Audio Recorder is a fully client-side application built with Next.js that explores in-browser audio recording, transcription, and summarization. It uses the Web Audio API to capture voice input and stores all recordings locally using IndexedDB for offline persistence.
Once a recording is complete, the app generates a transcript and then a summary using the @huggingface/transformers package, specifically leveraging the pipeline("summarization") function running entirely in the browser via WebAssembly.
⚠️ Note: The transcription and summarization functionalities are experimental and unstable. Processing is fully local and may be slow or fail, depending on the device’s resources.
🔗 Links
🧱 Tech Stack
- Framework: Next.js
- Audio: Web Audio API
- Storage: IndexedDB (local database)
- ML Inference:
@huggingface/transformersrunning in-browser (WASM)
🚀 Features
- Record audio directly in the browser using Web Audio API
- Store recordings locally in IndexedDB
- Generate transcripts from audio (experimental)
- Summarize transcripts using Hugging Face Transformers (unstable, no API calls)
- Fully private: no server, no network requests, all data stays on-device