What is the Learning System (coming soon)?

The Learning System will let VoicePad **learn from your corrections** and improve accuracy over time — all locally on your device. **In development.** Not yet active in v2.1.0. **How it will work:** After each transcription completes, a 👍 👎 feedback bar will appear below the transcript. | Button | Meaning | What happens | |--------|---------|--------------| | 👍 **Good** | "This is correct" | Confirms transcription — VoicePad reinforces the pattern | | 👎 **Fix** | "There's a mistake" | Opens editing mode — correct the text, VoicePad learns the correction | **Every 10th thumbs-up shows a special 🐘 elephant message** as a small reward for engagement. Yes, really. **What VoicePad will learn:** - Brand names (iPhone, WhatsApp, Spotify, PostgreSQL) - Technical terms (Kubernetes, GraphQL, PyInstaller) - Your name and contacts' names - Industry-specific jargon - Foreign words you use often in your language **How learning happens (technical):** Your 👎 corrections feed into the Custom Dictionary at `AppData\Roaming\VoicePad\custom_dictionary.json`. Over time, VoicePad builds a personal dictionary of your specific vocabulary and the Whisper mistakes you consistently see. **Privacy:** 🔒 **All learning happens locally on your device.** Your corrections are stored in your local custom dictionary. **Nothing is uploaded anywhere.** Not aggregated, not shared, not used to train models. Your vocabulary is yours. **When will it ship?** It will ship as an **automatic update** — no reinstall needed. **Founding members get it first.** Track status on the public roadmap. **Why this matters:** Generic Whisper is trained on millions of generic voices. It doesn't know YOU specifically. The Learning System personalizes Whisper to your voice, your vocabulary, your terminology — without sending a single byte of training data to anyone else's servers. It's federated learning the way it should be: local, voluntary, transparent, and entirely under your control.