1. What JetBrains offers for speech in October 2026
JetBrains’ AI Chat documentation describes typed prompts with attached files, folders, symbols and images, and no microphone. YouTrack issue JUNIE-796, opened in September 2025 and still open, says there is currently no functionality in Junie or AI Assistant to turn speech into text, and suggests a third-party speech-to-text tool as the workaround. Older requests ask for the same in the IDE itself: LLM-1562 (voice control for AI Assistant, open since November 2023) and IJPL-217393 (voice-driven coding, open since November 2025).
What JetBrains does document is screen-reader support on Windows, macOS and Linux: NVDA 2019.3 or later and 64-bit JAWS 12.0.1158 or later on Windows, Orca on Linux, and a Support screen readers option under Settings | Appearance & Behavior | Appearance. Screen readers read the interface aloud; they do not type your speech.
The one JetBrains product with a microphone is JetBrains Air, a separate agentic development environment that now installs on macOS, on Linux through an install script or the Toolbox App, and on Windows through the Toolbox App. In Air, open Settings, search for Voice Input, download a local Whisper model, optionally pick the model size and language, then use the microphone in the Chat task input. JetBrains says Air uses that local model and sends no audio to the cloud. It helps with Air task prompts only, not with code, comments or AI Assistant in IntelliJ IDEA.
2. Windows: Win+H voice typing in the editor and tool windows
Press Windows key+H with the caret in the editor, a commit message, the terminal tool window or the AI Assistant input, wait for “Listening…”, and speak. Microsoft says voice typing uses online speech recognition powered by Azure Speech, so it needs an internet connection, a working microphone and the cursor in a text box. Say “Stop listening” or select the microphone in the voice typing menu to stop, and press Windows key+Alt+H to move through that menu from the keyboard. The settings gear holds the voice typing launcher, automatic punctuation, a profanity filter and the default microphone.
On Copilot+ PCs, Fluid dictation is on by default in English: on-device small language models correct grammar, punctuation and filler words as you speak. That helps in a commit message but can rewrite a symbol you meant literally, so say “Undo that” or press Ctrl+Z to restore the original wording, or turn Fluid dictation off in voice typing settings while you work on code.
Microsoft documents voice typing for text boxes in general, and JetBrains does not document Windows speech tools at all. Test the exact field you care about before relying on it. One data point from JetBrains’ tracker: IJPL-58446, open since 2021, comes from a Dragon Professional user who could not dictate directly into the IntelliJ IDEA editor and had to paste from an intermediate window.
3. Windows: Voice Access for offline dictation and spoken control
Voice Access is Windows 11’s broader voice tool, available from version 22H2. Microsoft describes it as a way to control your PC and author text using only your voice and without an internet connection. That makes it the built-in choice when you need dictation on a train, on a locked-down network, or with your hands off the keyboard entirely.
It has three modes. The default mixes commands and dictation; say “Dictation mode” to treat everything as text for an editable field, or switch to commands mode for control only. The “Add to vocabulary” command and menu item bias recognition toward words you add, which is worth doing for class names, library names and acronyms your project uses. Some languages are ready to use and others need a speech pack download first.
- Windows 11 version 22H2 or later; no internet needed once the language is installed.
- Say “Dictation mode” before dictating a long comment, so phrases are not taken as commands.
- Add project words with “Add to vocabulary” before dictating them.
- Test it in the editor and in AI Assistant; JetBrains does not document Voice Access support.
4. Linux: native Wayland changes how text reaches the IDE
On Linux the IDE offers no dictation, so speech comes from a separate app that synthesizes keystrokes or pastes into the IDE window, and how that works depends on the display server. In February 2026 JetBrains announced that IntelliJ-based IDEs would run natively on Wayland, starting with the 2026.1 EAP, in supported desktop configurations, instead of as X11 apps through XWayland. To see which mode your IDE uses, open Help | About, click Copy and Close, and look for Toolkit: sun.awt.wl.WLToolkit (Wayland) or XToolkit (X11). To return to X11, add -Dawt.toolkit.name=XToolkit under Help | Edit Custom VM Options and restart; Remote Development mode does not switch to native Wayland automatically.
