Speech to text software, sorted by what leaves your computer

Speech to text software turns spoken words into written ones, and it is used for three jobs: typing by voice into whatever you are writing, turning a recording into text, and keeping a written record of a conversation. Tools also differ in where the speech is turned into text: on a server, which means the audio has to be sent there, or on the computer you speak into. Dalm is a Mac app that does all three jobs. The speech is recognised on your Mac and no audio is sent anywhere; the text is synced to our servers automatically. There is a Free plan.

Typing, a file, a conversation: the three jobs

Typing by voice. You speak instead of typing, and the words land where the cursor is. In Dalm you hold the right ⌘ key, speak and let go; the text is typed into whatever app you were in. The key can be changed. How dictation works, key by key, and voice to text, from setup to the text in the field.

A recording. You already have the audio — a voice memo, a call someone recorded, a video — and you want the words. Dalm takes any audio or video file macOS can read, up to four hours long, and it works on the Free plan. Turning a recording into text, step by step, three ways to get a recording transcribed, and transcription tools compared by name.

A conversation. People talk, and you want a record of who said what. Dalm records a call from your Mac — your microphone and the sound of the call — and nothing joins the meeting; in a room it takes the microphone. Ways to record a meeting, compared, and the kinds of AI note taker. When the job is choosing software to transcribe meetings, the whole path from the call to the transcript is on one page.

What leaves your computer: the audio, the text, or neither

Whatever the job, a speech to text tool is built one of three ways, and the difference is what leaves the computer you speak into.

	Recognised on a server	Recognised on the computer, text kept there	Recognised on the computer, text synced
Where the audio goes	To the server that turns it into text	Nowhere: it is not sent	Nowhere: it is not sent
What reaches a server	The audio, and the text made from it	No server is involved	The text
Your text on a second computer	Where the service stores it	Only if you move it yourself	Through your account

None of the three is the “right” one: the first needs no model on your machine, the second keeps everything in one place, the third lets the text follow you. What matters is knowing which one you are using before you dictate a contract or record a client. Dalm is the third: the speech is recognised on your Mac, no audio is sent anywhere, and the text is synced to our servers.

What Dalm sends in each of the three

• Typing by voice — the speech is recognised on your Mac and the text is typed into your app without going through the clipboard. The dictation is also saved in your list and synced to our servers, where no one but you can see it, even in a team.
• A recording — the file is read from your disk and not copied. It becomes a dictation or a meeting, and from there it is text like any other.
• A conversation — the voices are told apart on your Mac, and the transcript is synced to our servers. When a summary is written, a language model gets the text; the audio is not part of what it gets.

In all three jobs no audio is sent anywhere, and what goes to our servers is text.

Everything you say lands in one archive you can search: meetings, phone calls, conversations in the room, and what you dictate to a model.

What keeps working with no network, and what does not, is on a page of its own; which step Dalm does where is laid out step by step.

Speech to text accuracy: languages, names and your own terms

You do not tell Dalm which language you speak: “You never pick a language — it is detected for you.” What you choose is the coverage — “European languages”, which covers “English, Russian, German, French, Spanish and twenty more”, or “More languages”: “Slower, but over a hundred languages.” More on choosing the coverage for the languages you speak.

Names and in-house terms are the words you will want to check. Any dictation or transcript line can be edited. The dictionary holds a term and the ways it tends to be misheard; correct a word in a transcript and the app offers to add it, and a word can go in before you ever say it. The same dictionary works for dictation and for meetings.

Free, Pro or Team: starting with speech to text

All three jobs come with the Free plan, and Free is a plan, not a trial: there is no end date. Dalm needs a Mac with Apple Silicon · macOS 15.0 or later and an account, and while the app is not public, the way in is the waitlist. What the paid plans add is on the pricing page. For a team, the dictionary, templates and tags are shared, and dictations stay visible only to the person who made them — speech to text for a team.