The reason we chose on-device voice capture for Ideas!

The moment we knew voice had to be private

When I was building the first version of Ideas!, I kept a voice memo app open alongside my code editor. Every half-formed thought went into it. Podcast outline. A joke for next week's newsletter. A client pitch structure. Then I'd spend 20 minutes transcribing manually or paying for a third-party service to do it.

The friction was real, but the privacy concern was larger. I didn't want those half-thoughts, the vulnerable ones, the ones I'd delete before sending to anyone, sitting on someone else's servers. And more practically, I didn't want to pay per transcription on top of the app cost.

So we made a decision early: Ideas! would use on-device speech recognition. Everything stays on your phone. No upload. No account linking to third-party services. No per-transcription fees.

What on-device SFSpeechRecognizer actually does

SFSpeechRecognizer is Apple's built-in framework for speech recognition. It runs locally on your device, which means when you hit record in Ideas!, your voice never leaves your phone. The transcription happens right there, and that's the end of it.

In practical terms: you tap the microphone icon, speak your idea, and within a few seconds you see the text. No waiting for a server. No network requests. If you're on an aeroplane or in a place with no signal, it still works. The transcription quality is solid for natural speech, which is what an idea capture moment needs to be. You're not dictating formal documents; you're catching thoughts before they evaporate.

This also means we're not holding the keys to your ideas. Apple's framework handles it; your device handles it. We never see it, store it, or log it. That trade-off - a slightly narrower feature set in exchange for genuine privacy - felt like the right one for the creators we were building for.

Why speed matters more than perfection

Here's what we observed during early testing: people use voice capture when they're in motion. Driving. Walking between meetings. Thinking in the shower. The worst thing we could do is make them wait for transcription to finish, or worse, require them to stop what they're doing to upload and process somewhere else.

On-device transcription means Ideas! is ready the moment you finish speaking. The idea is captured and searchable within seconds. You can add a category, tag it, move on. No second step. No friction.

It's not perfect transcription. Sometimes a word lands wrong, and you need to edit. But it's fast enough to keep you in the creative moment, and editable enough that you can clean it up in a few taps. That's the actual job: capture the kernel of the idea before it's gone.

The categories and tags layer on top

Voice capture is just the beginning. Once an idea is transcribed, you can assign it to a category (podcast episodes, content series, writing projects, whatever you define) and tag it (urgent, collaboration needed, research required). The search is built right in, so you can find that half-remembered thought you captured three months ago without scrolling through hundreds of entries.

For creators working across multiple projects or formats, this becomes the connective tissue. Your podcast outline lives alongside your newsletter ideas and your long-form article seeds, all in one place instead of scattered across voice memos, notes apps, and email drafts to yourself.

What happens when you want to see your ideas everywhere

On-device transcription is private, but sometimes you need your ideas accessible from your laptop or tablet. That's where the Creator tier comes in. iCloud sync means your ideas move across your devices without uploading to a third-party service. Apple handles the sync; you still own the privacy.

If you're part of a content team or running a coaching programme, the community voting board lets you share specific ideas with collaborators without making every single capture public. You control what leaves your personal vault.