Bing Chat now lets you talk directly to AI

Microsoft just turned Bing Chat on desktop into a real conversation.

The company announced that users can now talk to Bing Chat using voice commands by clicking the microphone button in the chat box, with the AI speaking its answers back aloud.

Voice input meets longer memory

The feature currently supports English, Mandarin, German, and French, with more languages coming soon. It relies on text-to-speech to read responses directly to users.

Alongside voice support, Microsoft also raised Bing Chat's conversation limits.

The cap per session is now up to 30 turns, up from 20 previously. The daily limit has also jumped to 300 total turns.

These new limits apply to conversations saved in your Bing chat history, meaning you can jump back into older sessions that previously hit a wall and pick up right where you left off.

This update follows a busy few weeks for the tool. Microsoft recently rolled out Bing Image Creator across all chat modes and started testing richer visual cards with Bing Travel links for vacation queries.

It also comes exactly one week after Microsoft announced the retirement of the standalone Cortana app on Windows, paving the way for Windows Copilot.

Why it matters

We think adding voice to desktop Bing Chat is far more than an accessibility perk: it is part of a deliberate push to replace legacy voice assistants with generative AI.

For everyday users, being able to talk to Bing in Mandarin, French, German, or English reduces friction for multitasking and researching. The increase to 30 turns per session also gives users enough runway to work through complex, multi-step queries without the frustrating cutoff.

For the broader industry, ditching Cortana and feeding all voice interactions directly into Bing and Copilot shows Microsoft's aggressive consolidation strategy. Instead of maintaining a separate assistant ecosystem, everything is now converging into the same conversational core.

The catch is voice latency and hallucination. When an AI reads an answer out loud, users are much less likely to cross-check sources compared to reading text and clicking citations. Microsoft will need to prove that its voice synthesis can handle fast-paced research without sounding like a confident guesser.