Chaos

What are you looking for?

© xoio
© xoio

Exploring the dynamic role of AI in Xoio’s CG pipeline


Explore AI's potential in archviz and learn how xoio integrates it into their workflow, giving you insight into both its benefits and challenges. 


As the CG field develops rapidly, innovative companies are looking for new ways to improve their workflows and produce outstanding results with AI. This blog article delves deep into the use of AI by the archviz studio xoio to transform its workflow and explores the benefits the studio has gained from this integration.

comp-singleimage-Sleeppod_Lake_FINAL_by_xoio.png
© xoio

AI integration in xoio's CG pipeline

Initially, the perception of AI was that it would be disruptive and replace current workflows entirely. However, xoio has taken a more nuanced approach. The archviz studio has implemented several AI applications that fit seamlessly into conventional workflows, enhancing steps of the process and offering alternative options. 

xoio has established a system to differentiate between image components and the allowable AI contribution. On one end, there are components where the render must be kept as close to the origin as possible, such as a product or architectural representation. On the other hand, there are components that allow for AI creativity to a larger extent due to the lack of specificity. Between these two extremes are areas where various levels of AI assistance might be useful, such as making a texture look more natural or replacing a stand-in object.

© xoio. The heatmap represents areas that are off-limits (red), ranging all the way to green areas that would allow AI to handle them very creatively (green).

While embracing AI brings forth numerous benefits, xoio acknowledges the challenges encountered during integration. The rapid pace of AI development presents a daunting task for studios to stay abreast of advancements while maintaining operational efficiency. xoio stresses the importance of investing in R&D to overcome these challenges. 

comp-singleimage-Heatmap_Selective_AI_4_Result.png
© xoio

I guess the biggest challenge concerning AI is the crazy pace of development. Studios and individual artists alike are overwhelmed by the demand for flexibility while established workflows are put in question. This is exhausting! The only way to face this is to accept the challenge and invest some part of your power in R&D. Otherwise, you will fall behind.

Peter Stulz, CEO, xoio

The role of Corona in xoio's workflow

xoio's main render engine, Chaos Corona is the main backbone of their daily work, and the biggest part of their images are based on results from Corona. In conjunction with AI, Corona acts as a counterweight to AI – a sort of sparring partner. Even in areas that are partially enhanced by AI, Corona provides xoio with a solid base for further refinement. Additionally, the render elements can be used to enhance the AI creation process. 

Corona is perfect for us because it delivers uncompromising photorealism in areas where high accuracy and quality are required.

Peter Stulz, CEO, xoio

Corona’s materials and lights, in conjunction with its worry-free default render setup, deliver brilliant render results out of the box and make it the top choice for xoio as the main workhorse. 

Experimenting with Stable Diffusion and Magnific AI

xoio's Swiftpod project, an internal design study, is a prime example of the company's innovative use of AI. In mobility studies like these, the studio emphasizes storytelling, and AI helps flesh out the images. This allows the team to focus on other important aspects of the project.

xoio used Stable Diffusion for the bigger scale variations and Magnific AI for the detailed image refinement. Both of these tools were combined to create the Swiftpod illustrations.

In every studio, you will have illustrations that just don’t work out. We found that throwing them at Stable Diffusion and playing with the prompt and the creativity can yield helpful and inspirational results.

Peter Stulz, CEO, xoio

By leveraging Stable Diffusion, xoio's team navigated challenges in project development, gaining insights into areas needing improvement, such as lighting and color composition. Even when faced with AI misinterpretations, xoio was able to adapt its approach, enhancing its CG version based on insights from the AI-generated version. 

comp-singleimage-Backplate_Enhancement_2_Selection.png
© xoio

In the example below we needed to give some background skyscrapers a more futuristic look. So, we simply painted some rough color ideas and reflections on top of our render output, and Stable Diffusion created some variations. A selection was then manually patched on the original render.

Peter Stulz, CEO, xoio

© xoio

Magnific AI is a very approachable and easy-to-use web solution that gives the final touches to surfaces and details, especially small variations and imperfections that are hard to pull off. The tool does this within a minute. So you take your final image, enhance it in Magnific AI, and then decide by masking off in Photoshop with masks which parts you want to keep.

Peter Stulz, CEO, xoio

© xoio

Revolutionizing CG people with AI

By harnessing the power of AI, xoio redefined the way it integrates CG people. With Stable Diffusion and Magnific AI, xoio brings 3D-scanned individuals to life, achieving unparalleled photorealism. As Peter explains, this shift reduces reliance on traditional cutout people, offering a more seamless integration into scenes while enhancing overall realism and visual cohesion. A great side effect of AI engines is that they understand light setups and materials pretty well, as seen in the final result. 

Fueling creativity through AI variations

xoio shares insights into how AI prompts creative variations, such as replacing CG characters using Chaos Cosmos. By experimenting with textual prompts, xoio unleashes a wave of creative possibilities, resulting in dynamic and innovative outcomes.  

Cosmos was great for quickly dropping in a 3D character, which then offered a solid base for AI enhancement. Playing with the prompt resulted in a firework of funky variations in absolutely no time. That was great fun and felt very powerful!

Peter Stulz, CEO, xoio

For the Swiftpod project's onboarding illustration, a two-stage AI Process was used to enhance the CG characters. First, Stable Diffusion was used to turn the 3D scans into believable models. In the final stage, Magnific AI added that extra detail to the skin and clothes and ironed out the usual inconsistencies with faces. 

© xoio
comp-singleimage-Sleeppod_Downtown_FINAL_by_xoio.png
© xoio

Envisioning the future of AI in CG pipelines

Looking ahead, xoio recognizes the vast potential AI holds for the future of CG pipelines. With an expanding array of tools and applications, creators are empowered to streamline workflows and unlock new levels of creativity. The studio emphasizes, however, that in this day and age of technological innovation, the constant need to consider and evaluate the many choices popping up can be quite overwhelming.

Words of wisdom for fellow creatives

xoio's innovative approach to integrating AI into its CG pipeline demonstrates the immense potential of AI in enhancing workflows and delivering superior results. The successful use of AI in the Swiftpod project demonstrates the transformative power of AI in the realm of CGI. Drawing from experience, Peter offers some valuable advice to studios and artists:

Get started with AI now (if you haven't already) and dedicate some time to research and development. Also, try to stay open-minded, curious, and positive. There is no point in being afraid of AI. But you have to be prepared to throw some old workflows overboard and learn new ones. Technological revolutions have happened before - so will this one, there is no stopping it.

Peter Stulz, CEO, xoio

Peter also recommends not losing sight of traditional skills if you're overwhelmed by the technological buzz. Try to find time for some old-school techniques that you enjoy and that bring you peace. Peter tries to keep up his sketching habit to counterbalance his creative toolset. According to Peter, at the end of the day, it is all about combining the old, the new, and whatever you bring along as a unique individual.  

Begin your rendering journey.

Discover an easy and seamless workflow for photoreal renders.
Chaos
© 2024 Chaos Software EOOD. All Rights reserved. Chaos®, V-Ray® and Phoenix FD® are registered trademarks of Chaos Software EOOD in Bulgaria and/or other countries.