Everything 1AudioTool does: local Whisper transcription, speaker profiles, a personal dictionary, project context packs, correction memory, voice actions, and an accuracy report that tells you whether any of it worked.
Press Control twice from any application, hold Space in the window, or tap the menu-bar icon. The overlay appears over whatever you were using. Esc cancels and writes nothing.
AudioContext created at 16 kHz — whisper's native rate, so ffmpeg is never involved.
DC offset removed, outside silence trimmed, conservative gain applied. Quiet, clipped or mostly-silent input is reported as a signal problem instead of looking like a model failure.
Resampling uses OfflineAudioContext with a low-pass area fallback — no aliased noise in the speech band.
The clip becomes Int16 PCM with a WAV header, off the main thread, and crosses to Rust as a raw binary IPC body — no base64, no JSON array of a million numbers.
The temporary WAV is deleted after every request.
A Rust worker runs whisper.cpp against the model on your disk. Your personal dictionary and the active project pack go in as the initial prompt, so unusual spellings become likely instead of impossible.
base.en 141 MB · small 466 MB · large-v3-turbo 1.6 GB — downloaded in-app, switched per profile.
Recurring deterministic mistakes are rewritten by rules you taught this speaker. Every rule records its uses, its source project and its measured effect on error rate — and overly broad rules get flagged.
Never described as accent training. It cannot change Whisper's acoustic model, and it doesn't claim to.
The transcript is copied, or pasted at the caret in the app you came from once focus is restored. In Command mode it becomes a parsed intent with a preview — never a silent execution.
Auto-paste needs macOS Accessibility, which the app asks for rather than assuming.
The recorder. Push-to-talk, live transcript, output mode, destination, and the correction you make right after.
What still needs you: clips awaiting review, proposed spelling rules, and imported sources.
Every transcript you have ever dictated — searchable, filterable, replayable, and correctable.
The personal dictionary, and the project packs that make your codebase's vocabulary spellable.
What dictating has actually done for you — counted from what is already on this machine. Free at every tier.
Voice commands with a preview, a permission level, and an audit trail. Voice is never authorisation.
The day, read back to you — written first, so you can check it before you hear it.
Captions, translation and dubbing that start from the transcript you already corrected. The later milestone, shipping in stages.
Models, languages, microphone, storage, shortcuts, appearance, licence and updates.
Each space is broken down in full on the home page.
Items marked PRO need the one-time licence. Everything else is in the free tier, forever.
Free keeps one. Pro keeps as many as you need — work and personal, or everyone in the house.
Reshape what you said into a well-formed prompt, a written message, or a parsed action — locally, before it reaches the clipboard.
Point at a repository and 1AudioTool learns its identifiers, filenames and branch names.
Ask a question of everything you have dictated and get an answer that cites the transcript span it came from.
Turn a day of captures into a written digest you can read first, then have spoken back to you.
Free previews a parsed command and stops. Pro executes it, with the confirmation step and audit trail intact.
Pair a phone over the local network and dictate into this machine from across the room.
Watch the transcript build while you are still speaking instead of waiting for the clip to finish.
Move a profile to another machine as an AES-GCM encrypted file. Plain JSON export stays free.
Move every profile, transcript, dictionary, rule, context pack and setting to another machine.