Jan 13, 2023 21:00:00

Design Voice, a generation model that allows users to design their own completely new synthesized voice

As image and sentence generation AI is booming,

Eleven Labs , a software company that produces dubbing tools using artificial intelligence and machine learning, is creating a speech synthesis model `` Design Voice '' that can design new synthetic voices from scratch. clarified.

This Voice Doesn't Exist - Generative Voice AI
https://blog.elevenlabs.io/enter-the-new-year-with-a-bang/

Eleven Labs is a company that develops dubbing tools for movies and audiobooks. This tool is unique in that it can automatically reread in another language while preserving the character of the original speaker's voice.

According to Eleven Labs, the idea of a new speech synthesis AI came up by unraveling the speech synthesis / speech duplication method used for the dubbing tool. Eleven Labs, which actually moved to development, seems to be pursuing a method of learning a dedicated model and infinitely creating new voices.

The model under development at the time of writing the article can set basic parameters to establish a new voice identity, such as gender, age, accent, pitch, and speaking style. It can generate any voice, so even if you set the same basic parameters, you can get a completely new voice that has never existed before.

Click the link below to play a sample voice generated by Design Voice.

・Talking style
·news
·conversation

Eleven Labs says that it is useful for things that need to prepare ``unique voices'' such as voice recordings for news and commercials, and things that require long voices such as storytelling and video games, due to the characteristic that they can be generated from scratch. Appeal.

In addition, Eleven Labs is also looking at the future where voice actors can sign license agreements, train their own voice models, and receive fees as compensation. We respect intellectual property rights and are committed to safeguarding our technology from being misused, as well as watermarking all audio so that it is instantly recognizable as Design Voice. It seems that they are also working on

In the future, we are also considering allowing users to duplicate their own voices and let them speak freely. It will make it easier for us to create works that require our voice.

Eleven Labs said, 'By using AI, flexible thinking and free design are possible from the early stages of game development, and in the case of news and audiobooks, it was not possible until now to cover the cost of recording. More content will be free to participate in more projects, and you will be able to immortalize your voice.'