The Private Transcription Choice

What Privacy Meant to Us From the Start

When we started MRVL Technologies, we weren't interested in building another web-based note app. The market had plenty of those. What we saw instead was a gap: knowledge workers, especially people in sensitive fields like law, therapy, and consulting, were recording meetings on their phones and then uploading them to servers they didn't fully trust.

The therapist's question made something clear. Privacy wasn't a nice-to-have feature. For certain professions, it was foundational. You can't ask a client to consent to their therapy session being stored on someone else's infrastructure. You can't ask a lawyer to upload case notes to the cloud on day one.

So we made a decision early: the Free tier of Scribr would use on-device transcription. No cloud upload. No licensing agreements with third parties. Your audio stays on your phone, gets transcribed on your phone, and never leaves it. We built this using Whisper on iOS and Apple's native speech engine, depending on your device and iOS version. Both work entirely locally.

How On-Device Transcription Actually Works

This is where people often get confused. Whisper and Apple Speech are two different tools, and we support both. Here's why: Whisper is OpenAI's speech recognition model. On iOS, you can run it locally on your device, which means the processing happens right there, in your pocket. Apple Speech is Apple's native transcription engine, built into iOS. It's faster, lighter, and has been optimised for iPhone hardware for years.

When you hit the Quick Record widget or use the Siri shortcut to start recording, Scribr detects which engine is available on your phone and uses that one. Neither sends your audio anywhere. The transcription happens on device. You get your notes, searchable and stored locally in the app, with biometric lock protecting them by default.

On the Free tier, that's all you get. And for most people, that's enough. A sales rep recording a client call. A student capturing a lecture. A researcher noting down interview snippets. No subscription required. No data leaving the phone.

When On-Device Hits Its Limits

We're not pretending on-device transcription is magic. It has real constraints. Whisper works brilliantly for most accents and audio quality, but it's slower than cloud alternatives. Apple Speech is fast, but it's tightly integrated with iOS, so you're dependent on whatever Apple ships in each version.

There's also the matter of audio length. If you're recording a three-hour strategy session or processing dozens of calls a week, on-device transcription becomes a battery and storage question. Longer audio files take longer to process locally. And if you need AI summaries or action item extraction, that computation has to happen somewhere.

That's when we built the bridge: Pro and Team tiers unlock cloud transcription via Deepgram, plus AI summaries and action item extraction. But here's the key detail. You opt in explicitly. We don't flip a switch and suddenly your audio is in the cloud. You choose the tier that matches your workflow. Free users stay fully private. Pro users get speed and intelligence. Team users add compliance features like GDPR audit logs and encrypted Vault Mode.

A Real Decision from Week Three

During our first month in the app store, we got a support message from a legal professional. She'd used Scribr to record client conversations, and she asked whether she could upgrade to Pro without losing the on-device option. She didn't want all her calls in the cloud.

That's when we realised our architecture needed to be flexible. So we built it that way. You can be on Pro, have access to cloud transcription and AI summaries, but still choose to record certain sensitive calls on-device only. The app respects that choice. You own the decision about what leaves your phone and what stays.

It's a small detail, but it's the opposite of how a lot of modern software works. Most apps make the convenience choice for you. We give you the privacy choice and trust you to decide when convenience is worth it.

Why This Matters Beyond the Feature List

On-device transcription isn't particularly novel as a technology. Whisper has been public for over a year. Apple's speech engine is years old. What matters is the commitment it represents.

When we chose to make it the foundation of Scribr's Free tier, we were saying something: privacy is not a premium feature. It's the default. If you want to use this app to capture your work without worrying about data leaks, third-party servers, or subscription costs, you can. Full stop.

Everything else we've built since has been an extension of that principle. Biometric lock by default across all tiers. Vault Mode with AES-GCM encryption for Pro users handling sensitive material. GDPR compliance modes for Team accounts. We're not trying to create a walled garden. We're trying to build a tool that respects the nature of the work people do in meetings and phone calls.

The therapist who asked if it stays private? She's been using Scribr for eight months now. Still on the Free tier. She doesn't need summaries or cloud sync. She needs to know her client's words aren't being processed by anyone but her, on a device she owns, protected by her own biometric lock. That's the entire product to her. And we think that's enough.