Hold to talk, release to paste
No timer deciding you have finished. Pause for as long as you need to find the word - the recording ends when you let go of the key, and the cleaned text lands in whatever app has focus.
AI dictation for Mac
Speak It Once, Send It As Written
Hold a key, talk however you actually talk, and clean text appears in the app you are already in - fillers gone, punctuation right, and your names spelled correctly. Free, with no word limit.
Free · No word limit · English · macOS 14.2 or later · Apple silicon
Dictation
Held, spoken, pasted
What you said
“um so I think we should uh push the Kubernetes migration to next sprint because you know Priya is out and… and the staging environment isn't ready”
What was pasted
I think we should push the Kubernetes migration to next sprint, because Priya is out and the staging environment isn't ready.
“Kubernetes” and “Priya” came from your dictionary
What makes it usable
Dictation fails on the details.
Almost every dictation tool can turn speech into words. What decides whether you keep using one is whether it cuts you off mid-thought, whether the output needs editing, and whether it can spell your colleague's name.
Text you can send, not a transcript
Filler words removed, punctuation and capitalisation fixed, obvious grammar slips corrected - while keeping your wording and meaning. The difference between what you said and what you meant to write.
A dictionary that learns
Correct a misheard name once and Crodo offers to remember it. Learned terms are sent to the speech engine as recognition hints, so they come out right at the source rather than being patched afterwards.
Works the same everywhere
Mail, Slack, Chrome, Notion, your terminal, a form in a browser - the finished text is pasted into whatever has focus, so behaviour never changes app to app.
The custom dictionary
It learns the words you actually use.
Every job has vocabulary a general speech model has never heard. Product names, colleagues, clients, acronyms. This is the feature that decides whether dictation survives contact with real work.
01
It notices your correction
Fix a mangled name right after pasting and Crodo offers to learn it - one click, no settings screen to find.
02
Fixed at the source
Learned words are passed to the speech engine as key terms before transcription, so the model is listening for them. That is meaningfully better than find-and-replace on the output, which cannot tell a name from a similar-sounding word.
03
It builds up quietly
Colleagues, clients, product names, acronyms, medical or legal terms - they accumulate as you work until dictation simply knows your vocabulary.
How it works
Four steps, one of them yours.
01
Hold your dictation hotkey and speak - a sentence, a paragraph, or a rough thought with pauses in it.
02
Release the key. The audio goes to the speech engine along with your dictionary terms as recognition hints.
03
The transcript is cleaned up: fillers removed, punctuation and capitalisation fixed, grammar tidied, your wording kept.
04
The finished text is pasted into the app you were already in - or copied, if pasting is not the right move.
Where it fits
Anywhere you would have typed.
Email and replies
Speak the response, get a sendable paragraph. The bulk of dictation time for most people, and where cleanup earns its keep.
Docs and long-form drafts
Get the rough thinking down while it is fresh. Speaking is roughly three times faster than typing, and editing a draft beats staring at a blank page.
Tickets and bug reports
Describe what broke out loud instead of typing it twice - once badly, once properly.
Chat and Slack
Long replies without the typing, in an app where dictation usually behaves worst.
Notes while you think
Capture a thought at the speed you had it, before it turns into a half-remembered bullet point.
Anywhere your hands are busy
Holding a coffee, holding a baby, standing at a whiteboard, or just tired of typing.
Before you download
Two limits worth knowing now.
English only
Transcription is fixed to English today. If you dictate in another language regularly, Wispr Flow and Superwhisper both advertise 100+ languages and are the better choice - we would rather say that here than have you find out after installing.
Cloud, not on-device
Audio is sent to a cloud speech engine, so dictation needs an internet connection and your voice leaves your Mac. For work where that is not acceptable - legal, medical, locked-down networks - use a local tool like Superwhisper, VoiceInk, or MacWhisper instead.
Everything else on this page - the cleanup, the dictionary, no word caps, pasting into any app - costs nothing and has no tier to unlock.
Questions
The things people ask first.
Is Crodo's dictation free?
Yes. Free to use with no weekly or monthly word cap and no paid tier for any of the features described here. Most Mac dictation apps either cap the free tier - Wispr Flow's is roughly 2,000 words a week - or charge a subscription.
What languages does it support?
English only, today. Transcription is fixed to English, so if you dictate in another language regularly this is not the right tool - Wispr Flow and Superwhisper both advertise 100+ languages and are the honest recommendation there.
Does it work offline?
No. Audio is sent to a cloud speech engine, so dictation needs an internet connection. If on-device processing is a requirement, Superwhisper, VoiceInk, and MacWhisper all run locally, and macOS has built-in dictation that works offline too.
Does it stop if I pause to think?
No. Crodo is hold-to-talk: the recording runs until you release the key, however long you pause. This is the main reason people move away from Apple's built-in dictation, which ends the session after roughly 30 seconds of silence with no way to disable it.
Can I teach it names and jargon?
Yes. Add terms in the Dictionary tab, or just correct a misheard word after pasting and Crodo offers to learn it. Terms are sent to the speech engine as recognition hints before transcription, so they are heard correctly rather than corrected afterwards.
Which apps does it work in?
All of them. The finished text is pasted into whichever app has focus, so Mail, Slack, Chrome, Notion, a terminal, and browser forms all behave identically. There is a clipboard fallback for the occasional app where pasting is not appropriate.
Can I choose which microphone it uses?
Yes. Pick your input device in Settings - useful when a headset, webcam, and external mic are all connected. Crodo filters the list down to real microphones rather than showing every virtual and aggregate device macOS knows about.
How is this different from Apple's built-in dictation?
Three things: it does not cut off when you pause, it cleans up filler words and grammar rather than transcribing literally, and it can learn your vocabulary - Apple's has no custom dictionary at all. Both are free, so it is purely a capability question. The trade is that Apple's runs on-device and Crodo's does not.