For interviews you can’t afford to leak, the best transcription apps run the speech model on your own device, so the tape never reaches a vendor’s server. On a laptop, noScribe (free, built for interviews, with speaker labels) and MacWhisper (Mac, one-time Pro upgrade) are the strongest choices. On a phone, Private Transcribe (iPhone and Android) and Whisper Notes (Apple only) transcribe in the field with no upload. Cloud services such as Otter and Rev are faster to share and label speakers well, but they hold your audio.
Why it matters where interview audio is processed #
A cloud transcription service is a third party with a copy of your recording. That copy can be exposed in a breach. It can be kept longer than you’d like, or requested through legal process from the vendor rather than from you. Human transcription services go further: a person you’ve never met listens to the tape.
For a press conference that doesn’t matter. For a source who spoke on condition of anonymity, it might. Local transcription keeps the audio on hardware you control, which leaves you with one thing to protect instead of two.
The best private transcription apps compared #
| App | Platforms | Runs on your device | Speaker labels | Price |
|---|---|---|---|---|
| noScribe | Windows, Mac, Linux | Yes | Yes | Free, open source |
| MacWhisper | Mac | Yes | Pro only | Free; Pro is a one-time purchase |
| Private Transcribe | iPhone, iPad, Android | Yes | No | Free with ads; Pro $7.99 one-time |
| Whisper Notes | iPhone 12+, iPad, Apple silicon Mac | Yes | Yes | $7.99 one-time on iOS |
| Buzz | Windows, Mac, Linux | Yes | Yes | Free, open source |
| Otter | Web, iOS, Android | No, cloud | Yes | Free 300 min/month; Pro $16.99/month |
| Rev | Web, iOS, Android | No, cloud and human | Yes | Human transcription $1.99/min |
Prices were checked on each developer’s site in September 2026. Check again before you buy.
noScribe: best free tool for serious interviews #
noScribe is built for exactly this job. Its developer describes it as a tool for “high quality transcripts of interviews for qualitative social research or journalistic use.” It runs entirely on your computer and uses Whisper for the words and pyannote to tell speakers apart. It supports around 60 languages and includes an editor for checking the transcript against the audio.
The trade-off is speed. High-accuracy transcription on a laptop without a strong GPU is slow, so queue up long interviews and leave them running. It’s free under the GPL, and the project page has downloads for Windows, Mac and Linux. Our Linux interview guide walks through it step by step.
MacWhisper: the easiest option on a Mac #
MacWhisper is the most polished way to run Whisper on a Mac. You drag in a file and edit the transcript in the app. The free version covers basic transcription. Pro, listed at €64 one-time on its site, adds automatic speaker recognition, batch transcription of whole folders, meeting recording from Zoom and Teams, and exports to Word, PDF and Markdown. It only runs on Macs.
Private Transcribe: transcribe on your phone, in the field #
A lot of reporting happens away from a laptop. Private Transcribe runs Whisper on iPhone and Android phones, so you can record an interview, stop, and have text before you’ve left the building. It works in airplane mode, has no account, and no transcription server exists to receive your audio.
What makes it useful for interviews:
- Model choice. Professional (539 MB) is meant for interviews, and Best (574 MB, Whisper large-v3-turbo) is the most accurate tier for recent phones.
- Custom vocabulary so a source’s name or a technical term comes out spelled correctly.
- Timestamps for every line, so you can find a quote and play it back.
- Imports from a separate recorder: m4a, mp3, wav, aac, mp4 and mov files up to 90 minutes.
- Translation into English with Pro, done on the phone with a multilingual model from Quick to Professional. The Best tier’s large-v3-turbo model wasn’t trained for translation, so switch down for this. Treat it as a working aid and have a fluent speaker check anything you intend to quote.
Its main limit for interviews is that it doesn’t label speakers. In a two-person interview that’s a quick fix while editing. For a panel, a desktop tool with speaker identification will save time. Pro is a one-time $7.99 purchase that removes ads and adds .txt and .srt export.
Whisper Notes: Apple alternative with speaker labels #
Whisper Notes runs on iPhone 12 and later, iPad and Apple silicon Macs, and it transcribes offline. It labels speakers with timestamps for each turn and accepts files shared straight from Voice Memos. The iOS app is $7.99 one-time according to its site. The Mac download is sold separately. It isn’t available on Android or Windows.
Buzz: free, simple desktop transcription #
Buzz is a free, open-source Whisper app for Windows, Mac and Linux. It exports TXT, SRT and VTT and can identify speakers. It’s less interview-focused than noScribe but quicker to get going. Current Mac builds need Apple silicon.
When a cloud service still makes sense #
Cloud tools earn their place for material that isn’t sensitive. Otter gives free users 300 minutes a month, with a 30-minute limit per conversation. Pro costs $16.99 a month, or $8.33 a month billed yearly, per Otter’s pricing. Rev’s human transcription costs $1.99 a minute and promises delivery within 12 hours, per Rev’s pricing page. Both are good for press conferences, public meetings and on-the-record events where speed and easy sharing matter more than who else has the file.
Check the transcript before you quote from it #
Every speech model makes mistakes, and Whisper-based ones can occasionally produce words nobody said, especially over long silences or heavy noise. Never quote from a machine transcript without listening to that passage. Timestamps make this quick: find the line, play it, confirm it. Pay particular attention to names, numbers and negatives. A dropped “not” changes the story.
How to get more accurate interview transcripts #
- Put the recorder close to the person talking. A phone between two people on a table works. A phone in your pocket doesn’t.
- Record a backup on a second device for anything you can’t redo.
- Ask sources to spell names on tape. It fixes the transcript and your notes.
- Pin the language if the interview is in one language, and use a multilingual model if you and the source switch languages.
- Choose a bigger model for accents, noise or phone calls.
More on this in private transcription accuracy vs online services.
Frequently asked questions #
Are local transcription apps as accurate as Otter or Rev? #
For clear, one-on-one interviews, local Whisper models come close to cloud AI services. Human transcription from Rev is still more reliable on messy audio, and cloud tools are usually better at separating many speakers. Local tools win on who holds the recording and on cost over time.
Which private transcription apps label speakers offline? #
noScribe, MacWhisper Pro, Whisper Notes, Buzz and Vibe can all identify speakers without uploading audio. Private Transcribe and most phone keyboards can’t, so you add names as you edit.
Can I transcribe interviews in other languages without the cloud? #
Yes. Whisper’s multilingual models cover 99+ languages, and noScribe supports around 60. Accuracy is best for widely spoken languages. Some apps, including Private Transcribe Pro, can also translate speech into English on the device.
Is it safe to keep interview recordings on my phone? #
It’s as safe as the phone. Use a strong passcode, keep the OS updated, and check whether your phone’s cloud backup includes app data. Delete recordings once they’re transcribed and checked, unless your newsroom’s policy says to keep them.