llms.txt
Compare with other sites' llms.txt files →# 1AudioTool
> Dictation that learns the words only you say.
A local-first voice layer for macOS. It runs Whisper on your own machine, biases it with your personal dictionary and your project's identifiers, remembers the corrections you make, and measures whether accuracy actually improved. No account, no API key, nothing uploaded.
Platforms: macOS, Windows
Publisher: StoicSoft (https://stoicsoft.com)
## Pages
- [Home](https://1audiotool.com/): what the product is and who it is for
- [Features](https://1audiotool.com/features): every capability, in detail
- [Free tools](https://1audiotool.com/tools): what it does, task by task, on the free tier
- [Compare](https://1audiotool.com/compare): side by side with the tools people shortlist
- [Alternatives](https://1audiotool.com/alternative): switching guides, including when not to switch
- [Pricing](https://1audiotool.com/pricing): what it costs and what is free
- [Download](https://1audiotool.com/download): builds and availability
- [Changelog](https://1audiotool.com/changelog): every release, mirrored from the signed GitHub releases
- [Blog](https://1audiotool.com/blog): notes on local speech recognition and correction memory
- [Privacy](https://1audiotool.com/privacy): what is stored and what is sent
- [Terms](https://1audiotool.com/terms): licence terms
## Capabilities
- Fully local dictation with whisper.cpp — no account, no API key, no upload
- Global double-tap Option gesture and a floating overlay over any application
- Auto-copy to the clipboard and auto-paste at the caret in the app you came from
- Speaker profiles isolating languages, model, microphone, vocabulary and corrections
- Personal dictionary and gitignore-aware project context packs
- Correction memory with scoping, merging and a measured effect on error rate
- Recheck, which separates model adaptation from correction-rule rewriting
- A personal accuracy benchmark: word error rate, project-term recall and latency
- Up to three spoken languages per profile for mid-sentence code-switching
- Three Whisper models downloaded in-app: base.en, small and large-v3-turbo
- Verbatim, Clean, Prompt, Message and Command output modes
- Voice actions with four permission levels, previews and an audit trail
- Ask memory, Listen digests and spoken responses through the local system voice
- A phone as a microphone over the local network, paired by QR code
- Insights: speaking rate, applications, languages, streaks and correction effect
- Encrypted profile export and a full install backup that never includes audio
## Honest limits
- Transcription quality is Whisper's, and Whisper is measurably worse on accented speech that humans understand perfectly. The app measures your voice rather than asserting anything about it.
- Correction memory is deterministic phrase rules plus a decoder prompt. It cannot change Whisper's acoustic model and is never described as accent training.
- Speaker separation runs after a recording stops, not live.
- There is no mobile app. A phone can act as a microphone over the local network; it is not a client.
- No meeting bot, no calendar integration and no shared workspace — this records what your machine can hear.
## Free tools
- [Dictate into any app](https://1audiotool.com/tools/dictate-anywhere-on-macos): Double-tap Option from wherever you are. The overlay appears over the app you were in, and the text lands at your caret.
- [Transcribe a meeting](https://1audiotool.com/tools/transcribe-a-meeting-locally): Listen records, separates speakers after the fact, builds chapters from your markers, and lets you redo the whole pass with a bigger model.
- [Teach it your vocabulary](https://1audiotool.com/tools/teach-dictation-your-vocabulary): Correct a transcript once and the phrase is remembered for that speaker — as a readable rule you can scope, merge, edit and forget.
- [Dictate in several languages](https://1audiotool.com/tools/dictate-in-multiple-languages): Up to three languages at once on a multilingual model, with the pack for each pair a download you take deliberately.
- [Run speech-to-text offline](https://1audiotool.com/tools/run-speech-to-text-offline): whisper.cpp on a model file you downloaded. No account, no API key, no localhost server, and no audio leaving the machine.
- [Use a phone as a mic](https://1audiotool.com/tools/use-your-phone-as-a-microphone): A QR code on the local network, a pairing session with a clock on it, and your phone becomes the input device.
## Feature pages
- [Dictate from any app](https://1audiotool.com/features/dictate-from-any-app): A global shortcut that records over whatever you were using and delivers the text back to it, without switching to a dictation window.
- [Six output modes](https://1audiotool.com/features/output-modes): Verbatim, Clean, Polish, Prompt, Message and Command — each one explains what it does to your words before you spend a recording finding out.
