Getting Started
What is EchoKey?
EchoKey is a Windows voice dictation tool powered by OpenAI Whisper for transcription and Claude for text refinement. Hold a hotkey, speak, release — your words appear wherever your cursor is, cleaned up and ready to go.
Unlike browser-based dictation tools, EchoKey works system-wide across every application: code editors, email clients, Slack, browsers, terminals — anything with a text input.
EchoKey currently supports Windows 10 and Windows 11. macOS support is planned.
Installation
Download the installer
.exe installer.Run the installer
Add your OpenAI API key
(Optional) Add Anthropic key
Your first recording
Click into any text field — an email draft, a chat message, a document. Then:
- HoldAlt Right (default push-to-talk key)
- SpeakSay what you want to type, naturally.
- ReleaseEchoKey transcribes, cleans up, and types the result.
The first recording takes 1–3 seconds longer as the AI model warms up. Subsequent recordings are near-instant.
Recording
Push-to-talk
Push-to-talk is the default recording mode. Hold your hotkey to record, release to transcribe. The recording indicator appears on your screen while active.
To change the hotkey: Help → Shortcuts or Settings → General → Recording shortcut. Click the field and press any key combination.
| Default key | Right Alt (AltRight) |
| Max duration | 5 minutes (auto-stops) |
| Silence handling | Whisper handles silence — no need to fill gaps |
Hands-free mode
Hands-free mode lets you toggle recording on and off with a single key press instead of holding. Useful for longer dictation sessions.
| Default hotkey | Ctrl + Space |
| Start dictating | Press once to start |
| Stop & transcribe | Press again to stop |
| Change hotkey | Help → Shortcuts → Hands-free mode |
Microphone setup
EchoKey uses your system default microphone unless you specify one. To change it: Help → Microphone.
For best results use a headset or a dedicated USB microphone. Laptop built-in mics work but are more susceptible to background noise.
Whisper is robust to background noise. You don't need studio silence — a normal office environment works well.
AI Processing
Transcription (Whisper)
Your audio is sent to OpenAI's Whisper API for transcription. Whisper is a state-of-the-art speech recognition model that handles accents, background noise, and 99+ languages extremely well.
| Model | whisper-1 |
| Provider | OpenAI API |
| Cost | ~$0.006 per minute of audio |
| Languages | 99+ (see Languages section) |
| Audio format | WebM/Opus, 16kHz mono |
Text cleaning (Claude)
After transcription, Claude refines the raw text: removing filler words ("um", "uh", "like"), fixing punctuation, capitalising properly, and applying your Style preferences. This is what makes EchoKey feel like you typed it — not spoke it.
Text cleaning requires an Anthropic API key. If not configured, the raw Whisper transcript is used.
| Model | claude-haiku-4-5 (fast + cheap) |
| Cost | < $0.001 per typical dictation |
| Filler removal | Configurable in Settings → AI |
| Punctuation | Always applied |
The processing pipeline
The entire pipeline takes 1–4 seconds depending on audio length and API latency. Text is injected at your cursor position using clipboard paste or direct key simulation (configurable).
Features
Dictionary
The Dictionary teaches EchoKey words it might misspell — product names, technical terms, proper nouns, industry jargon. Every entry is automatically applied as a correction after transcription.
Adding words: Go to Dictionary in the sidebar. Click Add word. Enter the correct spelling. EchoKey will match common misspellings automatically.
Pro tip: add your name, company name, product names, and names of people you frequently mention. These are the words Whisper gets wrong most often.
Snippets
Snippets let you say a short trigger phrase and have EchoKey expand it into a full block of text. Great for email signatures, boilerplate responses, meeting templates, or code patterns.
Example: Set trigger my email → expands to your full email address.
| Trigger | The phrase you speak to activate the snippet |
| Expansion | The full text that replaces the trigger |
| Matching | Case-insensitive, partial matching at end of dictation |
Transforms
Transforms rewrite your last dictation using a hotkey. Hit the shortcut immediately after speaking and EchoKey replaces the typed text with an AI-rewritten version.
| Polish (Win+Alt+1) | Improve clarity and conciseness — ideal for emails |
| Prompt Engineer (Win+Alt+2) | Restructure as an optimised AI prompt |
| Custom slots (Win+Alt+3–9) | Your own prompts — create in the Transforms page |
| View diff (Win+Alt+O) | See before/after comparison for any transform |
Transforms require an Anthropic API key. You can create up to 7 custom transform slots with any prompt you like.
Style settings
Style lets you set how EchoKey formats text per app context. Configure different tone levels for personal messages, work messages, email, and other contexts.
EchoKey detects the active application and applies the matching style automatically — so casual Slack messages and formal emails get different treatment without any manual switching.
Notes
The Notes view lets you dictate and accumulate text in a persistent notepad, separate from injecting into other apps. Useful for journaling, brainstorming, or capturing thoughts without losing context.
