"VeriVoice-V1 Dataset"

"The VeriVoice-V1 dataset is a curated bilingual speech corpus developed to support research in digital identity recognition and verification systems. It comprises high-quality audio recordings from 80 native speakers of English and Urdu, collected under controlled conditions using a fixed, text-dependent identity verification script. Each participant provides seven bona fide (genuine) recordings per language, resulting in a total of 560 audio samples systematically divided into training, development, and evaluation subsets. The dataset is structured to facilitate reproducible experimentation, with organized directories by language, data split, and speaker identity, along with aligned transcript files for each recording.VeriVoice-V1 is specifically designed to enable advancements in speaker verification, speaker identification, and anti-spoofing research, including deepfake detection. The use of a consistent script simulating real-world identity verification scenarios,such as providing personal details and authentication purposes to ensures uniformity while preserving natural speech characteristics. The bilingual nature of the dataset further supports cross-lingual and multimodal identity verification studies.By providing clean, text-aligned, and controlled audio data, VeriVoice-V1 serves as a valuable benchmark for developing robust and secure voice biometric systems. The dataset can also be extended for spoof generation and adversarial research, making it a versatile resource for the speech processing and security research communities."