Discover our premium multimodal text-to-video dataset, meticulously curated to empower the next generation of generative AI models. This comprehensive collection features high-definition video clips precisely paired with rich, highly detailed text descriptions. Engineered for uncompromising visual excellence, every asset maintains a minimum short-edge resolution of 720p, with a substantial portion reaching pristine 1080p HD.
What truly distinguishes this dataset is its breathtaking, cinematic aesthetic. Each video clip is selected to meet the exacting standards of professional photography and blockbuster filmmaking. From flawless exposure and sophisticated color grading to masterful frame composition, the visual quality guarantees an extraordinary level of artistic beauty and visual appeal.
To support diverse and robust AI training requirements, the dataset is strategically divided into three comprehensive categories:
High-Quality ACG: A premium selection of visually striking Anime, Comic, and Games-style video content.
Organic & Natural Worlds: Dynamic footage capturing detailed human subjects, diverse wildlife, vibrant plant life, and sweeping natural landscapes.
Artificial & Urban Environments: Complex man-made scenes, everyday objects, and a wide variety of transportation vehicles.
Offering a perfect synthesis of semantic accuracy and cinematic brilliance, this dataset provides the ultimate foundation for training state-of-the-art text-to-video architectures.
Data Volume: 10 million samples
Data Format: MP4 / MOV video files + TXT / JSON annotation files
Resolution: Short edge ≥ 720p, partial up to 1080p
Visual Quality: Professional photographic & cinematic level, optimized in composition, color and exposure
Categories: High-quality ACG content; humans, animals, plants & natural scenery; man-made scenes, objects & vehicles
