A Private, One-Time-Cost Transcription Workflow for Journalists and Podcasters
The freelancer's transcription problem
The math on transcription as a freelance journalist or independent podcaster goes like this. You do a handful of interviews in a good month. Each interview generates between thirty minutes and two hours of audio. You need the transcripts for reporting, editing, show notes, fact-checking, and the occasional pull quote. A cloud transcription subscription costs between ten and thirty dollars a month. The service is built for teams, or for people who transcribe constantly. You are neither. You are paying for capacity you will not use, nine months out of twelve.
Then there is the other problem. A material fraction of the interviews you are doing are with sources who expect confidentiality, or about subjects where the audio contains things you would rather not be sitting on a third-party server indefinitely. When Otter had a security disclosure a couple of years ago about stored transcripts, the panic in the journalism community was instructive — not because anything catastrophic happened, but because everyone suddenly realized how much of their reporting was sitting on someone else's infrastructure, under someone else's terms of service, governed by someone else's privacy policy.
And on the other side of the workflow, you still have the pile. Interview recordings from last quarter you meant to transcribe. Episodes you meant to extract show notes from. Content you could repurpose into posts, newsletters, social clips — if only the transcripts existed.
This post is about a workflow that handles all of that without a subscription, without uploading anything, and without the structural weirdness of paying a SaaS company to hold your source audio for you.
Why cloud transcription is a weird fit for interview work
None of this is a case against the major cloud transcription services. For live meeting capture in a sales or customer success context — recording a demo call, getting notes to Salesforce, sharing a transcript with a teammate — those products are excellent and they earn their subscriptions.
But interview journalism and podcast production have a different shape, and the cloud SaaS model chafes against it in a few specific places.
The usage pattern does not match the billing model. Interview work is bursty. You do ten interviews in two weeks while reporting a piece, then nothing for a month. You pay for the month anyway.
Source confidentiality is not negotiable. A source who agreed to talk on background did not agree to have their voice sitting on the servers of a third-party vendor, subject to whatever that vendor's privacy policy says today and whatever it says after its next acquisition. The legally correct answer is that they still did not agree even if the vendor claims SOC 2 compliance. The practical answer is that most journalists just quietly hope nothing goes wrong.
You do not own your own archive. If you cancel your cloud transcription subscription, you lose access to the interface you have been using to search and navigate years of your own interviews. You can usually export the raw data, but the tooling goes with the subscription.
Latency and reliability matter when you are on deadline. Uploading four hours of conference audio over a hotel wifi connection at 11pm the night before a deadline is an experience most working journalists have had, and none of them enjoyed it.
Again, none of this is a scandal. These are simply friction points between a product built for one audience and a user from a different audience. The fix is not to complain about Otter. The fix is to use a different tool for the jobs that tool is not built for.
The two kinds of recordings in a production workflow
For journalists and podcasters, transcription needs split cleanly into two modes.
Field recording. A conversation in progress. An interview happening live — across a table, over Zoom with local-recording enabled, on a lav mic at a conference, on a phone call recorded to a hardware device. You want this transcribed eventually but accuracy and format matter more than speed.
File processing. Existing audio on your disk. Interviews you already did. Episode source audio you want show notes or a blog post from. Older recordings you are revisiting for a longer feature. You want to process these in bulk, on a laptop, without ceremony.
A transcription tool that handles one of these two modes well and the other badly — or not at all — is a tool that is going to slow you down. Most of the cloud services handle live capture well and file import acceptably. Most of the on-device apps have historically handled file import well and live recording acceptably.
MinuteONE is trying to handle both well. You can open the iPhone app and tap record at the start of an interview and watch the transcript appear in real time. You can also open the Mac app the following morning, drag in the exported Zoom file from yesterday's other interview, and have a full transcript plus a generated summary a few minutes later. The same app does both. Nothing uploads. No subscription. One-time five-dollar purchase.
A complete production workflow
Here is what this can look like end-to-end for a typical interview.
Before the interview. You do not need to do anything. Make sure MinuteONE is installed on whichever device you will use to record. If the interview is in person, that is probably your iPhone. If it is on Zoom or Google Meet, it might be your Mac, or it might be a hardware recorder whose output you will import later.