- [The recording overlay](https://1audiotool.com/features/floating-recording-overlay): A floating panel that shows listening, processing, result, destination and error over whatever app you were in — with live words while you are still speaking.
- [Phone as a microphone](https://1audiotool.com/features/phone-as-microphone): Pair a phone over the local network with a QR code and dictate into your Mac from across the room.
- [Microphone picker and pinning](https://1audiotool.com/features/microphone-picker-and-pinning): Every input listed by name and type with a live level meter, and a pin that stops a device connecting mid-session from moving the default out from under a recording.
- [Correction memory](https://1audiotool.com/features/correction-memory): Correct a transcript once and the phrase is remembered for that speaker — as a readable rule with its uses, its source project and its measured effect on error rate.
- [Personal dictionary](https://1audiotool.com/features/personal-dictionary): Your own identifiers, filenames and jargon, biased into the decoder as an initial prompt so unusual spellings become likely instead of impossible.
- [Project context packs](https://1audiotool.com/features/project-context-packs): Import a project folder and it learns that project's vocabulary (filenames, symbols, branches) so an identifier comes out spelled right the first time, with a separate pack per project and several in use for dictation at once.
- [Recheck reports](https://1audiotool.com/features/recheck-reports): Re-runs your own terms against stored audio and separates the model genuinely hearing more from text being rewritten afterwards — and reports both.
- [Room quality warnings](https://1audiotool.com/features/room-quality-warnings): Notices music or a TV behind you, a room your voice comes back off, and a background sitting nearly as loud as you are — and says so once, quietly, without stopping the recording.
- [Speaker separation](https://1audiotool.com/features/speaker-separation): A recorded meeting shows who said what. Voices are told apart on this Mac after it stops, each turn attributed, and renaming a speaker renames them everywhere.
- [The meeting page](https://1audiotool.com/features/meeting-page-and-chapters): A recording gets a page of its own: playback that every speaker turn seeks, markers that become chapters, and a second pass that rebuilds everything with a bigger model.
- [Live translation](https://1audiotool.com/features/live-translation): Talk in one language and the captions arrive in another, about a second behind your voice, translated on this machine during the pause after each sentence.
- [Captions, on screen and on the network](https://1audiotool.com/features/captions-window-and-network-captions): A captions window at reading-across-a-room size, and a read-only page anyone on the same Wi-Fi can open from a link or a QR code.
- [The model shelf](https://1audiotool.com/features/model-shelf): A Settings tab listing every speech pack the app can use, with its size, a line on what that size buys, a Download button and progress, plus Import model file… for a model you fetched yourself.
- [Local-only processing](https://1audiotool.com/features/local-only-processing): Audio goes from the system WebView to Rust over Tauri IPC, then through local whisper.cpp. No Electron, no Node runtime, no localhost server, no API key and no account.
- [Speaker profiles](https://1audiotool.com/features/speaker-profiles): Languages, model, vocabulary, correction history and metrics, isolated per speaker — and exportable without carrying your audio or your licence key.
- [Updates you read before you take](https://1audiotool.com/features/updates-you-read-first): A newer signed release opens a window with the version, its date and the full release notes — headings, bullets, emphasis and screenshots — then offers one button.
- [Transcribe audio and video files](https://1audiotool.com/features/transcribe-audio-files): Drop a WAV, MP3, M4A, FLAC, OGG or AIFF file, or the audio track of an MP4, MOV or WebM, on Review and it becomes a clip, transcribed on this machine with your model, your languages and your corrections.
- [SRT and VTT captions from any clip](https://1audiotool.com/features/srt-vtt-caption-export): Every clip in Review has an Export menu that copies SRT or VTT captions with the clip's timings, and Plan dub in Studio hands the transcript to the dub planner.
- [Music pauses while you dictate](https://1audiotool.com/features/pause-music-while-dictating): The moment the recording overlay appears, whatever is playing is paused, so a chorus never lands in your transcript, and a player you had already paused is never started again.
- [Correction intelligence](https://1audiotool.com/features/transcript-analysis): Review's Analysis tab reads what you corrected on this device, shows what is improving and which mistakes keep coming back, and proposes fixes that do nothing until you apply them, with an optional deeper pass on a local model or a coding CLI.
