Description
With recent advances in generative models, many industries must adapt to offer more-intuitive solutions to interact with their expert software. The challenge ahead is to combine this expert knowledge with artificial intelligence (AI) to include more-natural ways of interacting with our tools while remaining precise and controllable. In this talk, we will discuss our solution to achieve this for content creation in Maya software. We'll present a high-level overview of the architecture to better understand the integration of the Microsoft Azure OpenAI services in our product. Then, we will discuss how we improved the inference model bridging the gap between research and production. We'll also show a demo of our plug-in in Maya software on a complex scene. You'll leave this session knowing how to prepare a proper data set for code generation from prompt, how to design a structured context improving the accuracy of the generated response, and how to maximize custom embeddings to handle more-complex use cases.
Key Learnings
- Learn about the Azure OpenAI services.
- Learn how to prepare and augment your data set for prompt-to-code generation purposes.
- Learn how to design the proper context for prompt engineering.
- Learn how to maximize custom structures using embeddings for more-complex use cases.