Dictate
Speak directly into your Windows workflow
Hold your shortcut, speak naturally, and release. EYRO transcribes locally and returns the words to the active text field in email, documents, messaging, task tools, and browser editors.
Complete feature set
Dictation, writing review, and personal vocabulary—working together on your PC
EYRO is more than speech-to-text. It helps you capture a first draft, clean it, review it, remember the terminology you approve, and move through Windows with fewer interruptions.
Review
See writing issues before changing anything
Local suggestions underline possible spelling and grammar issues. For deeper work, review selected writing with an optional local model and inspect every proposed change before replacement.
Personalize
Teach EYRO the language of your work
Add names, technical terms, acronyms, and spoken-to-written aliases. EYRO prioritizes useful vocabulary and can remember a compact correction only after you confirm it.
Accelerate
Use the best reliable engine on your PC
EYRO calibrates local CPU and supported acceleration paths, keeps the speech engine warm, and falls back safely when a faster backend is temporarily unavailable.
Voice workflow
A faster way to get words where they belong
The floating companion stays available over your work without becoming another editor you must manage.
01
Hold-to-talk
Hold Ctrl + Win while speaking and release to transcribe and insert. Because recording duration—not processing time—drives Speaking WPM, the metric reflects how quickly you speak.
02
Hands-free dictation
Use the three-key lock shortcut for longer passages, then press it again to stop. The default shortcut family can be changed in Setup.
03
Floating status companion
See listening, transcribing, finalizing, and recovery states at a glance. The control stays above other windows and follows the screen containing your cursor.
04
Cross-app insertion with a safe fallback
EYRO returns text to the field you were using. If Windows or an application blocks insertion, the transcript remains available to paste instead of being lost.
Writing quality
From rough speech to writing you can confidently send
EYRO separates automatic cleanup from optional review, so speed never turns into invisible rewriting.
01
Configurable speech cleanup
Control fillers, stutters, self-corrections, punctuation, capitalization, web addresses, acronyms, units, mathematical symbols, Greek letters, and ending punctuation.
02
Local writing suggestions
A fast offline checker highlights possible spelling and grammar issues after transcription. Apply one suggestion, ignore it for that note, or add a legitimate word to your dictionary.
03
Private selected-writing review
Select a sentence or longer passage in a Windows editor and start Review selected writing. The optional Qwen3 4B model runs locally after a one-time verified download.
04
Change-by-change approval
A scrollable comparison underlines additions, replacements, and removals. Move through each change, keep the original, or click Replace selection only after you approve the result.
05
Protected details
Review safeguards reject blank output, excessive rewrites, and proposals that alter protected numbers, email addresses, links, mentions, hashtags, or identifier-like text.
Personal memory
EYRO adapts to your terminology without training on you
The speech model stays frozen. Your local vocabulary and explicit corrections make the system around it more useful over time.
01
Written words and spoken aliases
Save a correct spelling by itself, or tell EYRO that a phrase such as “night quest” should become “Nyquist.” Entries remain editable and removable.
02
Relevant-term prioritization
Personal terms are ranked by source, confirmed use, and recency so the highest-value vocabulary receives recognition priority within the speech model’s prompt limit.
03
Full-dictionary support after recognition
The complete local vocabulary remains available to EYRO’s deterministic correction and writing-suggestion layers, rather than becoming useless when the speech prompt is full.
04
Confirmed correction memory
When you make one compact correction inside EYRO, it can offer Remember this. Nothing is learned globally, uploaded, or saved without your approval.
05
Specialist vocabulary profiles
Choose the field closest to your work to activate relevant terms, then refine the result with your own names, products, and professional language.
Voice actions
Use your voice for more than inserting text
A related three-key shortcut can route a locally transcribed phrase to the destination you intentionally choose.
01
Direct web search
Hold the voice-destination shortcut, speak, and release to search in your default browser. Choose Google, Bing, DuckDuckGo, Yahoo, Brave Search, or a compatible custom HTTPS search address.
02
ChatGPT or Claude destination
Optionally route the finished prompt to the ChatGPT or Claude Windows app. EYRO verifies the destination before insertion and never sends microphone audio to the assistant.
03
Customizable shortcut family
Choose a two-key dictation shortcut, then separate third keys for hands-free mode and your voice destination. The family structure keeps the actions easy to remember.
Local workspace
A record of your work that remains yours
EYRO’s workspace helps you find and refine previous dictation without requiring a cloud account.
01
Searchable transcription history
Keep local notes, search titles and transcripts, filter by date or favourites, copy useful text, and delete entries you no longer need.
02
Speaking activity
Track total words, words today, speaking WPM, and your activity streak. Reset only the Speaking WPM measurement whenever you want a fresh comparison.
03
Undo and recording controls
Clear a recording, undo the latest insertion, switch English US or UK, and see live word and recording-duration feedback from the main recorder.
04
Guided first run
A short, replayable walkthrough introduces dictation, hands-free mode, voice search, selected-writing review, and the personal dictionary.
Models & reliability
A performance path for different Windows PCs
Choose the recognition quality your hardware can sustain while EYRO manages the local compute path underneath it.
01
Efficient
The included Efficient English model is designed for everyday CPU-only and integrated-graphics laptops.
02
Recommended
Download the Recommended English model once for a stronger quality-and-resource balance, then continue using it fully offline.
03
Maximum
Choose EYRO’s full-precision Maximum model on powerful hardware when your priority is the highest available recognition quality.
04
Hardware-aware calibration
Setup compares repeated local runs and can select CPU, broad Vulkan acceleration, optional NVIDIA CUDA, or supported Intel NPU acceleration when it proves reliable on that PC.
05
Continuity under failure
If an accelerator errors, times out, or resets, EYRO retries the same retained recording on a verified fallback and tests the faster path again later.
06
Warm engine and conservative VAD
The speech engine stays loaded between recordings. Longer, pause-heavy audio may use local voice activity detection, while uncertain results retry without it to protect recognition.
Clear boundary
Core dictation and writing review stay local
Internet access is used for downloads, licensing, purchase services, updates, and an explicit web-search or AI-destination action. Normal audio, transcripts, notes, and dictionary entries are not sent to CAWOSH.
Let EYRO prove itself before you pay for it.
Start with 2,500 successfully transcribed words every seven days. Keep using the free allowance indefinitely, or pay $19.99 once when the weekly limit becomes the problem.