- [Audio verification](https://1audiotool.com/features/audio-verification): Before a transcript is pasted, the passages most likely to be wrong, such as the seams between sentences and words the model was unsure of, are listened to again, and the verified text is what gets delivered. It is an opt-in preview while it is measured on real voices.
- [Transcript change review](https://1audiotool.com/features/transcript-change-review): Every change made to your words, whether by correction memory, a context pack, Clean or Polish, is listed beside the verified transcript with the reason it happened, and each one can be kept, reverted or made permanent.
- [Local Polish](https://1audiotool.com/features/local-polish): A Polish mode on Dictate tidies grammar with an instruction model you import and run on this computer, and checks that names, numbers, identifiers and meaning survive before anything is pasted.
- [Styles by application](https://1audiotool.com/features/styles-by-application): Give each app its own output mode and tone, such as Polish for email and Clean everywhere else, so dictation comes out right for wherever you are typing without switching modes by hand.
- [Run in background](https://1audiotool.com/features/run-in-background): Hide the window and keep dictating: the shortcut still works from any app, and the menu bar or tray icon brings the window back as you left it.
- [Voice API](https://1audiotool.com/features/voice-api): Your cloned voices, served to other programs on this computer through an ElevenLabs-compatible API. Code written for the hosted service works by changing only the base URL, and the speech is made on the device.
- [Train a voice, and see how close it sounds](https://1audiotool.com/features/voice-cloning): Clone a voice from up to five of your own samples on this computer, then see a score for how close each clone sounds to its reference before a script ever calls it.
## Comparisons
- [1AudioTool vs Apple Dictation](https://1audiotool.com/compare/apple-dictation): The dictation already on your Mac — press the microphone key, talk, text lands in the field.
- [1AudioTool vs Wispr Flow](https://1audiotool.com/compare/wispr-flow): Cloud dictation that returns polished writing, not a transcript — the category leader.
- [1AudioTool vs superwhisper](https://1audiotool.com/compare/superwhisper): Local-first Whisper dictation for Mac, Windows and iOS, with a mode for every context.
- [1AudioTool vs Voice In](https://1audiotool.com/compare/voice-in): A browser extension that types with your voice in text fields on the web.
- [1AudioTool vs MacWhisper](https://1audiotool.com/compare/macwhisper): On-device transcription for recordings — drag audio or video in, get a clean transcript out.
- [1AudioTool vs Aiko](https://1audiotool.com/compare/aiko): Whisper on-device for recordings — Sindre Sorhus's minimal transcription app.
- [1AudioTool vs Otter.ai](https://1audiotool.com/compare/otter): The cloud meeting assistant — a bot joins the call and comes back with notes.
- [1AudioTool vs Dragon Professional](https://1audiotool.com/compare/dragon): The incumbent — deep command-and-control and industry vocabularies, on Windows.
## Switching guides
- [Switching from Apple Dictation](https://1audiotool.com/alternative/apple-dictation): includes when not to switch.
- [Switching from Wispr Flow](https://1audiotool.com/alternative/wispr-flow): includes when not to switch.
- [Switching from superwhisper](https://1audiotool.com/alternative/superwhisper): includes when not to switch.
- [Switching from Voice In](https://1audiotool.com/alternative/voice-in): includes when not to switch.
- [Switching from MacWhisper](https://1audiotool.com/alternative/macwhisper): includes when not to switch.
- [Switching from Aiko](https://1audiotool.com/alternative/aiko): includes when not to switch.
- [Switching from Otter.ai](https://1audiotool.com/alternative/otter): includes when not to switch.
- [Switching from Dragon Professional](https://1audiotool.com/alternative/dragon): includes when not to switch.
## Releases
- [v0.7.0 — One StoicSoft bundle key, and caption times fixed for long recordings](https://1audiotool.com/changelog/v0.7.0)
- [v0.6.0 — A local Voice API for your cloned voices, trained from your own samples](https://1audiotool.com/changelog/v0.6.0)
- [v0.5.0 — Transcripts checked against the audio, Polish on this Mac and a reason for every change](https://1audiotool.com/changelog/v0.5.0)
- [v0.4.0 — Audio file import, captions from any clip and speech models you bring yourself](https://1audiotool.com/changelog/v0.4.0)
- [v0.3.0 — Live translation on this machine, a speech model shelf and updates you read first](https://1audiotool.com/changelog/v0.3.0)
- [v0.2.0 — Speaker separation, a microphone picker and sentence-by-sentence transcription](https://1audiotool.com/changelog/v0.2.0)
- [v0.1.0 — Local dictation, nine spaces and correction memory](https://1audiotool.com/changelog/v0.1.0)
## Blog
- [Your cloned voice, served on your own computer through an ElevenLabs-compatible API](https://1audiotool.com/blog/local-elevenlabs-compatible-voice-api): 1AudioTool 0.6.0 lets you train a voice from your own samples, see how close the clone sounds, and call it from any ElevenLabs client by changing only the base URL. The speech is made on your device.
