On the occasion of Vidu's launch surpassing 100 days, BioNum Technologies proudly announces the release of Vidu 1.5, a new version that achieves breakthroughs at a world-leading level, particularly in understanding diverse inputs and overcoming the "consistency" challenge.
The introduction of Vidu 1.5 marks the entry of visual models into a new "contextual" era, accelerating the advent of Artificial General Intelligence (AGI). Since its global launch, Vidu has possessed the capability to generate consistent characters by locking in facial features, addressing a key pain point in video generation. In September, Vidu globally pioneered the "subject consistency" feature, extending facial consistency to full-body consistency and expanding the scope to any subject including animals, objects, and virtual characters. Vidu's technological breakthroughs are mainly reflected in three aspects: precise control over complex subjects, natural consistency of facial features and dynamic expressions, and multi-subject consistency.
Vidu 1.5 demonstrates a new "emergence of intelligence" in visual models, showcasing its powerful contextual learning capabilities. This means that visual models not only possess the ability to understand and imagine but can also manage memory during the generation process. Vidu 1.5 continues its industry-leading generation efficiency, producing a video in under 30 seconds. Vidu adheres to the philosophy of universality, consistent with Large Language Models (LLM), unifying all issues into visual input and output problems, using a single Transformer to model variable-length inputs and outputs, and obtaining intelligence from video data compression.
The launch of Vidu 1.5 not only enhances the controllability of video models but also achieves consistent generation from multiple angles, with multiple subjects, and multiple elements through flexible diverse inputs. This marks the emergence of visual intelligence and accelerates the arrival of AGI. Vidu is no longer just a high-quality, efficient video generator; it can also integrate contextual information and memory during the generation process, a significant leap in visual modality intelligence. Visual models will possess stronger cognitive abilities, becoming a crucial piece in the puzzle of AGI.
Experience URL: www.vidu.studio