The typing tool has to match. The nerd-dictation README describes xdotool as simulating input in X11 and ydotool as working anywhere, including Wayland. Handy’s README recommends xdotool on X11 and wtype or dotool on Wayland, notes that dotool needs your user in the input group, and says wtype does not work on Ubuntu 26.04’s default Wayland session, where ydotool is needed instead. If speech is recognized but nothing lands in the editor, the typing tool is the first thing to check.
Choices that run on Linux today include Handy (free, MIT-licensed and offline; version 0.9.8 publishes deb, rpm and AppImage builds for x86-64 and ARM64), Spokenly for Linux (version 1.0.3 in October 2026, deb, rpm and AppImage packages for x86_64, free with local Whisper and Parakeet models, X11 and Wayland) and nerd-dictation (a script you bind to begin and end shortcuts yourself). JetBrains Marketplace also lists community voice plugins, including one that runs whisper.cpp inside IntelliJ IDEA. Vet the publisher first: on June 16, 2026 JetBrains removed 15 third-party AI plugins that had been built to steal users’ AI provider API keys.
5. Ctrl+Space, remapped paste keys and elevated IDEs
JetBrains binds Basic code completion to Ctrl+Space in its Windows, GNOME, KDE and XWin keymaps. IraVoice’s default dictation key on Linux is also Ctrl+Space, so in a JetBrains IDE the two compete. Change IraVoice’s key under General, or on Wayland, where global shortcuts may not reach IraVoice at all, create a custom shortcut in GNOME or KDE settings for the command iravoice --toggle-transcription and pick a combination the IDE does not use.
On Windows, IraVoice inserts text by pressing Ctrl+V and then restores your clipboard. If your IDE keymap gives Ctrl+V another job, as a Vim-style keymap can, switch IraVoice to a different paste method such as Shift+Insert, Ctrl+Shift+V or direct typing.
Some teams run Rider or IntelliJ IDEA as administrator, for example to manage local services. Windows blocks synthetic input from reaching an elevated app, so IraVoice cannot paste there; it leaves the text on the clipboard and tells you, and you press Ctrl+V yourself. Running the IDE without elevation avoids the extra step.
6. IraVoice in IntelliJ IDEA, PyCharm, WebStorm and Rider
IraVoice 1.2.2 is a desktop dictation app for macOS 13+ on Apple silicon, Windows 10 or 11 (64-bit), and Linux (beta). It recognizes speech on the computer with NVIDIA Parakeet; the default Parakeet Unified EN model (about 700 MB) downloads once, and after that dictation works offline. Hold Right Ctrl on Windows or Ctrl+Space on Linux, speak, and release: the text lands at the caret, whether that is the editor, a commit message, the terminal tool window or an AI chat prompt. It does not press Enter by default, and Esc cancels a dictation on Windows.
On Windows, IraVoice recognizes IntelliJ IDEA, PyCharm, WebStorm and Rider and spells the code names of the Git project the IDE has open: “get user by ID” becomes getUserById and “use effect” becomes useEffect. It keeps only file, folder and declared identifier names, never source text, skips secrets and dependency folders, and sends nothing off the computer. Other JetBrains IDEs on Windows still get dictation, but code names are promised only in the listed apps. When IraVoice mishears a word, press Ctrl+Alt+Shift+F or choose Fix Last Dictation from the tray menu, correct it, and it types the word right from then on.
On Linux, IraVoice is a beta that comes as an AppImage for any 64-bit distribution from 2022 or later, which updates itself, or as a .deb or .rpm package. It types with a helper it picks automatically (wtype, kwtype, dotool, ydotool or xdotool), which you may need to install, and Wayland support is limited in the beta. It cannot detect the focused app on Linux, so there are no code names there. On a Mac, JetBrains IDEs get code names too; see the Mac guide.
IraVoice Local costs $79 once for one computer, $109 for two, or $149 for three, and the 3-day free trial starts with your first dictation, with no email, card or account needed to download.