This dataset contains 88,000+ professionally sourced video recordings with high-quality audio and automated transcripts, spanning 44+ languages and speakers from 120+ countries. Each recording is delivered as MP4 video and Opus audio with accompanying per-recording metadata covering demographics, locale, language fluency, professional background, technical capture details, transcript text, integrity hashes, and consent status. Quality tiers and human-in-the-loop verification are available for higher accuracy. The dataset supports speech model training, emotion and speaker analysis, multimodal research, and regional performance evaluation, enabling mapping and geographic clustering by country, region, or city.

Key attributes include:

• High-fidelity MP4 video with Opus audio and 48 kHz sampling

• Whisper-generated transcripts plus optional human-verified tiers

• Speaker demographics and professional background across global coverage

• Location metadata enabling regional and country-level analysis

• Technical and integrity metadata including sample rates, SNR, and SHA-256 hashes

Machine learning teams, speech technology developers, and product owners benefit from a large, consented corpus for training and evaluating ASR, TTS, speaker recognition, emotion detection, and multimodal systems; the dataset reduces data collection time and improves geographic and accent coverage.