1. OpenAI has developed Voice Engine, allowing users to upload voice samples to generate synthetic copies responsibly.
2. The model powering Voice Engine has been in development for two years and is not trained on user data for privacy reasons.
3. OpenAI is taking steps to prevent Voice Engine from being misused, including watermarking clones and involving experts in risk assessment.
OpenAI is developing Voice Engine, an expansion of its text-to-speech API that allows users to create synthetic copies of voices from a 15-second sample. The company is aiming to ensure responsible deployment of the technology by understanding where it can be dangerous and implementing mitigations. The model powering Voice Engine has been used in projects like Spotify podcasts, ChatGPT, and other text-to-speech applications.
The training data for Voice Engine includes a mix of licensed and publicly available resources. OpenAI faces legal challenges regarding intellectual property as it uses copyrighted material to train its AI models. The company has agreements with some content providers and allows artists to opt-out of having their work used for training. Voice Engine is priced aggressively at $15 per one million characters, with no controls for adjusting voice tone or pitch.
Voice Engine could disrupt the voice actor industry by providing a cost-efficient alternative to human voice work. OpenAI is aiming to ensure ethical use of the technology by watermarking cloned voices and limiting access to a small group of developers. The company is also considering implementing a security mechanism to verify the awareness of users using the tool. As the preview stage for Voice Engine progresses, OpenAI plans to evaluate the technology’s safety issues and decide on its wider release based on feedback.
Overall, OpenAI is focused on ensuring the responsible and safe deployment of Voice Engine while exploring its potential impact on various industries, including voice acting and AI technology.