OpenAI has fixed a significant issue with ChatGPT's voice mode.
Switching between speech and text has never been easier on ChatGPT.
OpenAI has announced that ChatGPT Voice can now be used directly within the standard chat interface.
Users can speak to the AI, view responses on screen, reference past messages, and see visuals like images or maps in real time.
The update is being implemented for all users on mobile and web platforms. To access it, simply update the app. A waveform icon allows voice command use, enabling direct conversation with ChatGPT in chat. The update is available on the app and web. What is the purpose of such a change?
Previously, the orb interface only allowed listening to ChatGPT's spoken responses without a screen transcript. Users had to exit voice mode to see the transcript if they missed something. Now, the transcript is visible during voice interaction.
The AI can display real-time maps and weather, but the map function was imperfect. While the weather was accurate, map display didn't match expectations.
Instead of showing a map of local restaurants, the chatbot gave direction links, unlike the video demo on X. Refining the request to a specific restaurant still yielded direction links, not a map. Oddly, the map appeared using the exact command from OpenAI's video.
Voice mode had an initial surge in popularity, but user interest waned over time.
By integrating voice into the main chat, OpenAI hopes to increase its use. More usage means more voice data to refine its models.
OpenAI should be well-positioned by integrating voice. Users can opt out of having their voice data used for AI training:
In the app, tap the Customize icon, then your name to access settings. Choose Data Controls and disable "Include your audio recordings."
Users can revert to the prior orb interface by enabling the "Separate Mode" toggle in settings under Voice (mobile) or Personalization/Advanced (web).
The integration aims for more natural chatbot interactions, allowing seamless switching between text and voice. However, voice mode remains active until manually ended.
In one instance, the feature remained on, and ChatGPT began providing a tea recipe after overhearing a request to a family member.
Ideally, OpenAI will add an auto-deactivation feature after inactivity. Gemini already has "Gemini Live" with a separate screen and real-time transcript.
The author previously favored Gemini Live on their Pixel 10 for its speech and transcript. They may switch to ChatGPT now, preferring its voice.