Absentia isn’t exactly your ‘new-kid’ on the startup block. They’ve been around for a while, first coming into limelight around a couple years back for their Absentia VR platform and the Tesseract HMD that promised to change the VR landscape. Fast forward to 2017- after encountering some capital related issues with VR and stiff competition from international giants, they’ve shifted their focus to a different landscape – AI based game development and animation. And that’s where Norah.ai comes in.
What is Norah.ai?
According to Absentia, Norah.ai is a platform where “just a few lines of text/pictorial input can create a completely new set of interactive content such as Arcade games, Casual games, Puzzles, Animation, Graphics” etc. In its current state, Norah AI is a platform for web apps that can help you fulfill your basic 3D model animation needs quite easily, as long as the model is humanoid.
The animation section of Norah.ai currently comprises of four major components:
- Auto Rigger: This tool rigs your humanoid model with a skeleton without having to provide any additional instructions. All you need to do is upload a compatible .obj file where the model is in the ‘T’-posture and click on ‘Rig it’ and let the AI do the rest. Once the user uploads the 3D model, the AI first detects a humanoid from the vertex data points of a body and automatically detects the corners of body parts with their labels for rig placement and thereby outputting the rigged 3D model as output.
- Animation Repository: A huge library of locomotion, dance forms and combat animations catering to various requirements.
- Style Transfer: This part of the tool allows you to combine two animations to create a unique output. Say, for instance, you want to combine a Bharatnatyam move with a twerking move – if you have the two originals, you’ll get your result here. Although, there’s more to this than just selecting two animations to be combined.
Once the user selects two animations he is given an option to adjust the blend controls of each and every body part. The new animation generated from them happens almost in real time since AI has learned the representation of a complex animation data into a simpler form which formed a part of training it.
- Motion Editor: This part of the tool lets you chose an animation from the vast repository and tweak it according to your own requirements, playing with parameters as detailed as the planar orientation of each component of the skeleton at any point in the animation along the x,y and z axis.
Keep in mind that none of these components are complete animation tools by themselves. You’ll mostly be able to download the results to be used with an animation engine or a game development engine as you see fit. On the other hand, with something as simple as a character building tool in an RPG like Skyrim or Mass Effect, you’re getting an effective middleware platform that can be used to generate assets for all kinds of 3D-animation experiences.
And coming up…
As of now, that platform seems to be free as far as we have explored it. Going forward, Absentia will charge developers for downloading the results. And that’s not it – the startup has some other lofty ideas as well.
Along with the animation section, the platform will include a 3D modeling as well as a game making section in the future. Absentia plans to employ the neural network based AI approach to the upcoming sections as well, with a goal to eventually reshape the world of game design with a functionality that allows for automatic generation of animations, game maps, textures, characters, and stories.
How would that work?
For the upcoming modules of 3D modeling, extensive use of auto encoders comes into play since it involves procedural generation with semantic outputs. The system again here learns to encode its own input to learn the representation of a 3D model data for further computation. There also comes a requirement for a noise network here to generate sensible outputs and defining the threshold limits of each body part for sensible output data.
The AI then generates a countless number of 3D models at each input by the user, generating unique and sensible 3D models which can be directly imported into a 3D software/Game engine.
The complete workflow would be driven by AI playing the most part of it from modeling to animation to mission generation and deciding the NPC(Nonplayer character) behavior based on its mental state and also keeping in mind the asset optimisation part for Mobile and Other output platforms.
Will it happen?
If you were to tell animation engineers ten years ago that complex 3D animations would eventually be done in a browser window, you would promptly be thrown out of the animation studio. Right now, it is possible and it is free with Norah AI’s tools. So there’s no reason to believe that the rest of their upcoming toolset won’t be equally useful. Afterall, ‘outlandish’ ideas seem to be their signature. Let’s just keep our fingers crossed and hope that it’s reasonably priced.