Change the Game With AI
Part 2 – Animations

Bring your characters to life with AI animation tools. AI animations are the next big thing in game animation, harnessing the machine learning techniques of artificial intelligence to majorly improve the quality of movements. AI animations technology is still in its infancy, but when it is truly up and running (pun intended), AI animation tools will have the ability to create flexible natural movements. But where AI animation tools will really revolutionize animations, is that it will be accessible to every game designer, no matter the size of their budget. Check out these three AI animations tools to be at the forefront of game development.

Why are AI animations so powerful?

AI animations enhance your ability to enable your game characters to be flexible and move naturally across the screen. They can do so much more than that. AI animated characters can be taught to interact more naturally with their environments without the need for direct animation or coding. With AI, the movements your characters perform are no longer limited to what human beings (or animals) can do or imagine. The options are endless. AI animations don’t require human models to perform the actions first (as happens now with motion capture), making them accessible to everyone. Basic movements can be applied to all characters, drastically reducing the time it takes to make your characters move across the screen. And most importantly, AI animation tools free up your animators to work on the creative process rather than spending hours or days simply on ensuring that the mechanics are correct.

AI animations tools for the future

Cascadeur – 3D animation software for physics-based character animation

Cascadeur was created by games developer Nekki in response to their need for realistic animated character movements in their games. Described as the future of game animation, Cascadeur makes realistic action animation available to everyone (it’s free!) with convenient, user friendly AI animations tools to create believable actions from scratch. Cascadeur is a combination of an AI neural network reinforced by physics tools which are there to ensure that the movements created by the machine learning model are physically accurate, and to correct them if not. The physics tools are also used in creating more complex movements such as jumps or twists. Cascadeur is not a full neural network. Instead of the AI changing things as it learns, in Cascadeur animators have full control over the end result. The AI simply suggests improvements for existing animations to make them more realistic, and to remove the time taken to manually correct them. As the animator creates the key poses and movements for a character, the system will start to analyse the image for movement trajectories, physical components, and will show the animator improvements and options as the animator works on the pose. The animator can copy some of these suggestions, or continue to make changes, which the system will analyse and suggest further improvements to. This will ultimately save the animator time, and make creating routines far far quicker and more efficient. Cascadeur is free in Beta mode for early adopters. Get it now.

DAIN App

DAIN App is the Depth Aware Video Frame Interpolation method, creating smoother animations, whatever the size of your starting file. DAIN App interpolates new frames into the file in order to increase the frame rate, and improve the final effect. Automating this process means that animators don’t need to spend time creating additional frames for video footage that has a slow frame rate. It is also invaluable to game animators who can now animate fewer frames, and then interpolate them to higher rates without having to put in the time and effort to create every incremental movement, or with sacrificing the quality of their scene. How does it work? DAIN App studies two existing consecutive frames of footage, and calculates an intermediary frame to place between them. The app uses a neural network that has been trained to measure the depth of space in the video, looking at changes between frames, and then calculating how to create a convincing transition frame. Ultimately DAIN App can multiply the number of frames, creating smoother transitions, improving slow motion effects, and allowing low resolution games to be played on high resolution devices. The DAIN App is still in its Alpha version, and only works on Windows. DAIN App requires an Nvidia CUDA card to run the neural network.

NVIDIA OptiX AI-Accelerated Denoiser

This third product is slightly different from the other two. Instead of creating movements or animations, the NVIDIA OptiX AI-Accelerated Denoiser is used to clear images of noise or impurities, creating quicker, sharper scenes. It can denoise multiple image layers simultaneously, making the denoising process quicker and more accurate as it filters every layer in the exact same way. It uses GPU-accelerated artificial intelligence to reduce the time taken to render a high fidelity image, making it sharp, clear, and visually noiseless. This allows game animators to receive real time interactive feedback, so that the animator can see the final image almost immediately, make a decision about it, and finalize images quicker. The OptiX AI-Accelerated Denoiser was trained using tens of thousands of images rendered from 3D scenes to improve the quality of images in a scene. It is particularly useful for scene backgrounds, and scenes that have a lot going on in them. It can also be trained using an animator’s own renderings, teaching it the style and requirements of the user. The AI-Accelerated Denoiser comes as standard within NVIDIA OptiX, which is free for any commercial user.