Voice Dictation in VS Code Copilot Chat

No, not really. VS Code ships one official dictation path — the free VS Code Speech extension from Microsoft — and it does process your audio locally rather than in the cloud. But it's a general-purpose speech tool, not one built for reading out code, file paths, or long technical prompts, and that's precisely where developers say it falls apart. A GitHub community discussion opened in March 2026 calls it "effectively unusable when dictating technical prompts, code-related language, file names, symbols, and longer instructions" — and asks GitHub directly for native, Whisper-grade dictation in Copilot Chat.

The good news: you don't have to wait for that request to land. Dictating into VS Code — Copilot Chat's input box, inline comments, commit messages, PR descriptions — works today if you stop looking for an editor feature and use a local, system-level dictation tool instead. Below: what's actually built in, what the record shows about GitHub's own plans, and how to set up dictation that works everywhere in VS Code, not just one text box.

Does VS Code have built-in voice dictation?

Yes, in the form of an extension, not a core feature. VS Code Speech (ms-vscode.vscode-speech) is Microsoft's own extension, installed over 1.3 million times, and its listing is explicit about where processing happens: "No internet connection is required, the voice audio data is processed locally on your computer." It adds two things: dictation straight into the code editor, and speech-to-text inside GitHub Copilot Chat (plus text-to-speech for reading Copilot's replies back to you). Using it inside Copilot Chat requires the GitHub Copilot Chat extension installed and signed in — the editor-dictation mode alone doesn't.

Why isn't VS Code Speech good enough for coding prompts?

Because it's a general dictation engine asked to do a specialized job. The community discussion filed under GitHub's official feedback board lays out the complaint in detail: "There are multiple public bug reports around voice input reliability, including cases where no speech is converted at all," and for real development work, the experience is "effectively unusable when dictating technical prompts, code-related language, file names, symbols, and longer instructions." The same post explicitly points to the bar Whisper-based tools have set: "Whisper-style transcription raised the bar on what good speech input looks like: stronger handling of accents, background noise, long-form speech, and technical language" — and notes developers are already dictating into tools like Claude Code instead, then pasting the result into Copilot.

There's history here, too. GitHub already tried a dedicated voice product once: "Hey, GitHub!" (later Copilot Voice) ran as a GitHub Next technical preview starting in November 2022, aimed at hands-free pair programming and accessibility. GitHub Next concluded the preview on April 3, 2024, and folded its learnings into the VS Code Speech extension rather than shipping it as a standalone product. In other words, today's extension is largely what's left of that earlier, more ambitious effort — which helps explain the gap between what developers want and what ships by default.

Is native voice dictation coming to VS Code Copilot Chat?

Not on any confirmed timeline. Two GitHub-hosted threads track the actual state of demand:

• Issue #299944 — "Feature Request: Native 'Voice-Only' Mode for VS Code Copilot Chat," proposing a hands-free, wake-word-driven chat mode. It went through VS Code's standard triage: the maintainers' bot flagged it as a backlog candidate needing 20 community upvotes within 60 days. It didn't reach that bar and was auto-closed as not_planned. Notably, a commenter pointed out that a lighter version already exists — VS Code Speech supports a "Hey Code" wake word to start voice chat, paired with an accessibility.voice.speechTimeout setting to auto-submit after a pause — it's just underused and not marketed as a headline feature.
• Discussion #190615 — "Feature Request: Native voice dictation for Copilot Chat in VS Code is urgently needed," filed in GitHub's own Copilot Conversations feedback category. As of writing it has 27 upvotes and remains open and unresolved — the clearest public signal that current accuracy, not the lack of any voice feature at all, is the actual gap.

Read together, the pattern is: VS Code already has voice affordances, they're just not accurate enough for technical dictation, and the fix for that specifically hasn't shipped.

What about GitHub Copilot CLI's voice input?

