RaDUO (Radio Dataset Utility and Organizer) is a tool for building multilingual speech-to-text datasets from live radio broadcasts.
What it does
- 📻 Captures live radio stations across Asia, Australia, Europe & North America
- ✂️ Creates random-length speech segments, typically 20–30 seconds
- 🤖 Supports multiple STT providers, including OpenAI-compatible APIs, Qwen, ElevenLabs, Deepgram, OpenRouter & more
- 📝 Stores transcripts, language, station, regional & timing metadata
- 🔄 Supports retry, cleanup, review & export of failed/incomplete data
- 📊 Exports CSV, JSONL, Excel & Hugging Face
audiofolder metadataGUI
🖥️ Desktop GUI for managing the complete dataset workflow, including station discovery, radio sources, ASR models, fallback chains, API keys, transcript editing, statistics, retries and activity logs.
🎙️ Output: 16 kHz Mono WAV + transcripts + metadata
Built for ASR research, evaluation and model-training preparation.
🔗 Project: https://www.dkly.net/projects/9