Acousmos
AI dictation for Mac with a permanent voice archive
Today I launched Acousmos on Product Hunt: https://www.producthunt.com/posts/acousmos
It's AI dictation for the Mac with one opinionated bet: your voice is an asset, not disposable input. Every dictation — original audio plus text — lands in a local, searchable archive.
Honest context:
Solo and bootstrapped, launching with essentially no audience
Three months of dogfooding: 11,669 dictations, 188 WPM, ~162 hours saved vs typing
Priced above most rivals ($12.99–17.99/mo) because the archive and managed cloud are the product, not an upsell
What I want to learn today: whether "keep every word" resonates beyond me, and what would make a voice second brain useful to you. AMA about the build, the pricing call, or launching from zero.
I've been dictating everything on my Mac for three months — 11,669 dictations, 495k words, 188 WPM average against my ~40 WPM typing. The app's own stats page says 162 hours saved.
But the number that actually changed my mind wasn't speed. It was search: every dictation (audio + text) lands in a local archive, and I keep finding myself pulling up things I said weeks ago — the plot of a feature I'd talked through weeks earlier, the exact wording I gave a customer. Every other dictation tool throws that away.
So that's the bet Acousmos makes: dictation where your voice becomes an asset, not disposable input. Mac only, solo, bootstrapped, $0 MRR, launching properly soon.
Curious about one thing from folks here: when you dictate, do you ever wish you could find what you said again — or is that just me?
-
1On your retrieval question: what I'd want isn't better search, it's time and context as the index. Text search only helps if I roughly remember the words I used, and the dictations I most want back are the ones where I was thinking out loud and the wording was bad. The query that would earn my trust is "what did I say while I had that file open on Thursday afternoon" — app/window plus timestamp first, full text as the fallback. The other thing worth deciding early is what you do with the half-sentences. A lot of spoken thinking is restarts and "no wait, actually" — a cleaned-up transcript reads better but drops the exact moment the idea changed direction, which is usually why I'm digging it up. Are you keeping the raw pass alongside the tidied one?
-
1
What stood out to me is that your insight isn't about dictation—it's about memory.
Most dictation apps optimize for getting words onto the screen. You're optimizing for making spoken thinking retrievable weeks later. That's a fundamentally different job, and I think it's where the real long-term value is.
-
1
"Memory" is honestly a sharper word than any I've used for it — thank you for that.
You're right that it's a different job. The design consequence I keep coming back to: memory only works if capture costs nothing. Nobody maintains a separate "record my thoughts" habit for long. So the archive isn't a feature you use — it's a byproduct of typing by voice. You dictate because it's faster, and the memory builds itself on the side.
The retrieval moment that sold me: I half-remembered a phrase from a feature plan I'd talked through three weeks earlier — found it in three seconds, and replayed my own voice working out the idea.
Curious what retrieval would need to look like for you to trust it with your thinking — full-text search is table stakes, but I wonder what else.
-
1
Interesting question.
Your example made me think about one implication of building a product around retrieval that I don't think I can explain properly in a thread without oversimplifying it.
I think it's much easier to discuss against a real example than in a few public comments.
If you're interested, what's the best email to reach you on?
-
-
Comment