For developers
Voice coding that stays on your computer.
Talk to Claude Code, Codex, Cursor or any terminal. Yap turns your speech into text on your Mac or PC, so your voice, your code and your screen never go to anyone’s servers. It costs less than half of Wispr Flow.
Nothing leaves your computer
Speech is recognized on your machine. Yap doesn’t read your screen, your files or your clipboard, and uploads no audio or text. Under an NDA, on client code or on a plane, it works the same.
$6.90 USD a month
Or $69 USD a year. Wispr Flow Pro is 15 USD a month, Aqua Voice Pro 10. 3 days free with a free account, no card, then a 10-day trial with a card.
No round trip
A median 0.56 seconds from releasing the key to finished text on an Apple M5 Pro, 1.4 on an Apple M2 Pro. No upload and no queue, and no Wi-Fi needed while you talk.
Where your voice goes
What dictation sends while you code.
Most voice input for coding is a cloud service: your recording goes to a server to become text, and some apps send what is on your screen along with it, to get names right. Here is what each one says about itself.
| Tool | Speech becomes text | Sent along with it | A month |
|---|---|---|---|
| Yap | On your Mac or PC | Nothing. It doesn’t read your screen or your clipboard, and uploads no audio or text. | $6.90 USD, or $69 a year |
| Wispr Flow | Their cloud | Context Awareness, on by default: text around your cursor, on-screen text, and variable and file names in coding apps | 15 USD, or 12 billed yearly |
| Aqua Voice | Their cloud | Transcripts may be stored on their servers: Privacy Mode is off by default | 10 USD, or 8 billed yearly |
| Willow | Their cloud | Not stored by default, they say | 15 USD, or 12 billed yearly |
| Claude Code /voice | Anthropic’s servers | Your project and git branch names, as recognition hints | Included with a Claude.ai plan |
| Codex voice | OpenAI's servers | - | USD 0.05 a minute from your Codex usage |
| Gemini CLI /voice | Google Cloud by default; local Whisper optional | - | Included |
| Warp voice input | Wispr Flow's servers | - | Included in Warp plans |
| VS Code dictation | On your computer (desktop VS Code) | - | Free, inside VS Code only |
Wispr Flow: “Transcription always occurs on the cloud.” and “When Context Awareness is on, context can accompany your dictation: app info, textbox contents (before, selected, and after the cursor), on-screen text, variable and file names in coding apps …” “It's on by default.” Its data controls add that “your data (i.e. audio, transcript, edits) may be used to evaluate, train, or improve AI models, by Wispr.” Sources: data controls, Context Awareness, read October 19, 2026.
Aqua Voice: “Aqua processes audio in the cloud, where the Avalon model runs; there is no offline mode.” Source. “With Privacy Mode off (the default), transcripts may be stored on our servers to improve the product.” Source, read October 19, 2026. Willow: “Willow runs on secure cloud infrastructure in the United States.” and “Your transcripts are not logged server-side or retained by default after the dictation is complete.” Source; pricing, read October 19, 2026.
Claude Code: “Voice dictation streams your recorded audio to Anthropic’s servers for transcription. Audio is not processed locally.” and “your current project name and git branch name are added as recognition hints automatically.” Source, read October 7, 2026. Codex voice: “Voice in Desktop uses your existing Codex usage budget at $0.05 per minute.” Source. Gemini CLI /voice: “When using the Gemini Live backend, voice recordings are sent to Google Cloud for transcription.” Source. Warp voice input: “The transcription is powered by Wispr Flow.” Source. VS Code dictation: “On desktop, dictation uses an on-device speech recognition model by default, so you can dictate without sending audio to an online service.” Source. Read October 19, 2026.
Works everywhere you type
One shortcut for every agent.
Yap doesn’t plug into your tools. It types into the window in front, so Claude Code, Codex CLI, Gemini CLI, Aider, OpenCode, Cursor, VS Code with Copilot, Windsurf, JetBrains IDEs, Zed, and Xcode get your words the same way they get typing, in Terminal, iTerm2, Ghostty, Warp, kitty, Windows Terminal, and PowerShell, and over SSH, because your terminal passes the keys on to the remote shell.
Hold Fn on a Mac or Ctrl + Win on Windows, describe the change, release. Your words are typed at the cursor without touching your clipboard, so what you copied is still there. Yap never presses Enter: you read the prompt, then send it. Every dictation is also kept in History on your computer, in case focus moved.
The same shortcut writes your commit messages, pull request descriptions, docs and Slack replies, in any of its 25 languages, recognized per recording.
Working with recordings instead of live speech? Our free, open-source Yap MCP server lets Claude Code and other MCP clients transcribe audio files on your computer with the same engine.
Being fair
When something else fits better.
Claude Code /voice
Free with a Claude.ai sign-in and tuned for coding words. Fine if you only dictate to Claude Code, on your own machine, and don’t mind your audio going to Anthropic. How the two compare.
Talon with Cursorless
If you want to edit code itself by voice, without a keyboard, for example with RSI: a command grammar you learn over weeks. Programming by voice, explained.
VS Code dictation
On-device and free, if every word you dictate goes into VS Code. It doesn’t reach your terminal app, your browser or chat.
Questions developers ask.
- Does Yap send my code or my voice to the cloud?
- No. Yap recognizes speech on your computer with a model it downloads once, then types the text where your cursor is. It doesn’t read the text on your screen or your clipboard, and it uploads no audio, transcripts or usage statistics. It goes online only to check your plan, download the speech model once and fetch updates.
- Does Yap work with Claude Code, Codex and Cursor?
- Yes. Yap types into whichever app is in front, so Claude Code, Codex CLI, Gemini CLI, Aider, OpenCode, and Cursor receive your words the way they receive typing. It never presses Enter, so you read the prompt before you send it.
- Can I dictate prompts over SSH?
- Yes. Yap types into the terminal window on your own computer, and the terminal passes the keys on to the remote shell like any typing. Anthropic’s docs say Claude Code’s own /voice doesn’t work in SSH sessions.
- How much cheaper is Yap than Wispr Flow?
- Yap is $6.90 USD a month or $69 USD a year. Wispr Flow Pro is 15 USD a month, or 12 a month billed yearly, so Yap costs about 46% of it.
- How fast is Yap?
- On an Apple M5 Pro, the median time from releasing the key to finished text was 0.56 seconds over 7 dictations; on an Apple M2 Pro, 1.4 seconds. Nothing waits on a network, so it is the same offline.
- Does Yap know my project’s file and function names?
- Not yet: Yap has no custom vocabulary. Say names the way you would spell them out, or fix them before you press Enter. If your prompts are mostly identifiers, Claude Code’s /voice is tuned for coding words.
From the library
Voice Typing in Google Docs, Word and Every Other App
Google Docs, Word, Outlook, PowerPoint, OneNote, ChatGPT, Claude and Claude Code have their own dictation button, but each one types only inside its own app and usually needs an internet connection or a paid plan. Gmail on a computer, Slack, terminals and Linux desktops have none, so the reliable setup is the built-in button where it is good enough plus one system-wide dictation tool that types into whatever text field has focus.