Vidu 1.5 offers a major leap for filmmakers to create a consistent look and feel within an AI generated video.
Written by Tom White
From digital twins to game design development, generative AI has made some promising progress in the media and entertainment industry. But when it comes to video creation, the lack of coherence between frames has been a major hurdle, holding back many film producers and content creators who might otherwise incorporate the technology into their creative process.
The open secret is that post-production is a necessity if you’re inclined to use generated videos. But with Shengshu AI’s latest Vidu 1.5 update, the future of film production may soon look very different.
Vidu was ahead of the game, really being the first to launch after Sora – and that’s ahead of rival Kling. But compared to other text-to-image video multimodal models, Vidu 1.5 offers a major leap for filmmakers to tackle a major pain point: landing that consistent look and feel within an AI generated video between or in the middle of a movement.
With this challenge in mind, Shengshu AI’s Vidu 1.5 in many ways is a breakthrough. It offers more fluid continuity in character appearances, actions and scene transitions. Without getting too technical, Shengshu’s algorithm uses advanced semantic understanding, allowing for greater precision and control over both the generated characters and scenes, by supporting more nuanced text prompts. This means less manual post-production on AI footage by up to 50%, the company claims.
From camera angles, character actions to their expressions, Vidu 1.5 promises to interpret complex instructions and generate detailed, cohesive video footage from start to end. Taking a more granular look at its features, Vidu 1.5’s Multiple Image Consistency feature enables its users to upload up to three images from different angles of a subject (like a person). Vidu 1.5 will then be able to generate a single cohesive video and ensure 360 degrees of consistency without any blind spots, no matter the complexity in detail. And this also applies to facial expressions as well, which now look more natural and less jarring or “uncanny valley.”
Adding to this, its Advanced Character Control feature offers a few additional benefits. Users can enjoy the ability to more precisely control cinematic angles and camera movements through text inputs, so that zooming, panning, tilting and rotating, or a combination of the above are now possible. You can even prompt dynamic ranges of motion within a scene or for a character, making their movements feel more human or more natural. This is compared with other AI models that instead of filling in gaps between movements with AI, they take short cuts and use slow motion.
Compared to glitchy and grainy clips made by early AI video generators, the new Vidu version also permits filmmakers to output crisper 720p or 1080p videos. By supporting a high-resolution generated video, smaller studios, independent artists or just about any user can now level the playing field with a video quality that may soon rival professional studios.
Animation is another area where Vidu 1.5 is significantly improving the production cycle. Amid growing appeal among 2D animators and artists, Vidu is adding support for Japanese fantasy and hyper-realistic anime genres, along with added support for special effects like flame visuals and realistic lighting. It’ll enable creators to iterate scenes more easily, streamlining workflow with efficient A/B testing, and putting concepts to test quicker.
It seems that it could only be a matter of time before Vidu ends up being one of the standards in the film industry.
Photo Credit: Shengshu AI
Videos