Show HN: I've built a prototype for a physical e-ink todo list with voice input

A prototype of an e-ink todo device with voice input powered by local AI transcription. This web interface simulates the physical device experience with button interactions and monochrome display aesthetics.

• Single click - Toggle item completion
• Double click - Delete item
• Long press - Record voice to add/edit item
• Arrow buttons - Switch between lists

• AI Transcription: Transformers.js with Whisper Tiny
• Audio Processing: Web Audio API

1. Model Preloading - Whisper AI model loads automatically on page load with progress indicator
2. Microphone Permission - Browser requests mic access upfront for seamless recording
3. Voice Recording - Long-press triggers MediaRecorder API to capture audio
4. Audio Processing - AudioContext converts audio to proper format (16kHz Float32Array)
5. Web Worker Transcription - Offloads AI inference to background thread
6. Instant Feedback - Transcribed text appears immediately with visual confirmation

Open http://localhost:3000 with your browser.

Note: On first load, the Whisper model (~40MB) will download. Subsequent loads are instant thanks to browser caching.

1. Grant Permissions - Allow microphone access when prompted
2. Wait for Model - Yellow indicator shows model loading progress
3. Start Recording - When green "Ready to record!" appears, long-press any button
4. Speak - Say your todo item while holding the button
5. Release - Let go to transcribe and add the item
6. Manage Tasks - Click to complete, double-click to delete

E-ink todo list projects:

• blog.praccu.com
• E-ink Todo Board
• Invisible Calendar
• Building an E-ink Planner

MIT