By ShinobiTools Team · Last updated: July 2026
Record the meeting, upload it, and start from an accurate transcript instead of a blank page. Trim it into clean minutes with decisions, owners and action items. No account, no cap.
Transcribe audio now →You can't paste a recording into a document, but you can paste a transcript. ScribeGrab gives you the full text of the meeting so writing the minutes is editing, not typing from scratch: keep the decisions and action items, cut the small talk.
If your meeting was a video call, see transcribe a Zoom or Teams meeting for saving the recording. Everything you upload is deleted right after processing, so sensitive discussions stay private. See privacy.
Be clear about what the transcript is: a complete record of everything said, in the order it was said, which is almost the opposite of minutes. Minutes are short, reordered and selective. Nobody reads an hour of dialogue to find out what they agreed to. The transcript is the raw material; the twenty minutes you spend after it is where the minutes happen.
The structure that works is the same one in every organisation that takes meetings seriously, and it has four parts. Decisions — what was actually settled, phrased so it still makes sense in six months. Actions — each with a single named owner and a date, because an action with two owners has none. Open questions — what was raised and deliberately not resolved, which is the part almost everyone drops and the part that causes the same discussion to happen again next month. And context, one or two lines on why a decision went the way it did, so the next person doesn't reopen it.
Everything else can go. The scheduling talk, the tangents, the three attempts at explaining the same point: none of it belongs in minutes. A good rule is that an hour-long meeting produces about a page. If yours is producing four, you're still transcribing rather than minuting.
Two mechanical things make this much faster. First, work from the speaker labels: ScribeGrab marks the voices as Speaker 1, Speaker 2 and so on, up to eight, so a commitment stays attached to whoever made it. Find-and-replace the labels with real names once, and every action item you lift already has its owner. Second, keep the SRT file: each line carries a timestamp, so a disputed decision can be checked at source rather than argued from memory. The click-to-play transcript does the same in the browser — click a line, hear that moment — which is the fastest way to confirm the exact wording of something before you write it down as agreed.
Meeting audio is the messiest audio there is, and knowing where it breaks tells you which parts of your minutes need a second look rather than making you distrust the whole document.
Crosstalk is the main one. When two people speak at once, the transcript usually keeps the louder voice and loses the quieter, and the line gets attributed to whoever won. That matters because interruptions cluster around exactly the moments where a decision is being made or challenged. If a decision looks like it came out of nowhere, that's the place to re-listen. Similar voices on similar headsets can be merged into one label, most often two men or two women in the same room on the same microphone; the giveaway is a "speaker" who appears to disagree with themselves. And a meeting with more than eight distinct voices falls back to plain unlabelled text, so for an all-hands or a large workshop expect a clean transcript without a cast list.
Room audio beats everything else. One laptop microphone at the end of a table of eight is the single biggest cause of a bad transcript, and no amount of processing fixes it afterwards. If the meeting is hybrid, the remote platform's own recording is usually far better than a phone on the table, because each participant is captured at their own microphone. If everyone is in one room, put the recorder in the middle rather than in front of the chair. And prefer an audio-only download when the platform offers one: it uploads faster and transcribes identically.
One limit to plan around: files go through in one pass up to 90 minutes. A longer workshop needs splitting in any free audio editor, which is no real hardship since there's no daily cap — and splitting at the natural break gives you two shorter documents to work through instead of one that nobody opens.
It gives you an accurate full transcript to work from, so writing minutes becomes quick editing instead of typing from scratch.
Yes. ScribeGrab pulls the audio from a video file automatically.
Up to 90 minutes and 2 GB per file, no daily cap.
No. It's deleted right after processing; results wiped within 45 minutes.