Hello, Vidu!

Vidu, a new groundbreaking AI video generation model, has been recently introduced at the 2024 Zhongguancun Forum in Beijing. This large AI model is a collaboration between Tsinghua University and the Chinese AI startup ShengShu Technology. Marking a first for China, Vidu features extended duration, exceptional consistency, and dynamic video generation capabilities.

Developed domestically, Vidu excels at processing and creating content that includes culturally significant elements like pandas and the Chinese dragon, noted Zhu Jun, the Deputy Director of the Tsinghua Institute for Artificial Intelligence.

The foundational architecture of Vidu was initially proposed back in 2022, according to the company. The model's architecture seems to be based on U-ViT, which is similar to the Diffusion Transformer used by SORA by OpenAI.

If you're interested in the brains behind this, here's the Google Scholar profile of the scientific director who led this project.

Vidu AI Capabilities and Video Quality

Vidu, a new competitor to OpenAI's Sora, is capable of generating a 16-second video in 1080p resolution. It is designed to transform text descriptions into dynamic, high-quality videos. This text-to-video AI model does more than just visually interpret the content at basic level. It can create videos from text with a full range of scenes, characters, and actions based on the input text, making the videos impressively realistic. Vidu can generate video sequences that illustrate the story or instructions described in the text, complete with appropriate settings, interactions, and movements tailored to the storyline.

During a live demonstration, Vidu was able to mimic the real physical world, creating scenes that follow actual physical laws, including realistic lighting and shadow effects, and detailed facial expressions. Additionally, it can produce complex moving shots rather than just static ones.

Now, let's watch the video presented by Chinese developers:

ChatLabs team opinion: Exciting! If the demo is real, then the quality of l the large AI model Vidu is just a step away from the quality of SORA's generation.

The quote from press-release:

Since the release of Sora, the battle for "domestic Sora" has begun. But when the industry focuses on the "long" feature, they all ignore that behind Sora is actually the improvement of comprehensive effects, such as consistency, realism, aesthetics, etc. in long time series.

From the perspective of comprehensive effects, "Vidu" is the first and only video model to fully benchmark against Sora at the effect level, not only domestically, but also globally. It is also the first video model to achieve a breakthrough after Sora.

How to Use Vidu AI – Possible Or Not?

So, the video looks extremely promising, but the main question is different – can users get access to Vidu and test its capabilities? To put it briefly, yes and no. The new Chinese model is not freely accessible just by a link, however, anyone interested can apply for consideration for access to the new video AI model.

To apply for access to Vidu:
1. Follow the link: https://www.shengshu-ai.com/home
2. Click the blue button in the top right corner.
3. You will see a simple form with a few fields. Fill it out.

Vidu: Questions and Answers

• What is Vidu? A new advanced AI video generation model presented by Tsinghua University and ShengShu Technology that produces realistic and imaginative 16-second videos in 1080p resolution.

• How does Vidu work? It uses a sophisticated AI architecture combining diffusion and transformer models to generate videos directly from textual prompts.

• How can one access Vidu? Access is typically through collaboration with Shengshu Technology or via its commercial platforms. Visit https://www.shengshu-ai.com/home

• What are Vidu's uses? Ideal for film, media production, advertising, and creative arts for generating unique, realistic content.

• Can Vidu handle complex video tasks? Yes, it excels in producing coherent, detailed, and dynamic video content for complex scenes.

Pros and Cons of Vidu

Advantages:
– Coherent Narratives: Blends shots smoothly for coherent storytelling.
– Realistic Physics: Simulates real-world physics effectively.
– Creative Visuals: Creates imaginative, non-existent scenes.
– High-Quality and Evolving: Already produces content close in quality to Sora AI and is continuously improving.

Disadvantages:
– Missing Details: Sometimes overlooks small but important details.
– Dynamic Inconsistencies: May struggle with complex dynamic scenes.
– Resource Intensive: Requires significant computing resources, limiting broader use.

Will Vidu Be Available In ChatLabs AI?

ChatLabs offers a wide range of over 30 different AI models, including well-known ones like GPT-4, Gemini Pro 1.5, and Claude 3 Opus, and we're always adding the newest AI tools to our collection. We quickly integrate new models, typically within just 1-2 days, much faster than our competitors.

We've already requested access to the Vidu model and are currently waiting for a response. We encourage our users to keep an eye out for updates and be ready to explore the latest AI technology at ChatLabs.

We're looking forward to more details about this model!

More Articles by ChatLabs

• Details about Phi-3 – a new AI model by Microsoft

• GPT-5 Overview and Release Date

• How to Get Access to Llama 3