Subscribe
Sign in
Harker
Turn thoughts into text at the speed of speech
Start new thread
Manuele

Harker - Turn thoughts into text at the speed of speech

by
A minimal speech-to-text widget that works locally on your Mac. Activate with a shortcut, speak anywhere you'd normally type, and watch your words appear instantly. Perfect for AI chats, emails, and any text field.

Add a comment

Replies

Best
Manuele
Maker
📌
Hey Product Hunt community! 👋 I'm thrilled to share Harker with you today. As someone who spends way too much time typing, I got frustrated with how slow my fingers were compared to my thoughts. And with all the AI conversations we're having these days, we're typing more than ever! That's when I realized most speech-to-text tools send your data to the cloud, or interrupt your workflow. Harker is different. Your voice never leaves your Mac, it launches instantly with a keyboard shortcut, and works in any text field. The interface stays hidden until you need it, then appears with a clean, minimal UI. Whether you're writing emails, chatting with AI, or taking notes, just hit the shortcut, speak, and your words appear exactly where you need them. I initially developed Harker for myself, I've been using it for months now and honestly can't imagine going back to typing everything. Would love to hear what you think and answer any questions! What would you use voice-to-text for in your daily workflow? Thanks for checking it out!
Maxim Zubarev

The problem I notice with local AI / ML applications is they are processing pretty slow. Would be nice to see some benchmarks (not sure how to visualize that when benchmarking audio tho lol).

Nice launch tho, super nice product, was just recently looking for alternatives on subscription-based apps like Wispr Flow.

Manuele
Maker

@mxmzb Totally fair. Local AI can feel slow depending on the setup and model size. Harker gives you the option to download and choose the model that best fits your machine, so you can get the right tradeoff between speed and accuracy.

For English-only models, even the smaller ones are surprisingly precise and very fast on M1 or M2. With no loading screens or cloud roundtrips.

Running everything locally also means no server costs, which lets me offer Harker as a one-time purchase.
No subscriptions, the app is yours to use.

Benchmarking audio is tricky to show, but I’m thinking about adding a simple demo or side-by-side soon.
Thanks for checking it out, and yes, it’s built as a simpler, privacy focused, and no-subscription alternative to more complex tools like Wispr Flow

Jeremiah Patterson

@mxmzb  @mnlcpc you might want to take down your PH page with the launch discount code if you're not going to honor it.... PH50OFF is EXPIRED.

Manu Goel
Launching soon!

So, you are using local/offline SLM?

Manuele
Maker

@manu_goel2 yes, the speech to text model is embedded in the app.

Manu Goel
Launching soon!

@mnlcpc love it ... that's tech put to right use

Manuele
Maker

Thank you @manu_goel2 , appreciate it

Vladimir Vashnev

How do I test the quality/speed and supported languages of the app before buying?

Jason Dunn

Have you thought about benchmarking this for accuracy against the built-in macOS voice recognition? That’s what I use today, and it’s built in and free, so that’s your basic competition.

Anthony Sanna

Will it run on any Mac OS? I have an old 2008 Macbook running 10.4 and a newer iMac 2020 running Monterey 12.7.6.

Karl Clement

Testing out your app. When I open it with the shortcut keys, it shows up, I speak into the mic and all it always types is Thank you. I gave it permission with Privacy settings

Manuele
Maker

Hi @karlclement , thank you for trying out Harker.
That’s a common Whisper hallucination, caused by its training data and how it interprets silence, background noise, or clicks.
Can you confirm that your microphone is properly connected and working? Do you see the “Harker line” moving in sync with your voice while you're speaking?

Mitja Kukovec

I purchased Hacker and gave it a try. It works well for the English language, but I would say that multilingual support is lacking. Whenever I speak in Slovene (one of the Slavic languages), it always detects it as Russian. Thus, language detection is not functioning effectively. It would be better to have a configuration option that allows users to choose which languages are enabled.