- [Your transcript, checked against the audio, with a reason for every change](https://1audiotool.com/blog/check-transcripts-against-audio): 1AudioTool 0.5.0 can listen again to the passages a transcript usually gets wrong before it pastes, then shows every change to your words beside what made it, so you can keep it, revert it or make it a rule.
- [Turn any audio file into a transcript that already knows your words](https://1audiotool.com/blog/transcribe-audio-files-locally): Drop a recording on Review and it is transcribed on your machine with the same model, vocabulary and correction rules as your dictations, then copy SRT or VTT captions straight from the clip.
- [Live captions in another language, translated on your own machine](https://1audiotool.com/blog/live-translation-on-your-machine): Talk in one language and captions arrive in another about a second later. The translation runs on your computer during the pause after each sentence, and the transcript stays what you actually said.
## Questions
### Does anything actually leave my machine?
No audio and no transcripts, ever. Audio goes from the system WebView directly to Rust over Tauri IPC, then through local whisper.cpp. There is no Electron, no Node runtime, no localhost server, no API key and no account. A packaged build makes exactly two kinds of network call, both of which you can see: a signed update check, and — only if you buy Pro — licence activation.
### Which model should I use?
large-v3-turbo (1.6 GB) if you can spare the disk: it is still faster than realtime on Apple silicon, at roughly 5×, and it is substantially better on accented speech and uncommon terms. small (466 MB) is a good multilingual default. base.en (141 MB) runs at about 28× realtime and is noticeably rougher — measured here, it heard “audio workload… root the PCM” where large-v3-turbo heard “audio worklet… route the PCM”.
### Is this accent training?
No, and the app never says it is. Correction memory is deterministic phrase rules plus a decoder prompt — it cannot change Whisper's acoustic model. A true fine-tune would need consented audio with reference transcripts, a training pipeline, and model conversion and serving. What the app does instead is measure: Recheck separates the model genuinely hearing more from text being rewritten afterwards, and reports both.
### Will it work with my accent?
Whisper is measurably worse on accented speech that humans understand perfectly, and no product should promise otherwise. So 1AudioTool measures your voice rather than asserting anything about it: the benchmark scores word error rate, project-term recall and latency on your own speech, and Recheck tells you which of your terms are genuinely learned, which are being carried by a rule, and which are unstable. Explicitly selecting English rather than Auto-detect is the single biggest fix for non-native speakers on short clips.
### Can I mix two languages in one sentence?
Yes — select up to three spoken languages on the profile. Mixed speech decodes with a multilingual model, auto-detect, and a short multilingual prompt built from native-script anchors, because whisper.cpp can only lock one language token and locking the wrong one is exactly what garbles the other language. Prefer large-v3-turbo for code-switching. English-only models cannot mix, and the picker says so instead of failing quietly.
### Does it work offline?
Entirely, once a model is downloaded. That includes Pro: entitlements are verified locally against a signature, with a 14-day grace window so an offline stretch or a transient failure never costs a paying customer their features.
### What is the state of Windows support?
Written, but not yet shipped. On Windows the gesture is Control twice, because a lone Alt press opens the menu bar there. That gesture, insertion at the caret in the app you were last in, spoken responses through the system voice and the folder picker all run natively on Win32, and the release workflow can publish an NSIS installer alongside the macOS bundles. It is not in a public release yet, so this site does not offer it as a download. One caveat for when it lands: the installer is not Authenticode-signed, so SmartScreen will warn on first run.
### Can I move to a new machine?
Yes. Release the device under Settings → Licence, then activate on the new one. Profiles export as JSON — or as an AES-GCM encrypted file — and Pro can export the entire install, every profile and setting, as one file. Exports deliberately exclude microphone recordings, and never contain the licence key.
### Why a separate app instead of a feature inside something else?
Because
