Published on November 22, 2023, 6:18 pm

Introducing Stable Video Diffusion: High-Quality Ai-Generated Videos From Text Prompts

Stability AI, the developer of Stable Diffusion, is introducing a new generative AI technology called Stable Video Diffusion. This AI system consists of two models, SVD and SVD-XT, which can generate short-form videos based on a text prompt. The resolution of these videos is 576 x 1,024 pixels, and users have the flexibility to customize the frame rate speed between three and 30 FPS.

The length of the videos depends on the selected model. If you choose SVD, the clip will last for 14 frames, while SVD-XT extends it to 25 frames. However, regardless of the length, the rendered clips are designed to play for about four seconds before ending.

Stable Video Diffusion has impressed viewers with its high-quality results. Unlike some other AI-generated content that can be unsettling or unrealistic, Stable Video Diffusion produces visually appealing videos. One noteworthy demo showcases an Ice Dragon with intricate scale details and breathtaking mountain scenery in the background. While animation features are limited to a slow head bobbing motion or a slow panning shot in other demos, they still exhibit impressive visual quality.

Despite its achievements, Stable Video Diffusion does have limitations. It may not achieve perfect photorealism and struggles with generating legible text or accurately rendering faces. However, there are instances where the model successfully renders human faces without any flaws, suggesting that it might depend on individual cases.

It’s important to note that Stable Video Diffusion is currently in its early stages and not ready for wide release or commercial use. Stability AI emphasizes that it is intended for research purposes only at this time. The developer is taking precautions after a previous incident where their diffusion model was leaked online and misused to create deep fake images.

If you’re interested in trying out Stable Video Diffusion, you can join a waitlist by filling out a form on Stability AI’s website. While there’s no information on when access will be granted, the preview will include a Text-To-Video interface. In the meantime, you can explore the AI’s white paper to delve into the technical details and learn more about the project. Interestingly, the white paper mentions using publicly accessible video datasets as part of the training material, indicating an effort to be more cautious given previous legal challenges.

There is currently no launch date announced for Stable Video Diffusion. However, there are alternative options available. TechRadar has compiled a list of the best AI video makers for 2023 that you can explore.

In conclusion, Stability AI’s Stable Video Diffusion is an exciting generative AI technology that shows promise in creating short-form videos based on text prompts. While it still has some limitations, especially in achieving photorealism and accurately rendering faces or text, its high-quality visuals and ongoing research demonstrate its potential. With proper precautions in place, Stability AI aims to prevent misuse of its technology and ensure ethical usage.


Comments are closed.