I've been going down this rabbit hole too. I ended up building DictaFlow (https://dictaflow.vercel.app/) because I needed something that specifically works in VDI/Citrix environments where clipboard pasting is blocked (I work in finance).
It uses a 'character-typing' method instead of clipboard injection, so it's compatible with pretty much anything remote. Also kept it super lightweight (<50MB RAM) for Windows users who don't want to run a full local server stack.
Cool to see Handy using the newer models—local voice tech is finally getting good.
I built this because I was frustrated with existing AI dictation tools failing in my daily workflow. I work in a locked-down corporate environment involving a lot of Remote Desktop (RDP) and Citrix sessions.
Because those environments block clipboard sharing for security, most dictation apps that rely on "transcribe-then-paste" simply don't work.
How it works: Instead of pasting text, DictaFlow mimics human keyboard input (sending keystrokes).
1. Audio: Captures raw PCM audio (16kHz).
2. Processing: Runs a quantized Whisper model locally.
3. Output: Simulates keypress events with a tunable delay. This tricks the remote desktop into thinking a physical keyboard is typing, bypassing the clipboard restriction entirely.
Features:
- Context Aware: It handles code formatting surprisingly well (e.g., "function def" turns into python syntax).
- Privacy: It doesn't upload audio to the cloud for training.
- Local-First: Designed to be lightweight.
It's currently Windows-only. I'd love feedback on the latency or any edge cases you find in other VDI environments.
Hey HN! We're transforming how Canadians buy and sell homes. Here's how:
Deep Pricing Analysis: Our AI doesn't just price your home. It dives deep into renovation quality from photos, listing descriptions, and neighborhood analytics to give you the most accurate model price in Canada. Plus, a detailed comparables analysis to show you why.
AI Realtors: Chat with our AI chatbots for a jargon-free journey through real estate. Try our chatbot here:
https://chat.smartbids.ai/
Automated Showings: Want a tour? We'll arrange it with a real person.
AI-Enhanced Listings: Selling? Our AI crafts the listing, refines your photos, and pinpoints the optimal list price.
Tailored Recommendations: Buying? Share your budget and wishes. Our AI will spotlight homes tailored for you.
B2B Focus: Initially, we're reaching out to brokerages to integrate our solutions and redefine the real estate landscape.
It uses a 'character-typing' method instead of clipboard injection, so it's compatible with pretty much anything remote. Also kept it super lightweight (<50MB RAM) for Windows users who don't want to run a full local server stack.
Cool to see Handy using the newer models—local voice tech is finally getting good.