Everything You Need to Dictate Effortlessly
From local AI transcription to intelligent post-processing, Dikt gives you a complete speech-to-text workflow that respects your privacy.
Powerful Transcription Engine
Multiple transcription engines working together to give you the best possible results, whether you're online or offline.
- Local Whisper.cpp with multiple model sizes (tiny to large)
- OpenAI Whisper API for cloud accuracy
- GPT-4o Audio for high-quality cloud transcription
- Real-time streaming preview — words appear as you speak
- Automatic failover between providers
- Retry with exponential backoff
- Multi-language support (100+ languages)
- Model download manager
- Audio dimming — automatically dims system audio during dictation
- GPU acceleration — run local Whisper with CUDA or DirectML for faster transcription (Auto/CPU/CUDA/DirectML)
- Noise profile learning — record 3 seconds of ambient noise; Dikt filters it out of future recordings using spectral subtraction
- Auto-correction suggestions — word-level confidence scores highlight uncertain words before injection so you can review and fix them
AI-Powered Cleanup
Raw transcription is just the beginning. AI post-processing transforms your spoken words into polished, ready-to-use text.
Set up offline AI cleanup- Grammar and punctuation correction
- Filler word removal ("um", "uh", "like")
- Voice commands ("period", "comma", "new line", "new paragraph")
- Editing voice commands ("scratch that", "delete last word", "select all")
- AI Command Mode — transform selected text with voice instructions (Ctrl+Shift+Space)
- Context-aware formatting adjusts tone by app (formal for email, casual for chat)
- Custom vocabulary for proper nouns, technical terms, and brand names
- Snippets — text shortcuts with {date} and {time} variables
- Profanity filter — automatically removes curse words with built-in + custom word lists
- Swear Jar — fun metrics tracking words cleaned, cost, and daily trends
- Powered by GPT-4o-mini and Claude Haiku
- BYOK: Use your own API keys, pay only for what you use
- On-device cleanup — an optional 462 MB model (SuperWhisper S1-mini) strips "um" and false starts, resolves mid-sentence corrections, and writes numbers, dates and emails properly, without sending a word anywhere (English only)
- Real-time translation — dictate in any language and have the output automatically translated (e.g. speak Spanish, get English text)
- Word correction training — define wrong→correct pairs (e.g. "mispelt" → "misspelled") applied as a post-processing pass before cleanup
- Markdown voice mode — say "heading one", "bullet point", "open code block" and get proper Markdown syntax; auto-enabled in Obsidian, Typora, and VS Code
- Custom AI persona library — choose from 6 built-in personas (Technical Writer, Doctor, Journalist, Lawyer, Code Reviewer, Casual) or create your own
- Multi-turn AI Command Mode — AI Command Mode remembers your last 3 instructions for back-and-forth editing; say "start over" to reset
Works Everywhere You Type
Dikt integrates at the system level so you can dictate into any application without switching windows or copying text.
- Global hotkey (customizable, default Ctrl+Alt+Space)
- Push-to-talk or toggle mode — choose your style
- Push-to-talk grace period — quick-tap toggles recording without holding
- Automatic text injection at cursor position
- Append mode for continuous dictation
- Works with Notepad, Word, Slack, VS Code, Chrome, email, any text field
- Obsidian integration — dictate directly into your vault notes
- Notion integration — push transcriptions to Notion pages via API
- VS Code extension — dictate code comments, docstrings, and commit messages
- Output templates — format with {text}, {date}, {time} placeholders
- Clipboard re-inject queue — when injection fails (no focused window), text is queued; click the overlay to inject into whatever window you focus next
- Code dictation mode — natural language dictation auto-formatted as code when dictating in an IDE (Rider, Visual Studio, Cursor, IntelliJ)
- Wake word / always-on listening — say "Hey Dikt" to start recording hands-free (uses Windows built-in speech recognition)
Privacy-First Design
Your voice data is yours. Dikt is built from the ground up to keep your data private and secure.
- Local-first: transcribe without any internet connection
- On-device AI cleanup — polish your transcript with a local model, so even cleanup never touches the cloud
- DPAPI encryption for API keys at rest
- No data sent to servers unless you choose cloud mode
- BYOK: your keys, your data, your control
- Atomic file writes prevent settings corruption
- Local-only mode cuts off everything that reaches the internet — and says exactly what it covers, with LAN casting a separate opt-in that stays off unless you tick it
- Per-device API keys — every PC gets its own key, revocable one at a time from your account page without touching the others
Cast to Your Other PCs
Dictate on the machine with the good microphone; the finished text lands on whichever PC you are actually working on. Pair once and forget about it.
