Short episodes on transcription, audio, and getting the spoken word onto the page – for people who record interviews, meetings, podcasts, and research. From the team at Pepys.
Reading the feed…
Transcription work is bursty, so a monthly subscription bills you through every quiet month. Pay once for the minutes you actually use, and stop renting a tool you open twice a year.
A qualitative PhD runs a couple of dozen interviews, and typing them up by hand can eat weeks. Here is the real math, why your ethics board changes the tool you can use, and why transcription is a methods decision, not clerical work.
You don't need a notetaker bot silently joining your meeting. Record it yourself, upload the file after, and turn it into clean minutes, with consent you can stand behind and audio that stays yours.
How to record and transcribe an oral history so names, places, and the way someone actually spoke survive as a document that can sit beside the audio in an archive.
A single episode becomes show notes, a blog post, pull-quotes, and clips once you have the transcript. Here's the repurposing chain, and why one weekly show is a whole year of raw material.
The number one transcription complaint is that long recordings die halfway through. The fix is split-and-stitch: chunk the audio, transcribe the pieces, and re-offset the timestamps so a two-hour file comes back whole.
There's no true average deposition length, only a federal seven-hour cap and about one transcript page per minute. Legal work leans verbatim and stays confidential, and only a certified reporter's transcript is the official record.
The end-to-end researcher workflow for turning interviews into coding-ready transcripts: capture clean, get an AI first pass, pick a verbatim style, anonymize, and export a file your analysis software actually opens.
Transcribing a full qualitative study by hand runs into the hundreds of person-hours, roughly three to five working weeks of typing before analysis even starts. Here's the math that pushes researchers to automate.
A qualitative study quietly makes far more audio than you plan for. Count your interviews to saturation, multiply by length, and the real transcription pile appears, usually nine to seventeen hours.
Before you record a call or meeting there's a consent question, and it varies by where you are. The one-party vs all-party basics, plus how to keep sensitive recordings actually confidential.
Transcription pricing is all over the map – per minute, monthly subscriptions, free tiers with catches. Here's how the models actually work, so you stop overpaying for how you really use it.
Most transcription accuracy is won or lost before you hit record. How to get a usable transcript out of noisy rooms, strong accents, and crosstalk, and how to salvage a recording you've already got.
Transcript, captions, subtitles, translation – people use these interchangeably, but they're four different things with different formats and jobs. Here's what each one actually is.
Verbatim transcription keeps every um and false start; clean verbatim strips them out. Choose wrong and your transcript is either unreadable or missing what you needed. Here's how to pick.
Speaker diarization is the feature that labels who's talking, turning a wall of text into a readable conversation. How it works, why it matters for interviews and meetings, and where it still struggles.
AI transcribes in minutes; a human transcriptionist takes hours but catches what AI misses on hard audio. Here's how they really compare on speed, accuracy, and cost, and when each is the right call.
Accuracy is the first thing everyone asks about. Here's how it's actually measured with word error rate, why your audio drives almost all of it, and the simple things that move the number most.
Transcription sounds simple – turn audio into text – but the word hides a few choices that matter. Speech to text versus transcription, the structure that makes a transcript usable, and verbatim versus clean.
We read about a hundred Reddit threads where people vented about transcription tools. The same four complaints came up again and again. Here's what they are, and what people actually want instead.
