Handy STT
Free open source speech-to-text app for desktop runs fully offline, keeping transcription and audio data private. Extensible and customizable, supports accessibility, cross-platform use, and community-driven modifications. No subscriptions or internet requirement.
Handy STT News & Activities
Recent News
Recent activities
What is Handy STT?
Handy STT is a no-cost, open-source speech-to-text application that operates entirely offline. As a cross-platform desktop application, Handy STT is constructed with Tauri (Rust + React/TypeScript), offering a streamlined, privacy-centric speech transcription solution. By pressing a shortcut and speaking, users can see their spoken words appear in any text field, without the need to send their voice data to the cloud.
The creation of Handy STT was driven by the necessity for a truly open-source, expandable speech-to-text tool. As a free application, Handy STT believes in making accessibility tools available to all, not just those who can afford to pay. Its open-source nature encourages collective development; users can personalize Handy STT to their liking and contribute to a larger cause. The application respects user privacy, keeping voice data on the user's computer and providing transcriptions without the need to send audio data to the cloud. Handy STT's simplicity lies in its single-purpose design: to transcribe spoken words and input them into a text box.
While Handy STT does not aspire to be the top speech-to-text application, it strives to be the most adaptable one, encouraging users to customize and expand its functionalities according to their needs.







Comments and Reviews
Done. I've finished looking for an STT app, or trying to vibe-code my own. This is it.
Same
Works as intended, written with handy :).
You'd be hard pressed to find a better FREE speech to text that is as lightweight, straightforward, and accurate as Handy is.
Handy STT is a simple, lightweight, no frills text to speech app that handles day to day use well. To use it you do have to download at least one LLM. The models are presented to you to choose from so you don't have to go searching for them and each has a simple bar graph showing you (assuming you're an English speaker) how accurate and responsive each model is. Set up is as easy a picking a model, letting it download, changing settings like if you want the text copied to your clipboard or not then you are ready to go. You can easily be up and going in under 10 minutes. A few small things hold this software back from being a perfect 10 in my book ( The primarily issue being you're unable to train the model on how to spell words. As well as not being able to instruct the program to write numbers or special characters such as exclamation marks or question marks.) Finally while not an issue for me the current fastest and most accurate models used ( Parakeet V2 and V3) are known to work much better with English but they are working hard to add wider language support.
Pros:
##Cons:
Although this comment seems to be at least AI-reshaped, it reflects perfectly my experience, specially this part: "not being able to instruct the program to write numbers or special characters such as exclamation marks or question marks [or commas]" so, my favourite STT tool so far.
@mll It's not AI-reshaped I just tend to be talk a LOT and seem a bit robotic when on my ADHD medication. 😅 I'm trying to get better but as you can see, I have a long way to to improve.
It's a good app, but need some touch how on this working offline implementation a bit better. And yeah you need to download a model before it can work and it doesn't show how much space needed in hard drive for model to work.
The application may have been different when you made this comment but currently (Version 0.7.7) it does show how much space a model will use. The text is in the bottom right-hand corner of the model's tab. From a quick glance size ranges from the smallest being Moonshine V2 Tiny at 31 MB while the largest is Whisper Turbo at 1.6 GB. Once you download the model that size text does disappear.