Most of them do not. The majority of AI notetakers are built around text. They record the meeting in order to transcribe it, then keep the transcript, the summary and the action items. The meeting itself, especially the picture, is often not kept at all, or is kept only on a higher paid tier with limits on length.
If your calls involve anything on screen, a demo, a design, a spreadsheet, a proposal, this is the difference between having a record and having a description of one.
Video is expensive. A one-hour recording is hundreds of megabytes. Multiply that by every meeting for every customer and storage becomes a serious line item, while a transcript of the same meeting is a few kilobytes of text.
So the economics push the whole category towards text. Record it, transcribe it, summarise it, throw the recording away or charge extra for keeping it.
That is fine when the meeting was a conversation. It falls apart when the meeting was a demonstration.
Think about what a transcript of a screen share actually says. "So if we look at this column here, you can see the problem." The words are captured perfectly. The information is gone.
The same applies to design reviews, walkthroughs, onboarding calls, training sessions, anything where a person pointed at something. The transcript records that pointing happened.
There is a second thing transcripts lose, which is tone. A summary will tell you the client agreed to the timeline. The recording tells you how they said it.
As of July 2026, roughly: Jotted keeps it on every plan with no length limit. Bluedot keeps it on its video tier, with the cheaper tier capped. tl;dv keeps it. Fathom can capture video, but doing so means its bot joins the call. Otter, Fireflies, Granola and Notta are focused on transcripts or notes rather than keeping the recording.
This changes often, so check each tool's own page. The question to ask is specific: does the plan I would be on keep the video, and is there a cap on how long a single recording can be?
A lot of people end up running two tools. A notetaker for the transcript and summary, and a screen recorder like Loom for the picture. It works, and it is a nuisance: two things to remember to start, two places the meeting lives, and the job of matching them up afterwards.
That is the exact problem Jotted was built to remove. One recording, botless, with the screen and both audio tracks, then the transcript, the summary and the action items from the same file.
A Jotted recording works out around 180 MB per hour for the file that gets kept and backed up. The raw working files are larger while it processes, and are moved to your Recycle Bin afterwards so you get the space back. There is no cap on recording length on any plan, including the free one.
Otter is built around transcription rather than keeping the meeting recording. If seeing what was on screen matters to you, that is the usual reason people look for an alternative.
Fathom can capture video, but its bot-free modes are transcript only. To capture video the Fathom Notetaker joins the call as a participant.
Jotted keeps it on every plan with no length cap. Bluedot keeps it on its video tier. tl;dv keeps it but uses a bot. Check each tool's current plans, since this changes often.
Because a transcript of a screen share captures the words and loses the thing being shown. On demos, design reviews and walkthroughs, the screen was the meeting.
Roughly 180 MB per hour for a Jotted recording. There is no limit on how long you can record, on any plan.
Record free forever, no card and no bot. Join the waitlist and you'll be first in.
Free forever plan · Unlimited recordings · No bot, ever