Speak freely.Your words stayyours.
DictateAI turns speech into precise text at whisper-class accuracy — in your language, in any application. Audio is processed in seconds, then deleted. Nothing is stored, nothing is trained on.
Dictation software that treats your speech as private correspondence, not as raw material.
Most dictation tools record you and keep the tape. DictateAI inverts that: your audio exists for the seconds it takes to transcribe, then it is gone. The transcript is yours, in the file you asked for.
Your voice is not a dataset.
Audio is transcribed in seconds, then deleted. We never store recordings, never train on your words, never sell anything.
Faster than you can type.
Whisper-class transcription streams back in near real time — no queue, no waiting room, no batch processing.
One shortcut, every text field.
Editors, terminals, browsers, chat. DictateAI writes where your cursor already is.
Five commitments,
written as constraints.
Each one is enforced in the product, not in a press release. If a release can't satisfy them, the release doesn't ship.
- P1
Audio is deleted after transcription
Recordings are processed in seconds and then permanently deleted. There is no tape, no archive, no copy in a bucket somewhere.
- P2
No training on your words
Your audio is never used to train or fine-tune models. Transcription is a service, not raw material.
- P3
No selling, no sharing
We do not sell data, share recordings with advertisers, or build profiles from what you dictate.
- P4
Your transcripts stay yours
The text lands in your file, on your device, in the application you were using. We keep no copy of the content.
- P5
Verifiable, not merely stated
A published privacy policy, a data-retention schedule, and reproducible checksums so a sceptic can confirm the claims themselves.
- 01Hold the key
- 02Whisper-class engine
- 03Text lands in place
- 04The audio ceases to exist
Hold the key
A single global shortcut opens the microphone. A hairline indicator tells you precisely when audio is live — and it is never live otherwise.
Input · 16 kHz mono buffer
Try transcription live.
Hold the button, speak, release — hear the engine work. Audio is processed and discarded, nothing is stored.
Tap to start · tap again to transcribe
Everything a professional dictation tool should do — and nothing it shouldn't.
Custom vocabulary
Teach it names, acronyms and jargon once. Stored as a plain local file you can read, edit and version.
Local dictionary
Commands, not corrections
Say “new paragraph”, “scratch that”, “code block”. Editing happens as you speak instead of afterwards.
Inline control
41 languages, auto-detected
Switch languages mid-sentence. The engine detects what you're speaking and transcribes it accurately, no settings to change.
Multilingual
Formatted for the field
Terminal, email, code editor, chat. Punctuation and casing adapt to where the cursor is.
Context aware
Quiet by design
One hairline indicator, no floating windows, no sounds. The interface disappears while you work.
Interface
Fast on any hardware
Transcription runs in the cloud on whisper-class models — your laptop's CPU and battery are untouched.
Footprint
A privacy claim is only worth the proof behind it.
Audio is encrypted in transit, transcribed, and deleted. No recordings are stored, no transcripts are kept on our servers, and nothing you say is used for training. That is the contract, and it is the one we design against.
Read the privacy principles- Privacy policy
- Published, versioned, and reviewed on every release
- Audio retention
- Deleted after transcription. No archive, no backups of recordings
- Telemetry
- Anonymized page views only — no cookies, no IPs stored, no personal data
- Data retention
- Transcripts are not stored server-side; only usage counters for billing
- Dependencies
- Audited quarterly; no analytics SDKs of any kind
- Licence
- Perpetual plans; your transcripts and vocabulary stay yours