real speech and Voicebox-generated audio. The model will not be made publicly available until a solution is found.

The above is the detailed content of Meta releases voice AI model Voicebox to help virtual assistants communicate with NPCs. For more information, please follow other related articles on the PHP Chinese website!