Who spoke
A local speaker-detection model splits the recording into turns and labels them Speaker 1, Speaker 2 and so on, so a conversation reads like a script instead of one block of text. Two small model files, about 34 MB, download once.
New in CaringDictate
Your meetings, transcribed on your own computer
CaringDictate now notices when a call starts in any program, records it, and turns it into a readable transcript with speaker labels and notes. Every one of those steps happens on your PC. Nothing is uploaded, there is no per-minute charge, and it is included in the same $49 one-time price.
- Works with Zoom, Teams, WhatsApp, WeChat, your browser — any app
- Speaker labels, talk time and action items, all offline
- Off until you switch it on, with an indicator while it runs
- Where it runs
- Entirely on your Windows PC. No account, no server, no upload.
- Which apps
- Any program that makes a call. There is no approved-app list.
- What you get
- Audio, a transcript, speaker labels, talk time and notes.
- What it costs
- Included. $49 once for the whole app.
How it works
Four things happen, and you only start the first one
- 1
It notices the call
Once you switch it on, CaringDictate watches which programs are using your microphone and your speakers at the same time — the pattern a two-way call makes. There is no list of approved apps: Zoom, Teams, WhatsApp, WeChat, Discord, Slack or a call in your browser all look the same to it.
- 2
It records both sides
Your microphone goes on the left channel and the other side’s audio on the right, in one stereo file kept on your computer. The file is written as the call happens, so a crash or a flat battery leaves a playable recording rather than nothing.
- 3
It transcribes afterwards
When the call ends, the recording joins a queue and is transcribed in the background by the same speech model you dictate with — on your graphics card, when your chosen model and machine can use one. It waits while you are dictating, so it never fights you for the machine.
- 4
You read, search and export
Every transcript is searchable by any word inside it, including Chinese. Save one as a Word document, PDF, plain text, Markdown, or as .srt / .vtt subtitles.
Who spoke, and what it came to
A transcript that reads like a conversation
A wall of text is hard to use. With speaker detection turned on, each turn is attributed, the transcript breaks into speaker blocks, and the recording gets a short breakdown of who talked how much.
Turn the notes switch on as well and a small language model on your own disk writes a summary and the action items underneath — and renames the recording from “Microsoft Teams” to something you will recognise a month later.
Both are optional, both are off until you switch them on, and both run on this computer like everything else in the app.
What is in it
Everything the Meetings page does
Who talked how much
Each recording shows a share bar per speaker with their total talking time. Useful for interviews, and quietly revealing in a meeting.
Meeting notes
Turn on the notes switch and a small AI model on your own disk writes a summary and a list of action items into each recording after it is transcribed. It also suggests a title, so the row stops being called “Microsoft Teams”.
Your own vocabulary
The personal dictionary you built for dictation — names, product names, medical or legal terms — is applied to meeting transcripts too.
Files you already have
Drag in a WAV, MP3 or M4A you recorded somewhere else and get the same transcript, the same speaker labels and the same export choices. Up to five hours of audio.
Your files, your folder
Choose where recordings live. Keep them forever or have older ones cleared after 30, 90 or 365 days, or above a disk limit you set. Pinned recordings are never removed.
You are in charge of it
It records because you switched it on
Recording is off when you install CaringDictate. Before anything is captured you read a short note and accept it: recording rules differ by country and state, some ask that everyone on the call agrees, and that part is yours to handle.
After that it is straightforward. A small indicator sits on your screen while a recording runs, so you can always tell. Whether you mention it to the other people, and how, is your call — the app does not speak for you.
None of this is legal advice. If you are not sure about a particular call, check the rules where you and the other people are.
Being straight with you
What it does not do
Worth reading before you decide, because some of this is what people expect from a meeting tool and will not find here.
It is not a meeting bot.
Nothing joins your call as a participant, and no one else sees that CaringDictate is there. It records what your own computer can already hear, the way a recorder on the desk would.
It will not tell the other people for you.
How and when you mention a recording is your call, not the app's. On your own screen, a small indicator sits there while one is running, so you can always tell it is on.
Speaker labels are Speaker 1 and Speaker 2, not names.
The app works out how many different voices there are and which turn belongs to which. It does not know who anybody is, and it does not build a voice profile that follows a person between recordings.
Long recordings have a ceiling.
A recording is transcribed up to five hours. Speaker detection needs the whole recording in memory, so it works up to 90 minutes — past that you still get the full transcript, just without the labels.
Notes are a summary, not minutes.
The notes model is small enough to run on an ordinary computer, which is the point. Read what it wrote before you send it to anyone.
