Launching today
Stet
Smart open-source dictation that sounds like you, not AI.
22 followers
Smart open-source dictation that sounds like you, not AI.
22 followers
A minimalist, open-source voice input app for macOS. It listens locally, uses AI to refine your words, and preserves your natural voice. You can bring your own API key and use it for free, or subscribe to our cloud service for $6.99/month.
Stet
👋 Hey Product Hunt,
I’m Daniel, the maker of Stet, exciting to my first launch !
Most voice input apps are trying to do too much. They want to translate for you, rewrite your tone in different apps, and behave like a full AI assistant.
That was not what I wanted.
I just wanted an affordable simple voice input tool for macOS that could clean up my words a little, stay out of the way, and still sound like me. So I built Stet.
“Stet” is an editing term that means “let it stand.” That is the whole idea behind the product: your words, just cleaner.
Stet is open source, minimalist, and built for people who want better dictation without all the extra AI clutter.
If you already use an AI provider like OpenAI or Groq, you can bring your own API key and use Stet for free. If you do not want to set that up, you can use the hosted version for $6.99/month.
If you use the hosted option, your audio goes through my Supabase backend. I plan to open-source that part too, and it is designed for zero data retention.
I’m an indie developer, so I cannot offer an unlimited free tier, but registered users do get a free trial with 10,000 words.
If this sounds useful, I’d really appreciate a star on GitHub. And if you want to contribute, a PR would be even better.
Would love your feedback.
@naicheng_deng Hadn't thought about dictation this way before. I use Hermes for a lot of my daily workflow and the memory always ends up sounding nothing like me, probably because I clean everything up before I type it in. Feels like the human-mode output here could actually help with that.
Hi Naicheng! Congrats on the launch. preserve-your-natural-voice is the whole pitch, and also the thing most ai-refine tools quietly get wrong. they launder writing into generic corporate-voice, and the rambly personal bits are exactly what gets flattened. what's the rule for what to clean up vs leave alone?
BYO api key + open source in a category racing toward subscription lock-in is a good stance too. Good luck!
Stet
@keith_hiyamojo Hi Keith, thanks for your comment. We actually have two modes in this app, and it can tell whether your audience is human or AI based on cursor's position.
One is for person-to-person communication. We want to make sure your personal style is kept, even if there are some flaws in the words, since we believe those flaws are part of your personality.
The other one is for AI. In that case, it refines your words a little more and tries to make them more precise when you are talking to an agent.
@naicheng_deng two-mode split with cursor position as the audience detector is clever. most dictation tools apply one refinement regardless of context. and "those flaws are part of your personality" is exactly the right framing for human-mode. the gap between how we actually sound and how AI thinks we should sound is where voice dies in most of these tools.
bookmarked, will check it out as soon as I get chance. take care, hope it goes well for you!
great work! is MacOS the only operating system or you looking to explore it to other OS too? Do you also have an idea how much it would cost if I use my own API key - I'm just thinking because currently I pay $15 for Wisprflow so it'd be a nobrainer to do this if it's cheaper
Stet
@withstephen Thanks Stephen, honestly, Wisprflow is awesome. Yes, this app is written by swift/swiftUI, so not windows for now or future.
If you use your API key, it depends on the model that you choose. For example, if you choose groq and you choose GPT-OSS 20B, it's probably gonna cost you about $1.