Your voice,
becomes anything.
Not a dictation app — an assistant that lives one keypress away, in every app you use.Speak, and clean, faithful text appears at your cursor. Ask a question out loud and hear the answer on the spot. Translate as you talk. Highlight anything and just say how to change it. Mail, Slack, docs, code, the browser — wherever the cursor blinks, your voice now does the work.
Don’t type, just say it.
Hold a key, talk, and polished text lands at your cursor — in any app.
Four keys. One that you'll
use a thousand times a day.
Each action gets its own hotkey — bind them however you like. Dictation is faithful to what you said. The other three only fire when you press them.
You speak about 5× faster than you type.
Most people type around 45 words a minute — and speak closer to 220. Saytomo closes the gap, and its dashboard quietly tallies the hours you get back.
Works in every app you write in.
No plugins, no special editor. The hotkey turns any text field into a voice field, and a quiet floating indicator shows it listening — it never steals your focus.
Clean text. Never reworded.
Cleanup is reductive: it only removes and restructures what you said. Fillers go, your spoken corrections are applied, punctuation appears — but your words, claims and meaning are left exactly as you spoke them. When in doubt, it keeps the original.
Two languages, one breath.
Kept exactly as spoken.
If you mix English into Mandarin — like everyone in tech does — Saytomo keeps every word in the language you said it in. It will never quietly “normalize” your sentence into one language.
Stuck on a word?
Just ask out loud.
Reading a paper and hit a term you don't know? Select it, hold Ask, and say “what does this mean?” The answer appears in a floating panel — your document untouched. No tab-switch, no chatbot, no losing your place.
§ 4.2 — Method
Measuring what each component contributes
An ablation study isolates what a single mechanism is worth: disable one at a time, hold every other hyperparameter fixed, and the change in validation score is attributable to that mechanism alone.
Results are reported over five seeds; the sharpest degradation appears when the multi-head term is removed, consistent with prior work…
Saytomo
✕
A “remove-and-measure” experiment that quantifies what each component is worth:
- Disable one mechanism at a time, hold everything else fixed
- Compare the validation score before vs. after
- The bigger the drop, the more that part mattered
A few things people say to it
Answer · explains it in a line.
Answer · the gist, no fluff.
Replace · tightens in place.
Replace · writes the response.
Replace · swaps the language.
Answer · no selection needed.
Think in your language. Send in theirs.
Set a target language once, then press the Translate key and speak naturally. It cleans up and translates in one step — and keeps names, brands and your dictionary terms in their proper form.
When clean isn't enough, polish it.
Polish is the one key that rewrites — turning a rambling thought into tidy prose. It keeps your meaning, claims and facts exactly, never balloons past ~1.5×, and follows your saved style notes. Code-switching stays intact here too.
It learns the words
only you say.
Teammates' names, product names, acronyms, jargon. Saytomo biases the recognizer toward your terms and snaps phonetic near-misses to the right spelling. Edit a dictated word and it offers to remember it.
Formal in Gmail, loose in WhatsApp.
Saytomo notices which app you're in and adjusts the contractions and punctuation to match — same words, same meaning, just dressed for the room. (You can override it globally or per app.)
I will not make the 3 PM sync. Let us reschedule for tomorrow.
formal · no contractionsI won't make the 3 PM sync. Let's reschedule for tomorrow.
neutral · contractions ok
won't make the 3pm sync — let's push to tomorrow
casual · loose punctuationSee your voice add up.
A quiet home screen tallies what you've dictated and the hours you've saved against a 45-WPM baseline. Every number is computed locally on this device — there's no server endpoint behind it.
Computed locally · per device · never sent to a server
Fluent in 40+ languages.
Pick your languages from the curated list, or drop in any BCP-47 code. Speak one, or switch mid-sentence — Saytomo follows along and keeps every word where it belongs.
A system-wide voice tool you can actually trust.
Saytomo needs powerful permissions, so we're explicit about what it does and doesn't do with them.
No keylogging
The only keystroke we watch for is your hotkey. We never capture or transmit what you type.
No passive reading
Your selected text is read only at the instant you press Ask, used for that one request, and never stored. No screen contents, titles or URLs.
Audio isn't kept
Speech is processed in the moment and discarded. Transcripts are never written to disk on our servers.
No training on your words
Your data is never used to train models. Every company that touches data is listed on a public, auditable subprocessor page.
Start free. Upgrade when it clicks.
Start with 7 days of Pro, free — then dictate for free forever. Pro unlocks unlimited words all day, your dictionary synced across every device, and new features first. Your 7 days start the first time you dictate — not the day you sign up.
The whole app — every key, free.
You'll start with 7 days of Pro, free.
Billed monthly or annually · cancel anytime
Questions?
The honest answers.
Stop typing.
Start talking.
Free to download, faithful by default, and quietly counting the hours it gives you back.
macOS 13+ · Apple Silicon & Intel · v0.0.8