Song voice conversion tools have gained more and more popularity in the recent past. People have been uploading their self-made forgery songs on video websites, and these songs have been converted in timbre. However, singing voice conversion technology may cause copyright infringement of the songs. In order to protect the copyright of songs, the method of singing voice conversion detection needs to be investigated. We propose Rhythm and Pitch Aware Songs Conversion Detection (RPA-SCD), a dual-branch network for song voice conversion detection. RPA-SCD can predict forged song fragments through rhythm and pitch which are the global and local information of music. To evaluate the proposed method, we contribute a multilingual song conversion detection(MSCD) dataset. Our proposed model achieves the EER of 2.30% in the original domain of MSCD, which is lower than other benchmarks for speech forgery detection. The experiments show that our approach achieves state-of-the-art performance on the song conversion detection task. The MSCD dataset can be found at https://drive.google.com/file/d/1rFsvMYihVtk81uFbL7UpyUEs-qBgsX6H/view?usp=drive_link. The code can be found at https://github.com/Samantha-Du/RPA-SDD.

The emergence of artificial intelligence-generated content (AIGC) has raised concerns about the authenticity of multimedia content in various fields. However, existing research for forgery content detection has focused mainly on binary classification tasks of complete videos, which has limited applicability in industrial settings. To address this gap, we propose UMMAFormer, a novel universal transformer framework for temporal forgery localization (TFL) that predicts forgery segments with multimodal adaptation. Our approach introduces a Temporal Feature Abnormal Attention (TFAA) module based on temporal feature reconstruction to enhance the detection of temporal differences. We also design a Parallel Cross-Attention Feature Pyramid Network (PCA-FPN) to optimize the Feature Pyramid Network (FPN) for subtle feature enhancement. To evaluate the proposed method, we contribute a novel Temporal Video Inpainting Localization (TVIL) dataset specifically tailored for video inpainting scenes. Our experiments show that our approach achieves state-of-the-art performance on benchmark datasets, including Lav-DF, TVIL, and Psynd, significantly outperforming previous methods. The code and data are available at https://github.com/ymhzyj/UMMAFormer/.