Viktor.comAn AI coworker that actually does the work
Promoted
EVA
Your voice is enough to control your computer
3 followers
Your voice is enough to control your computer
3 followers
Eva sees your screen, hears your voice, and operates your computer - so people who have difficulty typing and clicking can still do everything.
Hey everyone! I'm one of the two people behind Eva. Really excited (and a little nervous) to share this with you.
Quick backstory: My co-dev temporarily lost the ability to type. Twice. First time was a ganglion cyst surgery — no keyboard for a month. Second time, a labrum tear that put his arm in a sling. Both times the frustration was identical: tasks that take 5 seconds — send an email, fill out a form, google something — suddenly took forever or required asking someone for help.
And here's the thing — he recovered both times. But 17 million people with cerebral palsy and 14.5 million with spinal cord injuries live with this every single day. Their brain works, their voice works, but every computer interface on earth assumes you can type and click.
So we built Eva.
What she actually does:
- You say "message my mom on iMessage that I need a ride at 2pm" and she just... does it. Opens iMessage, finds mom, types it, sends it.
- You can interrupt her mid-sentence and she stops *instantly* (sub-800ms). Try doing that with Siri. 😅
- She sees your screen — not through accessibility APIs or DOM parsing, but literally looks at it with AI vision and understands what's there. This means she works on literally any app, even legacy enterprise software with zero accessibility labels.
- She learns. Do a task 3 times and she saves it as a skill. Next time it's faster.
- She has a personality. Patient, warm, never condescending. Checks in during silence. Celebrates when things work. Like a supportive older sister who happens to be really good with computers.
The nerdy bit:
- Built on Gemini Live API for real-time voice (no STT→LLM→TTS pipeline — one model does it all), Gemini Flash for screen understanding, Google ADK for agent orchestration. Desktop app is Tauri 2 (Rust + React). We even found and patched a bug in Google's ADK (#4460) that was breaking voice interruption for everyone — not just us.
- The whole thing costs under $36/year (estimate) with a Gemini API key. Light users stay in the free tier entirely. A human assistant doing the same things? min. $3,600/year.
- Eva is fully open source (AGPLv3). You can grab it, run it locally, tear it apart, improve it.
- We built this in about a week and honestly we're a little amazed it works as well as it does. Would genuinely love your feedback — especially if you work in accessibility, know someone with a motor impairment, or just want to yell at your computer and have it actually listen.
What would you want Eva to do first?