# RTAI

> RadioTEDU's English technical showcase for live broadcast context and local, model-assisted audio production.

URL: https://radiotedu.com/rtai/
Publisher: RadioTEDU, the student-led radio and media laboratory at TED University in Ankara, Türkiye.
Language: English only.

## Two documented systems

1. RadioTEDU AI presents listener-facing broadcast context based on station status and current-track information.
2. RTAI Jingle is an open-source Windows production tool that turns a user-written script and voice direction into local QwenTTS narration, then uses FFmpeg for filtering, mixing and export.

## Why Local Models?

- Working scripts and voice direction stay on the production workstation during generation.
- After the model and dependencies are installed, voice generation can run offline.
- The generated narration and audio-processing stages remain inspectable before export.
- Local inference has real hardware, storage and setup requirements; the page states these constraints rather than hiding them.

## Verified RTAI Jingle details

- Model family: QwenTTS. The project README identifies Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign for the current release.
- Audio processing: FFmpeg filtering, compression, timing, gain, fades and music mixing.
- Export documented by the project: 48 kHz stereo MP3 at 320 kbps or 24-bit WAV.
- Source: https://github.com/radiotedu/rtai-jingle
- User guide: https://github.com/radiotedu/rtai-jingle/blob/main/docs/USER_GUIDE.md
- Methods and prompts: https://github.com/radiotedu/rtai-jingle/blob/main/ACADEMIC_METHODS_AND_PROMPTS.md
- Security guidance: https://github.com/radiotedu/rtai-jingle/blob/main/SECURITY.md

## Public boundary

The page documents status ingestion and listener context. It does not present public playout controls, pricing, customer claims or invented performance statistics.

## Related pages

- RadioTEDU AI: https://radiotedu.com/ai
- Technology Preview: https://radiotedu.com/technology-new/
- Global LLM guide: https://radiotedu.com/llms.txt