100 interviews at 45 minutes each is 4,500 minutes — one $24 pack of minutes.

At 45 minutes each, 100 interviews is 4,500 minutes of audio. One $24 pack of 5,000 minutes covers all of it, with 500 minutes spare. Nothing smaller reaches 4,500 — two $12 packs are 4,000 minutes, and $12 plus two $8 packs is also 4,000 — so $24 is the floor. Packs are one-time purchases with no subscription and credits never expire.

50 hours is 3,000 minutes. A $12 pack of 2,000 minutes plus an $8 pack of 1,000 is exactly 3,000 minutes for $20, and that is the cheapest route to it. Three $8 packs reach the same 3,000 minutes but cost $24, and a single $24 pack costs the same while giving you 5,000 minutes you may not need.

Same arithmetic, different totals. Forty 45-minute interviews is 1,800 minutes, covered by one $12 pack of 2,000 for $12 — two $8 packs reach 2,000 as well but cost $16. Four hundred interviews is 18,000 minutes: three $24 packs plus a $12 and an $8 is exactly 18,000 minutes, $92. Work in minutes rather than files and the number falls out.

They lower what you actually spend, not the pack maths. Your first transcript is free at any length and your second is $0.99 once per email address, so packs only need to cover the remaining 98 interviews — 4,410 minutes, which is still the one $24 pack because nothing smaller reaches it. Treat the free one as a real test on your own audio.

Connect Google Drive and import a whole folder in one go, or take an API key and submit programmatically from wherever the recordings already live. Everything lands in the same searchable Library either way. There is no CSV importer and no watch folder, so for a corpus this size it is the Drive folder or a short script against the API.

No. The transcript comes back as continuous text with timestamps, and it does not separate or label the people in the recording. If your analysis depends on attributing every line to the interviewer or the participant, you will be marking that up yourself after export.

Processing usually runs at about two minutes of wall clock per hour of audio, so a 45-minute interview is typically done in under two minutes and a three-hour session in well under ten. That is a measured average rather than a guarantee, and the Library shows each file's progress as it goes.

Each interview gets a summary, the takeaways, key quotes with the timestamp each one sits at, chapters marking where the conversation turns, and a chat over the transcript you can ask questions of. This is what makes 100 interviews searchable as a set rather than a folder of documents.

DOCX if you want to read and annotate the interview as a document, TXT for plain text, and JSON if you are feeding a script, a search index or an analysis tool that accepts structured input. All five formats — TXT, SRT, VTT, DOCX, JSON — are available on every transcript.

One file up to 5 GB and up to 12 hours; anything longer is split before processing without you doing anything, and language is detected automatically across 99 languages. Transcripts are never used to train AI. Anything your institution or client requires beyond that — ethics approval, data-handling rules, where recordings may be stored — is a question about your study, so check it against your own policy.