Autodesk AI Lab unlocks new AI-powered design tools

Whether it’s due to scenarios like a poorly placed mic, excessive background noise or, low-quality archive footage or any other reason, poor audio can be very difficult to clean up. Enhance Speech uses AI to magically remove background noise and improve the quality of poorly recorded dialogue, making it sound as if it was recorded in a professional studio. You can also use the mix slider to incorporate some of the background noise to make it sound just right. Create 3D content automatically at scale without any physical scanning or a team of 3D designers, to speed up processes and go to market faster. Style2Fab is driven by deep-learning algorithms that automatically partition the model into aesthetic and functional segments, streamlining the design process. Have you ever wondered how physical objects with hundreds of parts are assembled in CAD?

As 3D printers have become cheaper and more widely accessible, a rapidly growing community of novice makers are fabricating their own objects. To do this, many of these amateur artisans access free, open-source repositories of user-generated 3D models that they download and fabricate on their 3D printer. Creators of all levels can tap into these resources to produce high-quality outputs that meet the growing demands for content and virtual worlds in the metaverse.

Top requests from the community

Formed in 2018 as a part of Autodesk Research, the Autodesk AI Lab conducts fundamental and applied research in AI and machine learning with an aim to unlock a new era of AI-powered design tools for our customers. We’re committed to growing our reputation in the AI research community by expanding our teams, partnering with notable organizations, and publishing cutting-edge AI research. However, the deeper promise of this work is that, in the process of training generative models, we will endow the computer with an understanding of the world and what it is made up of. In addition to generating pretty pictures, we introduce an approach for semi-supervised learning with GANs that involves the discriminator producing an additional output indicating the label of the input. This approach allows us to obtain state of the art results on MNIST, SVHN, and CIFAR-10 in settings with very few labeled examples. On MNIST, for example, we achieve 99.14% accuracy with only 10 labeled examples per class with a fully connected neural network—a result that’s very close to the best known results with fully supervised approaches using all 60,000 labeled examples.

The common ingredient across AI pipelines from reconstruction to simulation is that meshes are generated from an optimization process. At each step of the process, the representation is updated to match the desired output better. Generative AI will touch every aspect of the metaverse and it is already being leveraged for use cases like bringing AI avatars to life with Omniverse ACE. Many of these projects, like Audio2Face and Audio2Gesture, which generate animations from audio, have turned into widely loved tools in the Omniverse community. An impressive feature of this suite of tools is the ability to generate animations.

Our papers focus on making designers’ jobs easier by using AI to reverse-engineer objects and assemblies into CAD models, as well as generate the designs themselves. The next two recent projects are in a reinforcement learning (RL) setting (another area of focus at OpenAI), but they both involve a generative model component. We show some example 32×32 image samples from the model in the image below, on the right. On the left are earlier samples from the DRAW model for comparison (vanilla VAE samples would look even worse and more blurry). The DRAW model was published only one year ago, highlighting again the rapid progress being made in training generative models.

More about MIT News at Massachusetts Institute of Technology

Generative AI models can take inputs such as text, image, audio, video, and code and generate new content into any of the modalities mentioned. For example, it can turn text inputs into an image, turn an image into a song, or turn video into text. Generative AI models use neural networks to identify the patterns and structures within existing data to generate new and original content. Alpha3D allows us to generate high-quality and accurate 3D models in a short amount of time, helping us speed up our processes significantly. Overall, we found Alpha3D to be extremely valuable and would recommend it to anyone who needs to generate high-quality 3D content at scale.

In response, workers will need to become content editors, which requires a different set of skills than content creation. We have made a tool, the Sloyd Maker, to create these in a way where parts are reusable, so if you are building a new object that needs for example a handle, a scope, or a hinge, chances are you can use an existing generator. This way, our generators can create so many variations that its very hard to get exactly the same result as someone else.

Yakov Livshits
Founder of the DevEducation project
A prolific businessman and investor, and the founder of several large companies in Israel, the USA and the UAE, Yakov’s corporation comprises over 2,000 employees all over the world. He graduated from the University of Oxford in the UK and Technion in Israel, before moving on to study complex systems science at NECSI in the USA. Yakov has a Masters in Software Development.

To help makers overcome these challenges, MIT researchers developed a generative-AI-driven tool that enables the user to add custom design elements to 3D models without compromising the functionality of the fabricated objects. A designer could utilize this tool, called Style2Fab, to personalize 3D models of objects using only natural language prompts to describe their desired design. To get around this limitation, the Point-E team trained an additional AI system to convert Point-E’s point clouds to meshes.

Though quicker than manual methods, prior 3D generative AI models were limited in the level of detail they could produce. Even recent inverse rendering methods can only generate 3D objects based on 2D images taken from various angles, requiring developers to build one 3D shape at a time. Mesh representations offer many benefits, including support in existing software packages, advanced hardware acceleration, and supporting physics simulation. However, not all meshes are equal, and these benefits are only realized on a high-quality mesh. Digital twins in the metaverse are providing physically accurate virtual environments that allow developers to simulate and test AI for software-defined technologies such as intelligent robots faster than ever before.

NEW!3DFY Prompt – from text to 3D model in an instant

3DFY AI uses advanced generative AI to produce high-quality 3D models from textual descriptions. By eliminating the need for costly, time-consuming, and impracticable manufacturing or scanning Yakov Livshits methods, 3DFY AI has made the creation of 3D content accessible to everybody. Masterpiece Studio only requires a few lines from its users to generate fully functional 3D models and animations.

But 3D assets, especially for games, need to be optimized and have great topology. That’s why we are making this dataset, by building parametric generators which we can combine with ML to create everything, in real-time, with game ready results. DCGAN is initialized with random weights, so a random code plugged into the network would generate a completely random image.

Bibliographic and Citation Tools

This game-changing program has the simplest UI of any 3D production software currently available, making it accessible to users of all skill levels. This solves the problem of finding random 3D assets in libraries or asset packs and having issues with details, whether it’s an issue with conversions or UV mapping. We believe that providing creators with these tools will both lower the barrier to entry for less experienced creators and free more experienced creators from the more tedious tasks of this process. This will allow them to spend more time on the inventive aspects of fine-tuning and ideating.

Earlier this year, Google released DreamFusion, an expanded version of Dream Fields, a generative 3D system that the company unveiled back in 2021. Unlike Dream Fields, DreamFusion requires no prior training, meaning that it can generate 3D representations of objects without 3D data. NVIDIA researchers trained GET3D on synthetic data consisting of 2D images of 3D shapes captured from different camera angles.

• As an evolving space, generative models are still considered to be in their early stages, giving them space for growth in the following areas.
• End users should be realistic about the value they are looking to achieve, especially when using a service as is, which has major limitations.
• For example, popular applications like ChatGPT, which draws from GPT-3, allow users to generate an essay based on a short text request.
• The revelation helps shed more light on the often unacknowledged human labor crucial to generative AI.
• It manipulates the aesthetic segments of the model in Style2Fab, adding texture and color or adjusting shape, to make it look as similar as possible.

It took the team just two days to train the model on around 1 million images using NVIDIA A100 Tensor Core GPUs. FlexiCubes generates high-quality meshes from neural workflows like photogrammetry and generative AI. GET3D will be available in Omniverse AI ToyBox along with existing generative AI research projects published by NVIDIA, such as GANVerse3D Image2Car and AI Animal Explorer.