ChatGPT gets support for voice prompts and image searches: Here’s what you need to know

The tech space has recently been abuzz with everything artificial intelligence, from OpenAI’s DALL-E 3 model to Apple’s Ajax conversational AI tool. In a space where competition is fast heating up, each player has consistently been updating its offerings in a bid to stay on top of the game.

Now, OpenAI has introduced a potentially game-changing update for its ChatGPT, available on both laptops, as well as smartphones. In addition to text-based prompts, the AI chatbot will also support voice-based prompts and image searches. Intrigued? Read on.

How voice-prompts work on ChatGPT

With the latest version of ChatGPT, one needn’t type prompts into the text box for answers anymore. All one has to do is tap a button and speak. It’s that simple.

What ChatGPT does, is convert your voice into text and feed it into the GPT-4 Large Language Model (LLM). Once it has an answer, it converts the same back into audio and speak it out loud to the user. OpenAI expects to provide an experience similar to talking to virtual assistants such as Siri or Google Assistant, except with better responses.

OpenAI’s speech-to-text model is called Whisper. The company says it can generate up to five different human voices with this. As per The Verge, the company is also working with Spotify to translate podcasts on the platform into different languages, while retaining the original sound of the podcaster’s voice.

What image search on ChatGPT can do for you

Coming to the second major update, and image search support on ChatGPT works in a manner similar to Google Lens. All you have to do is click a picture and feed it to ChatGPT. The AI tool will then try and figure out possible queries and provide responses to the same. Not just that, in order to refine your prompt, you can also use the in-built drawing tool to point out specific aspects of the image you may have questions about, and even speak or type questions about the image.

Now, while both of the features are extremely useful in their own right, they come with certain implications on security as well. The fact that one can generate voices, or deepfake audio on the platform, poses a potential risk of identity theft and fraud.

ALSO READ: DALL-E 3 launched for aspiring AI artists

Similarly, coming to image searches, the platform may also make direct, potentially problematic statements about certain public figures, events, objects, and locations, which may end up causing more harm than good. OpenAI, therefore, says that these features will be rolled out in a controlled manner, and possibly be restricted to certain individuals or organisations only.

The new features will be rolled out to ChatGPT Enterprise users in the next two weeks, while everyone else using the free GPT 3.5 model is expected to get access to the new features after.