New Delhi: OpenAI, the pioneering artificial intelligence startup, is introducing a feature in its ChatGPT app that enables the chatbot to not only understand and respond to spoken questions and commands but also engage in conversation using its own synthesized speech.
In the next two weeks, users of the chatbot app will have the option to select from five different personas, such as “Juniper,” “Breeze,” and “Ember,” to customize the voice of ChatGPT. This means that ChatGPT will generate audio content in the chosen persona’s voice, including tasks like reading AI-generated bedtime stories aloud. This feature will be available to subscribers of OpenAI’s ChatGPT Plus service, priced at $20 per month, as well as enterprise users.
OpenAI introduced the ChatGPT app in May, initially enabling users to communicate with the bot via voice-to-text. Now, with the addition of an audio response feature, OpenAI aims to enhance the conversational experience, making it feel more human. This move is intended to attract users who prefer on-the-go interactions, positioning ChatGPT in competition with established personal assistants such as Google’s Assistant, Apple’s Siri, and Amazon’s Alexa.
Users can make a variety of requests, like inquiring about Disneyland’s history during a drive to the theme park or seeking a cocktail recipe while in the kitchen. During testing, the tool effectively narrated a story involving a starfish and a rutabaga. However, it’s important to note that while ChatGPT can generate song lyrics, it won’t perform singing.
The voices of ChatGPT possess a mostly human-like quality, although upon closer scrutiny, a subtle robotic undertone can be detected. OpenAI collaborated with voice actors in the development of the text-to-speech AI model that forms the foundation of this feature.
The company has announced that in the coming weeks, paid and enterprise users will gain access to a feature involving GPT-4, one of the AI models supporting ChatGPT. This feature will allow users to submit a picture along with a related question. For instance, users can upload an image of pink sunglasses and ask the chatbot for outfit suggestions to match them, or they can submit a picture of a math problem and request assistance in solving it. OpenAI first revealed this feature earlier in the year when introducing GPT-4, and it will be accessible through both the ChatGPT app and website.