Now you can use ChatGPT Voice without leaving your chat (www.engadget.com)

OpenAI has updated ChatGPT Voice on the web and in its mobile app so voice interactions run inline with your existing chat instead of launching a separate interface. You start a voice session by tapping the waveform icon next to the text field, and ChatGPT will return an audible reply plus an on-screen transcript and contextual visuals (the demo showed a transcript followed by a map of bakeries and photos of pastries). If you prefer the older, separate “orb” experience you can re-enable it via Separate mode in Voice Mode settings. This change is a practical multimodal step: voice input and output now live alongside images, maps and text, preserving conversational context and making responses more actionable and scannable. For AI/ML users it signals emphasis on integrated multimodal UI/UX—voice is being treated as another modality that can be augmented with visual grounding rather than a standalone channel. Compared with Google’s more reactive Gemini Live overlays, OpenAI’s approach is less real‑time visual annotation but still improves informativeness and workflow continuity, which matters for building voice-first agents, accessibility features, and multimodal applications.