Shulou(Shulou.com)11/24 Report--

Stability AI rolled up the AI visual circle again! Stable Animation shocking release, the animation effect is amazing.

Stability AI put on a big move again!

Now, Stable Diffusion can also generate videos. The AI class is really too curly.

Yesterday, Stability AI released Stable Animation SDK, a tool designed for artists and developers. Users can create amazing animation effects in a variety of ways.

Before that, Runway's Gen-2 was quite cracked, and the slogan was very domineering-"say it,see it".

Now, Stability AI also has video generation tools comparable to Gen-2!

Stable Diffusion can make videos with the animation endpoints of Stability AI, and artists can use all Stable Difusion models to generate animation, including Stable Diffusion 2.0 and Stable Diffusion XL.

However, unlike the previously released open source model, Stable Animation is currently only available through SDK and paid API.

Now, in all three ways, you can create an animation--

Classic prompt uses text to generate animation just like using Stable Diffusion, Midjourney, or DALL-E 2.

Enter prompt, adjust various parameters, and you can generate animation.

Prompt + images provide an initial image as the basis for animation. After that, the animation of the final output can be generated by combining prompt with this initial image.

Prompt + video provides an initial video as the basis for animation. Similarly, through the adjustment of various parameters, plus the guidance of prompt, the output animation is finally obtained.

For 1 yuan, you can generate 100 frames of animation. The software still seems to be in the testing stage. Stability AI is not available through their browser platforms (such as DreamStudio or ClipDrop), but only software development kits (SDK) and paid API.

Of course, this does not prevent third parties from providing this animation model through services.

At present, however, the bar is high if you want to try this feature, because SDK and API are called through Python scripts.

Similar to Stability AI's image model, Stable Animation can also set many parameters, such as step, sampler, scale, or seed. In addition, features such as outpainting or prompt interpolation are provided.

Different parameters will also affect the price. So it's uncertain how much it will cost to create a video.

Stability offers a range of prices ranging from 3 cents to 18 cents per 100 frames under different settings.

Stability AI charges based on the points of the output animation, and the user-generated animation is converted into a point that costs $1 for every 1000 points.

The editor calculated that according to the officially announced integral calculation method, one yuan RMB can probably generate a 1024-1024 resolution animation of 100 frames.

But other parameters can also affect the price, and officials have given a table to show the parameters that support the adjustment and whether it will affect the price of the animation.

Stable Animation compatible with Stable Diffuision XL can be used with each version of Stable Diffusion.

The default version is 1.5, but we can also choose to use the new version of Stable Diffusion XL.

There are many models, including animation, comic books, low polygons, pixel art.

The original resolution is 512 x 512 pixels, which can be increased to 1024 x 1024 pixels. In addition, a frequency riser can be used.

The default value for Stable Animation is 72 frames (12 frames per second), but according to the instructions, it can be increased to infinity.

Before that, Stability AI already has Stable Diffusion-based animation tools. For example, we can generate a small piece of animation through prompt interpolation (constantly changing some properties of prompt).

The difference of this upgrade is that table Animation's ability to generate video is more comprehensive and mature.

Although AI has not evolved so that we can generate movies at the push of a button, the super evolution of Stable Animation, Runway ML, Phenaki, Imagen Video and other models has predicted that the video generation function of visual generation AI will be more and more amazing.

As soon as netizens' measured Stable Animation was released, netizens could not resist the small hands they were eager to try.

Npaka, for example, shares its own experience.

First, set the API key of DreamStudio to the native environment variable.

Then, prepare the Python virtual environment, install stability_sdk, and start Animation GUI.

$pip install 'stability_ SDK [Anim _ ui]' $python-m stability_sdk animate-- guiRunning on local URL: after opening the displayed URL in a browser, you can create the project in the "Project" tab.

Set the text prompt in the Render tab, and then press the Render button.

Finally, the animation is done.

In addition, some netizens also shared their works.

How far is it from AI to generate a movie? AI is so powerful that some netizens have planned to invest in AI animation studio.

This person said that the brain is ready to wait for a user-friendly animation tool.

The imagination of netizens is not far away. Recently, a lot of AI tools have been used in the new trailer of Wes Anderson's Lord of the Rings.

Andrej Karpathy, the former AI director of OpenAI, guessed that at least Midjourney / Stable Diffusion, ControlNet, ElevenLabs, D-Id, ChatGPT, Adobe and other tools were used in the trailer.

It seems that AI visual tools will not be too far away to replace Hollywood directors if they continue to evolve in this way.

Reference:

Https://stability.ai/blog/stable-animation-sdk

Https://twitter.com/StabilityAI/status/1656691563995512834

Https://the-decoder.com/stable-animation-stability-ai-makes-images-move/

This article comes from the official account of Wechat: Xin Zhiyuan (ID:AI_era)

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.