dIKta.me
Stop Typing and Start Talking - Local, Private, Versatile
1 follower
Stop Typing and Start Talking - Local, Private, Versatile
1 follower
Open-source local AI voice dictation for Windows. Speech-to-text-to-speech with local LLMs to do some other cool tricks, no cloud fees, no monthly subscriptions, no compromise on privacy. Option to expand to Cloud and larger models using BYOK options.
Everyone's racing to make AI smarter. Nobody's fixing how you talk to it.
We've gone from GPT-3 to high reasoning models in 3 years. Context windows went from 2K to 2M tokens. Multimodal, real-time, on-device — the intelligence side of the equation is moving fast.
But here's what hasn't changed: you still type at it like it's a search engine from 2003.
You open a browser tab. You paste text. You upload a screenshot manually. You switch apps to copy context. You pay five different subscriptions for tools that should be one.
The bottleneck was never the AI. The bottleneck is the input layer. Human-AI Interaction (HAI) is the next frontier — and it's embarrassingly underbuilt.
The models are ready. The interface isn't.
dIKta.me is built to close that gap for knowledge workers:
🎙️ Voice → text, formatted, injected at your cursor. 3-4× faster than typing.
👁️ Vision → point at anything on your screen, ask a question, get an answer. No copy-paste.
✂️ Selection → highlight text, refine/translate/ask about it. One hotkey.
🏠 Local-first → your models, your hardware, your data. No cloud required.
One app. Every input method. Any model. No new subscriptions.
Better HAI starts here.
How was it born?
I was tired of paying 4-5 monthly subscription for basic productivity tools, I gathered 100 dollars to pay for Claude Code compute and built a windows native App that saves me, and hopefully you, some good cash. With Privacy, security and speed in mind, dIKta.me is model agnostic and ready to ride the AI Tsunami, if we get better model, faster, smaller, we are ready to make good use of them without much extra code needed.
A single payment for a tool that frees your voice and your Local inference, no Subscriptions, no limits, no compromises on your privacy. And the evolution that is planned is going to be crazy. We are going Modular, adding a Memory, a Connectors and a Conversational orchestrator model called "Chaviz" so you can talk to it and ask for things to be done in your computer.
I'm a marketing & business executive with 20+ years in IT consulting, digital media, and project management. Not a software engineer by training. dIKta.me is my first desktop application, built from scratch with C#, WinUI 3, and AI coding tools. The product decisions come from two decades of building businesses and shipping products — not a CS degree.