News
Speak Your Vibe: VS Code Insiders Talks Up Voice-Driven Development
Visual Studio Code developers appear to have plenty to talk about in the July Insiders cycle -- quite literally, as voice-related technology is the most conspicuous theme in the early VS Code 1.131 updates.
The still-evolving Insiders notes, last updated July 24, list 16 changes so far. Eight of them involve dictation, Voice Mode, microphones, speech processing or text-to-speech, an unusual concentration for a development tool whose recent updates have been dominated by AI agents and coding automation.
Microsoft has not formally characterized 1.131 as a voice-focused release, and several of the entries are related parts of the same broader engineering effort. Still, voice technology clearly stands out from the otherwise scattered assortment of Agents window, terminal and chat improvements appearing in this iteration.
VS Code's voice-heavy turn is not happening in isolation. GitHub made voice input generally available in Copilot CLI in June, while AI-focused editor Cursor has introduced voice control for its coding agents and recently extended spoken instructions across its Agents window, visual Design Mode and mobile app. The developments fit a broader move toward multimodal AI: IDC predicts that 80 percent of production-grade foundation models in Asia/Pacific will support combinations of text, images, audio and video by 2028. Rather than reviving voice coding as a replacement for typing source code, the emerging pattern uses speech as another natural-language control surface for directing AI agents and describing higher-level development tasks.
Moving Voice into VS Code
The most substantial change expands VS Code's built-in, on-device dictation engine beyond chat. Microsoft is extending it to support dictation in the editor and terminal, while also adding a Read Aloud text-to-speech capability.
According to the associated GitHub issue, the goal is to cover speech functions that currently depend on the VS Code Speech extension. Microsoft said bringing those functions into the product would provide one consistent speech stack, eliminate overlapping microphone controls and keybindings, and eventually make the extension redundant for dictation and Read Aloud.
The on-device approach also keeps audio processing local rather than sending it away from the machine, according to Microsoft, although fallbacks may still be required in environments where built-in transcription is unavailable.
Much of the remaining work is aimed at making dictated input less like a raw transcript and more like finished text. VS Code can filter verbal filler such as "um" and "uh," refine punctuation and capitalization while the user is still speaking, and recognize when dictated content should be formatted as a bulleted or numbered list.
Other changes refine the surrounding experience. Users can select or disable a microphone directly from voice controls, built-in Voice Mode can disable a conflicting extension-provided implementation, and Voice Mode no longer automatically sends a turn by default when hands-free operation is disabled.
Everything Else So Far
The non-voice changes include easier creation of custom session groups in the Agents window, streaming command output, single-file diff views, fuzzy file search in the Agent Host, folder selection for new agent sessions, timestamps on older chats and an option to disable the terminal dimensions overlay.
None appears to represent a major standalone feature, leaving the cluster of speech improvements as the clearest story in the release so far.
That "so far" remains important. Insiders notes are updated throughout the development cycle, and Microsoft warns that the page will continue to evolve as new features are added. The final balance could look different by the time VS Code 1.131 reaches stable release, but the July iteration currently suggests that Microsoft wants talking to the editor to become a more native part of working in it.
About the Author
David Ramel is an editor and writer at Converge 360.