By ShinobiTools Team · Last updated: July 2026
Turn an interview recording into an accurate, quotable transcript: for journalism, research, hiring or content. Upload the audio or video and get the text back in minutes. No account, no cap.
Transcribe audio now →Transcribing an interview by hand takes hours. ScribeGrab does it in minutes so you can pull exact quotes, code qualitative research, or review a candidate's answers as text. The transcript is yours to download as TXT.
Clear one-on-one interviews transcribe very accurately; talking over each other is harder, so check those parts. Interviews are sensitive, so your file is deleted right after processing and wiped within 45 minutes, nothing kept or shared. See privacy and accuracy.
The reason journalists and researchers transcribe interviews isn't convenience, it's defensibility. A quote you can point to in the recording is a quote you can publish. A quote you half-remember is a liability, and no amount of AI accuracy changes that: the transcript is a working document, not evidence.
So verify before you quote. The click-to-play transcript on the results page makes this cheap: click any line and it plays that exact moment of the audio. Do it for every sentence that will appear between quotation marks, and do it especially for numbers, dates, names and anything negated. "We did not approve the budget" and "we did approve the budget" differ by one short word, and one short word is the kind of thing speech recognition drops when someone speaks quickly. Keep the SRT file too, because it carries a timestamp on every line: a quote with a timestamp beside it in your notes is one you can re-check months later without listening to the whole recording again.
The speaker labels are what make an interview transcript usable at all. ScribeGrab separates the voices automatically and marks them Speaker 1, Speaker 2 and so on, so an interviewer's question and the answer to it don't dissolve into one paragraph. Replace the labels with real names once, at the top, and the whole document becomes quotable. Two cautions worth taking seriously: labels are estimated from voice characteristics rather than from any participant list, so speakers with similar voices can be merged, and where two people talk over each other the line usually goes to whoever was louder. Both failures cluster at handoffs, which is exactly where an attributed quote goes wrong, so check the label at the start and end of anything you plan to use.
Recording law varies more than people expect and it isn't a technical question, so treat this as the practical shape of it rather than legal advice. In much of Europe, including the Netherlands, a participant in a conversation may generally record it, but publishing or sharing that recording is a separate matter and personal data rules still apply. Several US states require the consent of everyone on the call, not just one party. Professional codes often go further than the law: journalism and academic ethics boards typically require explicit, recorded consent regardless of jurisdiction.
The workable habit is simple and costs ten seconds: ask on the recording. Start by stating that you're recording, why, and what you'll do with it, and let the answer be part of the file. It settles the question permanently and it's the first thing any editor or ethics reviewer will look for. If the interview is remote, say it before the substance begins rather than at the end.
On our side, the handling is deliberately boring. Your upload is deleted the moment processing finishes, and the transcript and subtitle files are wiped from the server within 45 minutes — long enough to download them, short enough that there's nothing sitting around afterwards. There's no account, so there's no profile to attach an interview to and nothing to delete later, and nothing you upload is used to train anything. Transcription runs on our own GPU hardware using a pre-trained open-source model, not through a third-party transcription API. For a sensitive interview that's the relevant detail, and the privacy policy spells out the retention in full.
Very accurate on clear speech, but always proofread a quote before you publish it, especially names and technical terms.
No. It produces one accurate transcript; add speaker labels yourself if you need them.
Up to 90 minutes and 2 GB per file, no daily cap.
No, it's not kept. The file is deleted right after processing; results wiped within 45 minutes.