If a hosted meeting assistant that joins the call, keeps a shared team workspace and produces live captions is what you need, that is a different category of product and there are good ones — start with whichever your team already uses. CaringDictate is the option for people who want the recording and the transcript to stay on one computer.
Before you start
What you need
- Windows 10 or 11, 64-bit. The same machine you already dictate on.
- A speech model. The one CaringDictate already downloaded for dictation is the one it uses for meetings.
- About 34 MB more for speaker labels — two small files, fetched once in the background.
- A one-time download for notes, sized to your computer, only if you turn the notes switch on.
- Disk space for the recordings themselves. Audio is compressed, and you can cap how much is kept.
- No internet during a meeting. Connection is needed to fetch the models once, and nothing after that.
Questions
Recording and transcribing meetings — the honest answers
Does CaringDictate record meetings without sending anything to the cloud?
Yes. The recording is written to your own hard drive, transcribed by a speech model already on your computer, and the speaker labels and the notes are produced by models on the same machine. No audio, transcript or note is uploaded anywhere — there is no server involved in any of it. You can pull the network cable out and it all still works.
Which meeting apps does it work with?
Any of them. CaringDictate does not keep a list of approved apps — it watches for the audio pattern a two-way call makes, so Zoom, Microsoft Teams, WhatsApp, WeChat, Discord, Slack, Skype, a call in Chrome or Edge, and anything else all work the same way. You can also press Record now and capture anything at all, including a conversation in the room.
Is call recording switched on by default?
No. It is off when you install the app, and nothing is recorded until you have read and accepted a short one-time note about being responsible for consent where you are. Switching it back off stops any automatic recording straight away.
Am I allowed to record a call?
That depends on where you and the other people are. Some countries and US states ask that everyone on the call agrees; others only need one person to. We cannot know your situation and this is not legal advice — if you are not sure, ask, or check the rules where you are before you record.
Will the other people know I am recording?
Not unless you tell them — that is your call to make, in your own words, and the app does not speak for you. On your own screen, a small indicator sits there while a recording runs, so you can always tell it is on. You can turn that indicator off if you would rather not see it.
Does it label who said what?
Yes, once the speaker models are installed — two files, about 34 MB, fetched once in the background. Transcripts then come out as Speaker 1, Speaker 2 and so on, with a breakdown of how long each person talked. The labels are positional, not identities: the app has no idea who anyone is and keeps no voice profile between recordings.
Can it write meeting notes?
Yes, if you turn the notes switch on. A small language model downloads once and then, after each recording is transcribed, writes a short summary and a list of action items into that recording and suggests a title for it. It runs on your processor like everything else here, so it takes a few minutes on a slower machine. For recordings you already have, there is a Create notes button on the recording itself.
Can I transcribe an audio file I already recorded?
Yes. Drag a WAV, MP3 or M4A file onto the Meetings page, or choose it with the button. You get the same transcript, the same optional speaker labels, and the same export choices — Word, PDF, text, Markdown, or subtitles. Files up to five hours are supported.
What can I do with a transcript once it is made?
Read it in the app, copy it, search every transcript you have by any word inside it, or save it as a Word document (.docx), a PDF, plain text, Markdown, or as .srt / .vtt subtitles. The audio stays alongside it in its own folder, which you can open at any time.
Where are the recordings kept, and how much disk do they use?
Each recording gets its own folder — in your CaringDictate data folder by default, or anywhere you choose. Inside it are the compressed audio file, the transcript, and the notes if you enabled them. You can keep everything forever, or have recordings older than 30, 90 or 365 days removed automatically, or set a disk budget of 5 to 50 GB and let the oldest go. Anything you pin is never deleted.
Does recording and transcribing slow my computer down?
Recording is light — it is writing audio to disk. Transcription is the heavy part, and it happens after the call, one recording at a time, in the background. Where your chosen model and machine allow it, the work runs on your graphics card, which is several times faster than the processor, and the app now says on its Voice models page which one is doing the listening. It pauses while you are dictating so it never competes with you.
Does this cost extra?
No. Meeting recording, transcription, speaker labels and notes are part of CaringDictate at the same $49 one-time price, and they arrived in a free update for people who had already bought it. There is no per-minute charge and no monthly plan, because there is no server for anyone to pay for.
Related reading
- How CaringDictate works — the three-minute walkthrough of dictation itself.
- Privacy — exactly what is stored on your computer and what leaves it.
- Offline dictation, explained properly — why on-device stopped being the slower option.
- Protecting your privacy when you dictate.
- Pricing — $49 once, what is included, and the refund.
Try it on your next call
Record a meeting, read it back, and see where it went.
The free 7-day trial is the whole app — dictation, meetings, transcripts and all. No sign-up, no card.
$49 one-time after the trial · 30-day money-back guarantee