Learn how to pair your PCs- Dictate on one PC and the finished transcript arrives on your other PCs running Dikt
- Pair once by comparing a six-digit number shown on both screens — nothing to type, nothing to copy between machines
- End-to-end encrypted with AES-256-GCM; keys are agreed directly between your PCs and never leave them
- Travels straight across your local network when a paired PC is reachable — no round trip to anyone's server
- dikt.app relay as a fallback when no PC is reachable (Pro) — it stores only ciphertext, for 60 seconds, and can never read your text
- Each PC decides for itself what arriving text does: ignore it, copy it to the clipboard, type it at the cursor, or hold it for review
- LAN casting works on the free tier with no account at all
- Received transcripts are never re-cast and never replay editing commands, so a "scratch that" can't edit the wrong document
Your Words, On Every Device
Optional cloud sync keeps your history and settings with you across machines — and lets you publish a feed of your dictations if you want to.
- Encrypted transcription history sync across all your PCs (Pro)
- Browse, search and export your full history from the web dashboard
- Cloud sync for vocabulary, snippets, context profiles, and custom profanity lists
- Device management — see every PC on your account and revoke any one of them instantly
- Optional public profile at dikt.app/u/your-handle, off by default
- RSS and JSON feeds for your public profile, so anything you publish can be followed anywhere
- Usage dashboard on the web with words, time, provider breakdown and cost estimates
Beautiful & Functional
A modern, polished interface that looks great and stays out of your way when you need to focus.
- Modern Fluent Design (WPF-UI)
- Light, dark, and system theme support
- System tray operation — runs quietly in the background
- Dashboard with usage charts and statistics
- Full transcription history with search and playback
- Export in TXT, CSV, JSON, SRT formats
- Daily dictation streak — track your streak and earn milestone badges
- Usage dashboard with words, time saved, provider breakdown, and streak stats
- Cloud sync for vocabulary, snippets, and profiles across devices (Pro)
- Voice profiles — multiple users per machine, each with own settings and history
- TTS preview before inject — hear your transcription read back before it's typed; choose Windows built-in voice or high-quality ElevenLabs neural voice
Accessible to Everyone
Designed to be usable by everyone, with full support for assistive technologies and flexible configuration.
- Full keyboard navigation
- Screen reader compatible
- Setup wizard for first-time configuration
- Notification overlay — configurable type (Minimal/Line/Box), style (Pill/Box/Text), 9 screen positions, multi-monitor display selector, and size scaling
- Compact overlay — persistent floating status window with live waveform indicators (8 styles), Box-mode transcription preview, controls, and history
- Compact overlay features — click-to-open, lock position, auto-hide, multi-monitor support, and size scaling
- Typing animation — letter-by-letter text display in overlays
- Overlay toggle in status bar for quick access
- Customizable notification styles (toast, popup, overlay)
Team Workspaces
Share vocabulary, snippets, and formatting profiles across your team. Everyone stays consistent, no one starts from scratch.
- Shared team vocabulary — proper nouns and jargon available to everyone
- Shared snippets — team-wide text shortcuts with variables
- Shared context profiles — consistent tone across apps
- Per-seat billing — add or remove members anytime
- Team admin controls — owners and admins manage shared settings
- Personal overrides — your personal settings always take priority
- Automatic sync — team settings pull alongside personal cloud sync
- Invite by email — members join with a simple invitation flow
Built to Be Automated
Dikt exposes a local API so your own tools, scripts, and editor plugins can drive dictation and read transcripts as they happen.
- Local HTTP API server on a configurable port — start and stop recording, read transcripts, drive Dikt from anything
- WebSocket support for real-time streaming, so your tool sees words as they are spoken
- Editor integration protocol for third-party editor plugins
- Official VS Code extension — dictate comments, docstrings, and commit messages
- Export history as TXT, CSV, JSON, or SRT for downstream processing
- Settings import and export as plain JSON, so a machine can be provisioned from a file
Ready to Get Started?
14-day free trial. No credit card required. All features included.