Short answer: put the cursor anywhere text should land in Windows Terminal, press Ctrl+Alt+R, speak, press it again. AICHE inserts a clean transcript at the cursor. Works in every profile - PowerShell, CMD, WSL, and any AI coding agent running inside a tab.
What This Is For
Use voice for prose in Windows Terminal, not shell commands:
- AI agent prompts in Claude Code, Cursor agent, Copilot CLI, or similar (full context, constraints, test commands)
- Git commit bodies opened via
git commitwithout-m(subject + why + rollback notes) - Project notes in README, config comments, or scratch files open in the editor pane
- Documentation you would skip because typing breaks flow
Do not dictate git push, npm install, or PowerShell one-liners. Commands stay typed; AICHE inserts only where a text cursor accepts paragraphs.
AI Coding Agents Running In Your Terminal
Claude Code, Cursor agent sessions, GitHub Copilot CLI, Warp Agent Mode, and similar tools run inside Windows Terminal tabs. The shell commands are short; the agent prompts are not. Each prompt that needs repo context, files to avoid, test commands, and edge cases takes minutes to type. Most people shorten it and pay in extra agent rounds.
Click the agent input line, press Ctrl+Alt+R, speak the full brief, press again. AICHE inserts at the cursor. Hit Enter. Same workflow for Cursor's Agents Window when you run Cursor from a terminal tab on Windows.
Math: speaking lands around 150 WPM. Typing in a terminal prompt lands around 40 WPM. A 200-word prompt that takes five minutes to type takes about 80 seconds to speak.
How It Works
- Open Windows Terminal on any profile - PowerShell, CMD, WSL, a Claude Code session.
- Click into the pane where you want the text. For an agent prompt, that is the input line. For commit messages longer than a short subject, run
git committo open your editor and dictate there. - Press Ctrl+Alt+R.
- Speak. The recording is a toggle - you do not have to hold the key down.
- Press Ctrl+Alt+R again. AICHE drops clean text at the cursor.
- For short one-line subjects, you can send directly. For multi-line commit context, dictate in the commit editor and save.
No plugin. No setup inside the terminal. No difference between profiles or tabs.
Git Commit Messages
The commit message is the most obvious place to use voice in a terminal. Most repositories have one-line commits that say "fix" or "update" because writing a real explanation after every change is slow and nobody does it.
Use git commit (without -m) for anything longer than a short subject. Let Git open your editor, then dictate subject plus body safely in separate lines. This avoids shell-quote breakage and gives you a real commit body in history.
If you use a graphical editor for commits - VS Code opens automatically when you run git commit without -m - dictation works there too. Place the cursor in the message field, press the hotkey, speak, press it again.
Writing Notes And Documentation
Windows Terminal is where a lot of project work happens, and project work generates text that never gets written: what a script does, why a particular setting is set the way it is, what someone needs to install before the project runs. That text stays unwritten because typing it feels like extra work after the actual task is already done.
With voice it stops being extra work. Place the cursor in whatever file is open, press the hotkey, say what you want to say, press it again. A paragraph of context that would have taken ten minutes to type takes about a minute to speak.
This works in any file open in the terminal - a README, a config file with a comments section, a notes file you keep next to a project, a text file you are drafting for something else. Wherever the cursor is, that is where the text goes.
When You Run Agents All Day
If your main use of Windows Terminal is running AI coding tools, the prompt is where all your time goes. The agent builds in seconds. You spend the next few minutes typing the next instruction.
Voice changes that ratio. You speak the next prompt while the agent is still working on the previous one. Better prompts because you are not editing in your head to save typing time. Faster turnaround because speaking is three to four times faster than typing. Less fatigue at the end of the day because your hands are not doing all the work.
The Software Development profile (Pro) tunes recognition for code-heavy speech - library names, flag names, technical terms. Custom vocabulary (Pro) lets you add your own repo names, service names, and internal jargon so they are spelled right every time.
Works Across Every Profile And Pane
AICHE inserts text into whatever field has focus at the OS level. Windows Terminal can have PowerShell, CMD, WSL, and an agent session all open at once in different tabs or panes. You do not need to configure anything differently per profile - click into the pane you want, press the hotkey, speak, press it again. The same hotkey works in all of them.
That matters when you switch between profiles frequently. A typical session might be PowerShell for running builds, WSL for the actual development work, and a separate tab for the agent. Moving between them does not change anything about how AICHE works.
The same applies to split panes, focus mode, and quake mode (wt -w _quake). Click into the pane, press the hotkey.
For Non-English Thinkers
Turn on Auto-translation in AICHE settings. Speak in German, Mandarin, Spanish, Japanese, or whichever language your brain runs in. AICHE outputs English text into the terminal. Your git history, agent prompts, and notes read as if you wrote them in English the whole time.
This is useful any time you are working with an English-language codebase or team but think more naturally in another language. No translation step in your head while typing.
Common Questions
Q: I just started using Windows Terminal. Will this work for me?
A: Yes. Install AICHE, open Windows Terminal, click into the input line, and press Ctrl+Alt+R. That is the whole setup. You do not need to configure anything inside Windows Terminal.
Q: Does this work with Claude Code or other AI tools running in a tab?
A: Yes. AICHE inserts text into whichever pane has focus. Place your cursor in the agent input, press the hotkey, speak your prompt, press it again. The agent receives it as if you typed it.
Q: Does it work in WSL - the Linux side inside Windows Terminal?
A: Yes. AICHE runs on the Windows side and inserts text into whichever pane is focused. It does not matter if that pane is running PowerShell, CMD, or a Linux shell through WSL.
Q: Does it work in a split pane or just the main tab?
A: It works in whichever pane has focus. Click into the pane you want, then press the hotkey.
Q: I do not want my audio stored anywhere. What happens to it?
A: Audio is streamed for cloud transcription, processed, and discarded immediately after processing, within 1 second. No persistent audio copy.
Q: Is there a push-to-talk mode?
A: No. Recording is toggle-based - press once to start, press again to stop. You do not have to hold the key while you talk.
Result: the text that lives inside your Windows Terminal sessions - agent prompts, commit bodies, notes, documentation - stops being the part that slows you down. Commands get typed. Everything longer gets spoken.
Try it now: open Windows Terminal in any repo with uncommitted changes, run git commit, press Ctrl+Alt+R in the commit editor, and dictate a subject plus a short body explaining what changed and why.