| license: other | |
| license_name: rtcfake-license | |
| license_link: LICENSE | |
| task_categories: | |
| - audio-classification | |
| tags: | |
| - speech-deepfake-detection | |
| # RTCFake | |
| This repository contains the **RTCFake** dataset, presented in the paper [RTCFake: Speech Deepfake Detection in Real-Time Communication](https://huggingface.co/papers/2604.23742). | |
| RTCFake is a large-scale speech deepfake dataset tailored for real-time communication (RTC) scenarios, totaling approximately 600 hours. The dataset is constructed by transmitting speech through multiple mainstream social media and conferencing platforms (e.g., Zoom), enabling precise pairing between offline and online speech. | |
| The dataset is divided into training, development, and evaluation sets. The evaluation set includes both unseen RTC platforms and unseen complex noise conditions, providing a realistic and challenging benchmark for speech deepfake detection. |