The meeting that stays on your phone

Why we chose the phone, not the cloud

The therapist's question revealed something we'd suspected but hadn't articulated clearly enough: people need to trust where their words live. Not eventually. Not after reading terms and conditions. From the moment they hit record.

When we started building Scribr, we faced a choice. We could architect it like most note apps do: record on the phone, send audio to a server somewhere, transcribe in the cloud, sync everything back. It's the standard playbook. It's also the playbook that makes people nervous when they're discussing their salary negotiation, their health, or their client's sensitive information.

So we didn't do that. Not for the free tier, anyway. Instead, we built the transcription engine directly into Scribr using Whisper on iOS and Apple Speech on the same platform. Your audio file never leaves your device. It's processed locally. The transcription stays on your phone unless you explicitly choose to sync it or share it. This isn't a feature we market loudly. It's how the app works by default.

What on-device transcription actually means

I should be clear about what this does and doesn't do, because there's often confusion here.

On-device transcription means your audio file is converted to text using a model that runs directly on your iPhone. Whisper is Meta's open-source speech recognition system. We integrated it into Scribr so that when you record a meeting, a voice note, or a phone call through the app, the conversion from sound to words happens silently in the background. No upload. No server request. No timestamp logged anywhere except on your device.

What it doesn't mean: it doesn't mean Scribr is a dumb recorder. You still get full biometric lock protection (Face ID or Touch ID) across the app. You still have a Quick Record widget on your home screen. You can still start recording by just saying "Hey Siri, record a note with Scribr." The experience is smooth. It's just that the intelligence part stays put.

The trade-off is real, though. On-device transcription is slower than cloud transcription, and it works best with clear audio. If you're in a noisy cafe or you're transcribing a 90-minute workshop recording, you might hit limits. That's why Pro users can upgrade to cloud transcription via Deepgram for longer or messier audio. But for everyday use, on your phone, on-device just works.

The week we realised privacy needed to be boring

Three days after we shipped Scribr, a user asked us in a support email: "If I don't pay for Pro, does my transcription get sent to any servers at all?" We answered honestly: no, it doesn't. Her reply was just two words: "Good. Keep going."

That's when it clicked. Privacy wasn't going to be a selling point for us. It was going to be the baseline. The thing that happened without us needing to shout about it.

We've built the app so that the free tier is fully private by design. You record. Your phone transcribes. Your notes live in your Notes app or in Scribr's local storage. If you want to use AI features (summaries, action item extraction, meeting intelligence), you'll need to upgrade to Pro, and at that point, you're choosing to send that data to our cloud service. But until then, it's yours alone.

The consequence of this choice is that we have no telemetry about what you're recording. We don't know how many calls you had last week or which topics you discuss most. We don't train models on your audio. We can't even see it. Some people might say this is a competitive disadvantage. I'd argue it's the only way to build a meeting tool that people actually want to use for the conversations that matter.

How the pieces fit together

Let me walk through what actually happens when you hit record in Scribr.

You open the app. You tap the record button or use the Quick Record widget from your home screen. Your iPhone starts capturing audio. In the background, Whisper (or Apple Speech) is working. By the time you've finished your meeting, the transcript is usually ready, sitting on your phone, protected by your biometric lock.

If you want to share that note or sync it across your devices, you can. But you're choosing when and where that happens. You're not locked into any particular workflow. Some users just keep their notes local and export them as PDFs. Some upgrade to Pro and use Scribr Cloud so they can access notes across devices. Some use the Contact Intelligence feature (available on Team tier) to see if notes from previous calls with the same person exist.

The on-device transcription is the foundation. Everything else is optional.

What this means for the work you actually do

Most people don't think about transcription technology when they're in a meeting. They think about whether they're remembering the key points. They think about whether they need to follow up. They think about whether the person across from them heard what they said.

A meeting app should disappear. You should be able to record without worrying about where your words are going. You should be able to use the transcription without negotiating terms and conditions.

That's what on-device transcription gives you. It's not a headline. It's not a competitive advantage we're going to shout from our website. It's just how Scribr works if you don't want to pay for anything. You get a tool that captures your voice, transcribes it, and keeps it safe. No negotiation. No mystery.

The therapist who emailed us in November? She's still using Scribr. She never upgraded. She didn't need to. For her work, the on-device transcription and a biometric lock were exactly enough.