OpenAI is quietly launching its highly anticipated voice mode for ChatGPT, starting with a select group of ChatGPT Plus subscribers. This move marks a significant step forward in AI interaction, offering a more natural and intuitive way to engage with the popular chatbot.
Announced at the GPT-4o launch event in May, the voice mode initially sparked controversy due to its resemblance to Scarlett Johansson’s voice, prompting concerns about potential misuse and ethical implications. OpenAI responded by delaying the rollout to address these concerns and refine its safety protocols. “We’ve made it so that ChatGPT cannot impersonate other people’s voices, both individuals and public figures, and will block outputs that differ from one of these preset voices,” said OpenAI spokesperson Taya Christianson.
The enhanced voice mode boasts impressive capabilities, including low latency for real-time conversations and the ability to interrupt the AI seamlessly. During the May demonstration, OpenAI employees showcased these features, engaging in dynamic back-and-forth exchanges with the chatbot.
More Read
To ensure a responsible launch, OpenAI has implemented several safety measures. The voice mode is currently limited to four preset voices, all created with the consent and involvement of voice actors. This restriction aims to prevent voice cloning and unauthorized impersonations. Additionally, OpenAI has incorporated filters to detect and block requests for inappropriate or copyrighted content, further mitigating potential risks.
The company emphasizes its commitment to user safety and responsible AI development. Over 100 external “red teamers” were involved in testing the voice model’s capabilities, identifying vulnerabilities, and refining its security measures. OpenAI also implemented new filters to prevent the generation of copyrighted audio content, addressing concerns raised during the initial unveiling.
While the voice mode is currently exclusive to ChatGPT Plus subscribers, OpenAI plans to expand its availability to all users in the fall.
