Voice-to-text that respects EU privacy
Dictaro is a Windows-native dictation app backed by Whisper large-v3 running on a GPU server in Slovakia. We built it because the best voice-to-text tools were locked to macOS, and the Windows alternatives sent your audio to US clouds without a real GDPR story.
Why we built this
Speech recognition got genuinely good around 2024. Whisper large-v3 transcribes Slovak, Czech, Hungarian, Polish, and German at a level that finally beats typing for long-form work. But the tools shipping that quality were Mac-only, expensive, or required uploading every recording to a US data centre, a non-starter for anyone working with patient notes, legal drafts, internal meeting recordings, or anything else that should not leave the EU.
So we built the version we wanted: a Windows desktop app with a global hotkey, paste-where-the-cursor-is behaviour, no audio retention, and a server you can point on a map.
What makes Dictaro different
| Where audio goes | Slovakia. Streamed to a self-hosted GPU server, transcribed in memory, never written to disk. |
|---|---|
| Languages | English, Slovak, Czech, Polish, German, Hungarian, all first-class. Whisper large-v3 quality for every one. |
| Platform | Windows-native. PySide6 desktop app, global hotkey, system tray, auto-paste at cursor, works in VS Code, Word, browsers, terminals. |
| Hosting | EU-only. Slovak GPU for inference, German Supabase for auth, Cloudflare EU edge for the public site. No US sub-processors. |
| Retention | Audio: zero. Transcripts: not stored on our side. Usage counters only. |
Who is behind it
Dictaro is built by MEDIA.COM Slovakia s.r.o., an IT company based in Bratislava. We have been running infrastructure, networks, and custom software for clients across Slovakia for over a decade. Dictaro is our first widely available product, but the operational discipline behind it (monitoring, backups, on-call response) comes from years of running production systems that customers depend on every day.
Founder: Andrej Ardan, CEO of MEDIA.COM Slovakia.
The technology
- Speech model: Whisper large-v3 (OpenAI, MIT-licensed) running on faster-whisper, an optimised CTranslate2 build.
- GPU: NVIDIA RTX-class hardware in Slovakia, with redundant backup nodes for failover.
- Desktop client: Python + PySide6 (LGPL), packaged with PyInstaller into a single Windows installer.
- Auto-correction: LanguageTool for grammar in supported languages, Hunspell for diacritic restoration.
- Hosting: Cloudflare Tunnel to a Docker stack with nginx, an admin portal, a Next.js user portal, HAProxy load balancing across whisper nodes, and a Supabase project for authentication.
Roadmap
- Now (free launch): 100 transcriptions/month free, no credit card. Self-service signup, dashboard, API keys.
- Q2 2026: Paid plans go live - Individual (€9/mo), Business (€29/seat/mo), Enterprise on request.
- Post-launch: Custom vocabulary cloud sync (Business+), comparison content for SK/CZ/EU markets, more keyboard layouts and dictation commands.
Get in touch
- Product questions and support: support@dictaro.eu
- Privacy and GDPR requests: privacy@dictaro.eu
- Business and enterprise: sales@dictaro.eu
Curious whether Dictaro fits your team or your workflow? Send a short note describing what you would dictate and we will reply with an honest answer, including cases where it is not a fit yet.