Insights
Your Usage
The Your Usage tab shows your dictation statistics: words per minute, total words dictated, AI fixes made, time saved, desktop usage breakdown by app category, and a GitHub-style activity heatmap.
| WPM | Calculated from Whisper-reported audio duration vs word count |
| Streak | Consecutive days with at least one dictation session |
| Activity heatmap | Last 91 days, amber intensity scale |
| App categories | Code editors, AI prompts, Email, Messages, Browser, Documents |
Your Voice
The Your Voice tab unlocks at 500 words dictated. It generates a personalised AI voice profile describing your dictation patterns, a "catchphrase" based on your most-used phrase patterns, and your peak dictation time.
Your Voice uses Claude to analyse your dictation history. The profile is generated once and cached — it won't regenerate unless you reset it.
Sharing stats
Click the Share button in the top-right of the Insights page to open the share modal. You get 6 shareable stat cards:
- —EchoKey Highlights — overview of all 4 key stats
- —Total words dictated — with milestone context
- —Dictation Streak — your current streak
- —How fast you Flow — WPM + percentile gauge
- —Your dictation activity — heatmap card
- —Where you Flow — top app categories
Use Copy image to copy as PNG, Download to save, or share directly to X or LinkedIn.
Settings
API keys
| OpenAI API key | Required for Whisper transcription. Get one at platform.openai.com |
| Anthropic API key | Required for Claude text cleaning and Transforms. Optional but recommended. |
| AI provider | Choose OpenAI or Anthropic for text processing. Whisper always uses OpenAI. |
Hotkeys
Access via Help → Shortcuts or Settings → General. Changes take effect immediately — no restart needed.
| Push-to-talk | Default: Right Alt. Hold to record, release to transcribe. |
| Hands-free toggle | Default: Ctrl+Space. Press once to start, again to stop. |
| Transform 1 | Win+Alt+1 — Polish |
| Transform 2 | Win+Alt+2 — Prompt Engineer |
| View diff | Win+Alt+O — Show transform before/after |
Language
EchoKey supports 99+ languages via Whisper. Set your language in Help → Languages or Settings → General → Language.
Enable Auto-detect to let Whisper identify the language automatically — useful if you switch between languages frequently.
Text injection
EchoKey types text into your focused application using one of two methods:
| Clipboard (default) | Copies to clipboard, pastes with Ctrl+V. Works everywhere. May overwrite clipboard. |
| Key simulation | Simulates keystrokes character by character. Slower but clipboard-safe. |
Keyboard Shortcuts
Recording
| Alt Right (hold) | Push-to-talk — record while held |
| Ctrl + Space | Hands-free toggle — start/stop recording |
Transforms (after dictating)
| Win + Alt + 1 | Polish — improve clarity and conciseness |
| Win + Alt + 2 | Prompt Engineer — structure as AI prompt |
| Win + Alt + 3 – 9 | Custom transform slots |
| Win + Alt + O | View diff — before / after comparison |
Navigation
| Esc | Close any open modal |
| F12 / Ctrl+Shift+I | Open DevTools (developer mode) |
Team & Billing
Plans
Free
$0
- ✓2,000 words / week
- ✓Whisper transcription
- ✓Dictionary & Snippets
- ✓Style settings
Pro
$10/mo
- ✓Unlimited words
- ✓AI Transforms
- ✓Cloud sync
- ✓Priority support
- ✓Everything in Free
Team
$8/seat/mo
- ✓Everything in Pro
- ✓Shared dictionary
- ✓Shared snippets
- ✓Team management
Team setup
Create a team
Invite members
Shared resources
Troubleshooting
Common issues
Recording hotkey not working
Another application may be holding the key globally. Check Task Manager for apps that use global hotkeys (Discord, gaming software). Try changing your hotkey in Help → Shortcuts.
Text not appearing in my app
Try switching the injection method (Settings → General → Text injection) from Clipboard to Key simulation or vice versa. Some apps (terminals, games) require key simulation.
Transcription is slow or failing
Check your OpenAI API key in Settings. Verify you have API credits at platform.openai.com. If the key is valid, check your internet connection — Whisper requires an API call.
Text isn't being cleaned up / Claude errors
Verify your Anthropic API key in Settings. Text cleaning is optional — if unavailable, the raw Whisper transcript is used instead.
WPM shows "—" on the home screen
WPM is calculated from the duration reported by Whisper. It only populates after recordings that include duration metadata. Make a few new recordings and it should appear.
App shows 0 words after restart
This is usually a brief loading state — history loads from disk after the app starts. Wait 2–3 seconds and the stats should populate.
FAQ
Is my audio stored anywhere?
No. Audio is sent directly to the OpenAI Whisper API and is not stored by EchoKey. OpenAI's data retention policies apply.
Does EchoKey work offline?
No. Transcription requires an internet connection to call the Whisper API. A local offline mode is planned for a future release.
How much does it cost to run?
Typical usage costs $1–5/month in API fees (OpenAI + Anthropic). Heavy users dictating 100K+ words/month may spend $10–20.
Can I use it in languages other than English?
Yes. Whisper supports 99+ languages. Set your language in Help → Languages or enable Auto-detect.
Does it work in games or fullscreen apps?
EchoKey works in any app with a text input. Fullscreen games may require borderless window mode and clipboard injection.
How do I report a bug?
Use Help → Report an issue in the app, or email ambrose@nguzo.tech.
EchoKey by Nguzotech · Built with Claude
ambrose@nguzo.tech