It's a real, local option — for a narrower job than "dictate in VS Code." GitHub's own Copilot CLI voice input documentation states plainly: "Transcription runs entirely on your local machine. Your audio is not sent over the network." Setup happens through the /voice slash command inside an interactive Copilot CLI session, which downloads a voice runtime and a voice model once. The catch: it currently supports English and Spanish only, and it dictates into Copilot CLI's own prompt box in a terminal window — including one you can open inside VS Code's integrated terminal — not into the editor, inline comments, or the Copilot Chat sidebar directly.

What actually works: local dictation at the OS level

The pattern above repeats for a reason: every option so far is scoped to one surface — VS Code Speech to the editor and Chat, Copilot CLI's voice input to its own terminal prompt. None of them dictate into a commit message editor, a PR description text box in your browser, or a plain // TODO comment. The fix isn't a better VS Code extension; it's stepping outside VS Code entirely, to a dictation tool that runs at the operating-system level and types wherever your text cursor happens to be — the same way your keyboard does. That covers Copilot Chat's input box, inline comments, the commit message editor, a PR description, a terminal, or any other app, with one tool and one hotkey.

If accessibility or repetitive strain is what's pushing you toward dictation in the first place rather than raw speed, that same system-level approach is also the one covered in our guide to hands-free coding for RSI.

Setting up local dictation for Copilot Chat, comments and commits

Inkvox is one example built specifically for this: a small tray app for Windows 10/11 that runs Whisper large-v3-turbo, quantized, directly on your own GPU through Vulkan — NVIDIA, AMD, or Intel — with a CPU fallback if you don't have a compatible card. Here's how the local transcription pipeline works. On a mid-range GPU like an RTX 3070, a sentence transcribes in roughly 0.3–0.4 seconds. Nothing is uploaded: 0 bytes of audio leave your machine, no account is required, and once the ~800 MB model is downloaded on install, it runs fully offline in 100+ languages.

In practice: place your cursor in Copilot Chat's input box, a code comment, the commit message field, or a PR description tab — tap the hotkey, speak your prompt or your commit message the way you'd say it out loud, tap again, and the text lands exactly where your cursor is. Same tool for all of it, no copy-pasting between a "good" dictation app and VS Code.

VS Code dictation options, compared

Frequently asked questions

Does VS Code have a built-in dictation feature?
Yes. Microsoft's free VS Code Speech extension adds speech-to-text to the editor and to GitHub Copilot Chat, and according to its own marketplace listing, no internet connection is required — the voice audio data is processed locally on your computer. It's real and it's local, but it isn't tuned for dictating code, file paths or technical prompts, which is exactly where developers say it breaks down.

Can I dictate directly into GitHub Copilot Chat?
Yes, with the VS Code Speech extension installed and GitHub Copilot Chat signed in, you can click the microphone icon or say the "Hey Code" wake word to start a voice chat session. It works, but a GitHub community discussion describes it as effectively unusable when dictating technical prompts, code-related language, file names, symbols and longer instructions — which is why some developers dictate elsewhere and paste the result in.

Is the VS Code Speech extension processed locally, or does it use Azure's cloud?
According to Microsoft's own documentation for the extension, processing happens locally on your machine with no internet connection required. That's a different product from Windows' own Voice Typing (Win+H), which does route audio through Microsoft's Azure Speech cloud service — the two are unrelated.

Does GitHub Copilot CLI have offline voice input?
Yes. Copilot CLI's /voice command downloads a voice runtime and a voice model once, then transcribes locally — GitHub's documentation states transcription runs entirely on your local machine and audio is not sent over the network. It currently supports English and Spanish, and it's scoped to Copilot CLI's own prompt box in the terminal, not to VS Code's editor or Chat sidebar.

What's the best way to dictate into VS Code comments, commit messages and PR descriptions?
Since no built-in option covers every text field, the practical fix is a system-level dictation tool that runs at the OS level and types wherever your cursor is, instead of being scoped to one extension. That covers Copilot Chat's input box, inline code comments, the commit message editor, a PR description in your browser, and the terminal, all with the same tool and the same hotkey.

If you'd rather dictate everywhere in VS Code today than wait on a feature-request thread, join the Inkvox waitlist — local dictation that doesn't care which text box your cursor is in.