Stop typing what you can say.
Echo is push-to-talk dictation for Windows. Hold a key, say it the way it comes out of your mouth, release. Clean, finished text lands in whatever app you were typing in. And when a call starts, Echo takes the notes.
Hold. Talk. It lands.
Echo lives in your tray and stays out of the way. The only interface you see day to day is a small pill at the bottom of your screen that tells you what the mic is doing.
Hold a key
Default is Pause, or pick any key or chord you never use. The pill expands and the wave starts moving: that is the only time the mic is live.
Say it messy
Talk the way you actually talk. Ramble, restart sentences, think out loud. Echo transcribes with Whisper-class accuracy in 90+ languages.
Release
The cleanup pass fixes grammar, drops the filler and formats spoken lists as real lists. Finished text pastes at your cursor, in the app you were in.
One pill. Dressed your way.
This is the live overlay again, not a video. The wave swaps in place from the Appearance page while the pill keeps running. Rapids is the flagship; Cloud is the calm one.
It types like you on your best day.
Dictation tools give you a transcript. Echo gives you the message you meant. The cleanup pass is app-aware: an email reads like an email, a terminal command stays a command, a code comment keeps its backticks.
"um so basically I think we should uh move the launch to friday because the the demo isn't I mean the demo needs another day and um also can you tell walter to to check the feed first thing"
I think we should move the launch to Friday. The demo needs another day. Also, can you tell Walter to check the feed first thing?
The dashboard behind the pill.
Everything Echo hears and learns lives in one window: your day at a glance, every dictation with its raw transcript, and the corrections Echo taught itself from your edits. These are real screens, in the dark theme; a clean light one is a click away in Settings.
It learns from every fix
Edit a dictation once and Echo keeps the correction: "EMV file" becomes ".env file" forever, per app profile, with a count of every time it saved you.
A dictionary that fills itself
Echo notices what Whisper mishears and queues it as a one-click chip. Your jargon, product names and replacements ride every transcription after that.
Echo takes the notes.
A call comes to the front and Echo offers. It hears your mic and the other side through your speakers, tells the voices apart, and when you stop the note is there: a summary, the decisions, the next steps with who owns them, the open questions, and the whole transcript with every line timed and named. Nobody joins the call and no audio is kept.
Who said what
Every voice gets its own label, on your mic and on the call. Name a voice once and Echo knows it next time; your own voice it learns from your dictation.
A library you can scan
Meetings grouped by day with folders, tags, people, stars and open to-dos. Search every word ever said and land on the line, or ask a question and get the meeting and time cited.
Notes you can act on
Type what matters while it listens and Echo builds the note around it. Pick a shape (standup, one on one, sales call, brainstorm, interview) and re-run it any time. Every note is a Markdown file.
Every meeting, one row
The time, the title, a line of the summary, who was there, where the call happened, the folder and the to-dos. Six fit where two used to.
- Nobody joins your call. No bot in the participant list, no calendar access, no account to connect. Echo listens on your machine, in any call app or in person.
- Speakers are fine. Echo removes the mic's echo of the far side, so the other person is kept once, on their own track, with or without a headset.
- The room stays out of the call. A TV or the kids in the background are not part of the notes, one tap away if you want them.
- Dictation never stops. Hold the key mid-meeting and dictate into the chat. The pill shows a Notes clock while Echo listens and goes back to it after every paste.
- Stops when the call does. The call window closes, or everyone goes quiet, and Echo writes the note. A dropped call picks up in the same note.
Everything a daily driver needs.
Meetings
Echo listens to your mic and the call, tells the voices apart, and writes the note: summary, decisions, next steps with owners, open questions, and the full transcript, searchable across every meeting.
Dictionary that learns
Echo notices the names and jargon it misheard, suggests them as one-click chips, and never mishears them again.
Snippets
Say "sign off" and your full signature block lands. Your boilerplate becomes a spoken word.
Scratchpad
Think out loud into Echo itself. Blocks pile up as you talk; copy them out when the thought is finished.
App-aware profiles
Echo reads the window it is pasting into. Slack gets a casual register, email gets structure, VS Code keeps filenames and flags intact.
History with receipts
Every dictation keeps the raw transcript above the cleaned text, so you always see exactly what the cleanup changed. Repaste or transform any entry.
90+ languages
Pin one language or let Echo detect per dictation, so you can answer your mom and your standup in the same minute.
Hands-free
Double-tap the key to lock the mic, tap once to stop. Think out loud for minutes at a time without holding anything down. The pill shows an honest Hands-free state the whole way.
Fix last
One key undoes your last paste, lets you edit the text, and pastes it back. The fix also teaches Echo, so the same mishear stops coming back.
A tool you own, not a feed you're on.
Echo is built by one person for people he knows. That shapes everything about how it treats your data.
- Audio is never kept. Your voice is transcribed and discarded. A meeting is held on your disk only in short pieces until each one is transcribed, then deleted. There is no recording library, on your disk or anywhere else.
- No telemetry, no analytics, no account in the app. Echo phones home for exactly one thing: checking whether a new version exists. (This website is different: it uses Google Analytics, Microsoft Clarity, and the LinkedIn Insight Tag to see how visitors use these pages, as the xsvStudio Privacy Policy describes.)
- History lives on your machine. The searchable dictation history is a local database in your own AppData folder. Delete it whenever you want.
- A local fallback exists. Prefer nothing leaving the machine? Echo can transcribe fully offline with a local model, trading some speed for total privacy.
07:41:02 PTT press target=Slack 07:41:06 release 4.1s audio 07:41:07 stt 41 words groq-whisper 07:41:07 audio discarded 07:41:08 cleanup profile=slack-casual 07:41:08 paste landed at cursor 07:41:08 telemetry none. there is none.
Anyone whose thoughts outrun their hands.
Get Echo.
Echo is invite-only. Send one sentence about what you'd use it for, and you'll get a personal reply with a download link and your activation key. No list, no drip campaign, a person answers.
