A Quiet Dictation Keypad for Aider, Not a Silent One

WhisperKey costs $87.90 and is a clear handheld voice-workflow controller used near the mouth, with microphone, tick, cross and Enter legends. It may support a separate dictation path after verification. It does not make speech silent or prove an included or integrated microphone.

How is this different from Aider's built-in /voice?

Aider officially provides `/voice`, which starts recording and asks the user to press Enter when finished; installation and voice-input configuration are documented. WhisperKey is external hardware whose exact key output, audio path and compatibility are unconfirmed.

Test Aider's built-in feature first. Do not buy a controller to duplicate a workflow that has not proven useful.

How do you dictate an Aider prompt without the room hearing it?

You cannot guarantee that. Speaking near a microphone may allow a lower voice, but coworkers can still hear content and nearby microphones can capture it. Room acoustics, gain and recognition vary. A handheld controller is not a privacy device.

Use invented text during tests, follow workplace policy and type confidential prompts. A private room may be the only suitable voice setup.

What are the four keys for in a terminal workflow?

The legends suggest microphone, accept, reject and Enter, not verified outputs. A safe trial may use one key for push-to-talk or focus after key-down and key-up are confirmed. Accept, Enter and automatic submission should remain off handheld hardware.

Leave unused controls inactive and preserve a mouse or keyboard stop path.

Where does dictation still fall over on code words?

Identifiers, punctuation, casing, paths and commands demand exact recognition. A plausible transcript can still be technically wrong. Review every character before sending it to Aider, especially when shell text or file scope is involved.

The WhisperKey is worth evaluating only after `/voice` or another approved dictation path works. Confirm microphone, output and host details, and walk away if the office or privacy context makes audible prompting inappropriate. Compare three paths using the same harmless prose: typing, Aider `/voice` and any external dictation service. Count corrections, stop failures and privacy constraints, not imagined speed. Aider lets the input device and voice language be configured, so verify those official settings before adding another controller. Delete test audio or transcripts according to policy and never speak secrets during setup. Include awkward samples such as mixed-case names, dotted paths and spoken punctuation. A low error rate on ordinary sentences says little about code. Repeat the trial with the door open and normal background conversation, then ask another person whether the spoken prompt remained intelligible at their seat. That observation is more useful than the word “quiet.”