新的人工智慧語音工具經過訓練可以模仿英國地方口音

A British technology company called Synthesia has developed a new artificial intelligence tool that can accurately copy a wide range of British regional accents. This tool, named Express-Voice, was created because most AI voice systems are trained using voices from North America or southern England, which makes many artificial voices sound similar and not very representative of the UK's diverse accents. To solve this problem, Synthesia spent a year building its own database of voices from across the UK. They recorded people in studios and also collected voice samples from the internet, making sure to include as many different regional accents as possible.

Express-Voice can either clone a real person's voice or create a completely new synthetic voice. These voices can be used in things like training videos, customer support, and business presentations. According to Synthesia, their customers have been asking for more accurate regional voices because they want their own accents to be preserved in digital content. Youssef Alami Mejjati, the head of research at Synthesia, explained that people want their voices to sound like themselves, including their unique accents. He also mentioned that French-speaking customers have noticed that AI-generated French voices often sound Canadian rather than French from France. This is because most companies that make these AI tools are based in North America and use data from their own regions, which creates a bias in the voices.

Some accents are much harder for AI to copy, especially the rare ones, because there is not enough recorded material to train the system. There have also been reports that AI products like smart speakers have trouble understanding people with different accents. In the United States, a company called Sanas is working on a different kind of tool that changes the way people sound in call centers, making their accents more neutral. The goal is to help workers with Indian or Filipino accents be better understood by callers and to reduce discrimination based on accent. However, there are concerns that as technology like this becomes more common, it could lead to the loss of unique languages and dialects. According to UNESCO, almost half of the world's seven thousand languages are endangered, and only a small number are supported by major AI systems. Experts warn that as AI voice technology improves, it could also be used by scammers to trick people. Synthesia says their product will not be free and will have protections against hate speech and explicit content, but there are already many free voice-cloning tools online that are not as secure. This rapid development in AI voice technology is raising new questions about safety and the future of language diversity.