Saidby writes down your Google Meet calls as they happen, with each person's name, and records a separate track for every speaker. Speech recognition runs on your computer, so the audio and the text are never uploaded.
Free for sessions up to 30 minutes and the first 30 minutes of each file. No account. Needs a computer with WebGPU — most computers from the last few years.
Saidby hears each person in a Meet call separately and labels every line with the name Meet shows.
Lines appear in the side panel a second or two after each phrase, while the call goes on.
Echo removal drops other people's words that your microphone picked up from the speakers.
Your transcripts are kept with their recordings. Reopen, save or delete any of them from the side panel.
The speech model runs on your computer's graphics chip (WebGPU). It's a one-time 1.26 GB download; after that, nothing about your calls is sent anywhere.
Transcripts and recordings are saved in Chrome on this computer, not in someone's cloud.
Copy the transcript or save it as .txt or Markdown.
Audio gives you one .m4a file with an "Everyone" track plus a track for each speaker. All packs the text, the recording and a separate file per speaker into one .zip.
A lecture, a webinar, a video or a call on another site: open it, click the Saidby icon and start. Saidby writes down what plays in that tab.
In this mode it records only the tab, not your microphone, and can't tell people apart, so lines are not named.
An interview, a podcast, a voice memo: press Transcribe file in the side panel, or drop the file on it. It's transcribed on your computer by the same speech model, with a progress bar and the time left.
Once the file is copied in, you can close the panel and the work goes on. The transcript joins your history, with times from the start of the file. A file is transcribed as one voice, without names.
Speech recognition happens on your computer. Audio and transcripts are never sent to any server — ours included.
Before each call you confirm that you've told everyone you're transcribing. While it works, a "Transcribing on this computer" label stays on the Meet page.
No sign-up, no analytics, no ads.
Prices in USD. Taxes may apply depending on your country. Payments are handled by Paddle, our reseller. 30-day money-back guarantee — see the refund policy.
For another tab: open the page that plays sound, click the Saidby icon and start the same way. For a recording on your computer: press Transcribe file in the side panel, or drop the file on the panel.
Chrome with WebGPU, which most computers from the last few years have, and about 1.3 GB of free disk space for the speech model. If WebGPU isn't available, the side panel tells you before you start; check that hardware acceleration is on in Chrome's settings.
25: Bulgarian, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hungarian, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Spanish, Swedish and Ukrainian. The language is recognized automatically.
Yes. Many countries require everyone's consent to record or transcribe a call. Saidby asks you to confirm you've told everyone before it starts, but it doesn't notify the other people for you.
Only on your computer, in Chrome's storage for Saidby: the transcripts in the extension's storage, the recordings in its private file storage, the speech model in the browser's private storage. While a file is transcribed, a temporary copy of it is kept in that private file storage and deleted when it's done. Delete a transcript and its audio from the side panel; uninstalling Saidby deletes everything.
Google Meet, with names and your microphone on its own track. Any other tab with sound — a video, a webinar, a call on another site — without names and without your microphone.
Most audio and video files that Chrome can decode: recordings from a phone or a recorder, podcasts, saved meeting recordings, videos (Saidby uses their sound). If a file's audio can't be decoded, the panel says so; converting it to MP3 or M4A usually helps. One file at a time, transcribed as one voice without speaker names.
It depends on your graphics chip: a two-hour recording takes a few minutes with a fast graphics card, and about 15 minutes on a laptop's built-in graphics. The panel shows the time left, and you can close it while the file is transcribed.
A live session stops, and the transcript and the recording up to that point are saved; start again to continue. For a file, Free transcribes its first 30 minutes. Pro has no limit for either.
Click the Saidby icon to open the side panel, click the gear (Settings), paste your key under License and press Activate. Lost your key? Get it again on the purchase page.
Cancel the monthly plan any time from the link in your receipt email; Pro stays on until the end of the paid month. Any purchase can be refunded within 30 days — see the refund policy.
The people, the meeting, the lecture and the files in the screenshots are fictional. Speech recognition uses NVIDIA Parakeet (CC BY 4.0).
Saidby