During the interview. If you are using MinuteONE for live capture, tap record. The transcript appears on screen as you go. This is useful for two things: you can glance at it to confirm capture is working, and you can mark the moments where something important was said by checking the timestamp. If you are recording on hardware and importing later, just make sure your primary recording method is working — MinuteONE is not your fallback, it is your transcription layer.
After the interview. On the Mac, import the file (or finish the live recording). MinuteONE produces the transcript, then generates a summary, a list of any action items mentioned, and any decisions that were made. For a journalism interview, the summary and the list of decisions are less useful than the raw transcript, but the summary is surprisingly good as a way to find the specific moment you were looking for in a long conversation.
Editing and export. The transcript can be exported as PDF or plain text. For fact-checking, PDF is nice because it preserves the formatting and you can annotate it. For pulling quotes into your draft or sending to an editor, plain text is what you want. Audio can be exported as M4A if you need to send a clip somewhere.
Archiving. MinuteONE syncs across your devices via iCloud if you want it to. You can tag and filter the meeting library, which becomes a searchable archive of everything you have transcribed. Because the app has no server, your archive is your archive — it lives on your devices and in your own iCloud, and it does not depend on an external company continuing to exist.
On confidentiality, explicitly
The reason some of you are reading this particular post is not the five dollars. It is the part where nothing leaves your device.
MinuteONE processes everything locally. The transcription runs on Apple's on-device speech recognition framework, using the Neural Engine in your iPhone or Mac. The summarization uses Apple Intelligence, also on-device. There is no server component to the app. If your device is offline, the app works exactly the same as if it is online.
What this means in practice, for interview work:
- A recording of a conversation with a confidential source never touches a third-party server.
- There is no company that could be subpoenaed for access to your interview audio, because there is no company holding it.
- There is no data breach scenario that exposes your sources, because there is no central database to breach.
- When you tell a source "this conversation stays with me," the technical architecture of your transcription pipeline does not contradict that statement.
This does not free you from any of your other journalistic obligations around source protection — your device itself still needs to be reasonably secure, your iCloud account still needs good credentials, you still need to be thoughtful about where and on what you work. It just removes the particular third-party-cloud-provider risk from the equation.
For podcasters specifically
Two use cases that deserve their own mention.
Show notes and episode descriptions. Feeding the finished episode audio into MinuteONE produces a summary that is genuinely useful as the starting point for show notes. You will want to edit it — summaries are mechanical and show notes have voice — but the skeleton is there. The extracted decisions and action items, in particular, can map well onto the "what did we actually say" kind of episode description that indexes well for search.
Back catalog repurposing. If you have two years of episodes that never got proper transcripts, MinuteONE plus an afternoon is a cheap way to fix that. Transcripts are good for SEO. Transcripts are good for accessibility. Transcripts are good for finding that one thing you said in episode 37 that a listener is now asking about. One-time five dollars is the right price for a back-catalog transcription project.
FAQ
Does MinuteONE work for interviews with multiple people?
Yes. Apple's on-device speech recognition handles multi-speaker audio reasonably well when voices are distinct and overlap is limited. It does not currently produce per-speaker labels the way some cloud services do. If you need speaker attribution in the transcript itself, you will want to add it in editing.
Can I record phone calls?
Phone call recording is subject to device, platform, and jurisdictional limitations that MinuteONE does not bypass. For Zoom, Google Meet, or similar, record the call using the platform's built-in local recording and import the file to MinuteONE. For in-person conversations, the iPhone mic works fine.
What accuracy can I expect?
For clean audio — lav mics, quiet environments, native English speakers — very high. For noisy environments, heavy accents, or poor mic positioning, substantially lower. This is true of all AI transcription, cloud or on-device. Audio hygiene at the recording stage is the single biggest input to transcript quality.
What about languages other than English?
Apple's on-device speech recognition supports a number of languages, and MinuteONE inherits that support. Specific language availability depends on your device and iOS/macOS version.
Can I export for use in other tools?
Yes — plain text, PDF, and the original audio as M4A. No lock-in, no account.
Is there a team plan?
No. MinuteONE is a single-user app at a single price. If you need shared workspaces, collaborative notes, and team integrations, a cloud SaaS is genuinely the better fit for that use case.
MinuteONE is on the App Store for iPhone, iPad, and Mac. For the broader case for on-device transcription, see the main post.