Why we built auto-transcription into Ideas!
Last autumn, a podcast producer named Sarah sent us a message that stuck with me. She'd been using Ideas! for three weeks and said: 'I get my best ideas while walking. I used to fumble with my phone, start a voice memo, then forget to transcribe it for days.' She asked, simply, 'Can you just write it down for me?' That question shaped how we built the core of Ideas!.
The note-taking graveyard nobody talks about
Most creators keep voice memos scattered everywhere. Phone voice app, a notes app, WhatsApp messages to themselves, maybe even an email draft. Each one a small friction point. You capture the thought in a moment of inspiration, but then you have to manually transcribe it later, which means the idea loses momentum. By the time you've typed it out, you've already moved on to the next thing.
We realised early that Ideas! wasn't going to be a general note-taking app. There are enough of those. What creators actually needed was one place to capture an idea fast, in any format, and have it ready to develop later. Voice was the obvious path. But voice without transcription is still just a recording you have to manage.
Staying private, staying fast
We had a choice early on. Send voice to a server somewhere, transcribe it in the cloud, get back results in seconds. Reliable, standard, but it meant sending your voice data away every time you capture an idea. For creators, that felt wrong. Your ideas are your work. They're often half-formed, vulnerable, sometimes confidential.
We chose to use on-device transcription instead. iOS has SFSpeechRecognizer built in, and it's genuinely good. It runs locally on your phone. Your voice never leaves your device. It's free. And it's fast enough for the job. The trade-off is you need an internet connection to download the language model the first time, but after that, transcription happens in seconds without any server involved. It felt like the right choice for a creator tool.
The moment we knew it was working
During our first week of wider testing, we watched a YouTube creator use Ideas! on a walk. She spoke an idea into her phone. Tapped transcribe. Within three seconds, her entire thought appeared as text, ready to edit or tag. She didn't have to remember to come back to it. She didn't have to sit down at a desk. The idea was captured, transcribed, and already in her Ideas! library, searchable and tagged, before she'd walked another fifty metres.
That's when we understood what we'd built. It wasn't a transcription feature. It was a friction eliminator. For people who think on the move, who create while they're living, auto-transcription meant ideas could finally stay with them without the manual work that usually kills momentum.
Why this matters more than it sounds
A lot of creators told us they use voice memos because typing is slow. Fair enough. But the real issue isn't capturing the voice. It's what happens next. Voice without text means you have to listen to it again later to remember what you meant. Text means you can scan it, search for it, tag it, share it with others, build on it. Transcription is what turns a voice memo into an actual idea you can work with.
We've had pastors planning sermons capture a thought while driving home from work. YouTubers capture a video concept while doing dishes. Coaches capture a lesson idea between sessions. In every case, transcription meant they could carry that thought forward without dropping it. The idea stays alive.
What we learned building this
We learned that creators don't want perfect transcription. They want fast, usable transcription. If it gets 90 percent right, they'll fix the 10 percent in seconds. They learned to speak clearly into their phone. Some told us it actually made them sharper about what they were trying to say. You can't mumble into a transcriber the way you might mumble into a note-taking app.
We also learned that categories and tags matter as much as the transcription itself. An idea captured and transcribed but lost in a messy list is almost as bad as no capture at all. That's why we built search and tagging alongside transcription. Ideas! isn't just about getting your thoughts out of your head. It's about getting them into a system where you'll actually find them again and use them.
Auto-transcription works best when it's paired with a real workflow, not just a bucket for voice recordings. What would change about how you capture ideas if you knew you'd never lose them, and could find them again in seconds?