Modo renderer gets an overhaul!

Many have been asking “what’s new with the Modo render these days?” 

It’s a question that I’ve actually been eager to answer for a while now and I am really excited to say that next week we will be showing a technology preview of what Allen Hastings, our Head of Rendering at Foundry, has been working on as of late. Next week during the Siggraph 2019 conference the Modo team will be showing a completely re-designed path-tracing based renderer that is fueled by the extremely fast ray-tracing hardware (Optix RTX™) created by Nvidia.

If you want to see the path tracer tech preview in person be sure to sign up for one of the following: 

• Nvidia Limelight  - Date: July 29th 2019 | When: 6:00pm-9:00pm | Venue: Los Angeles JW Marriott
• Nvidia Partner Pavilion - Date: July 31st 2019 | When: 9:30am-12:00pm | Venue: Los Angeles Convention Center, Expo Floor
• Modo Exhibitor Session - Date: July 30th 2019 | When: 4:00pm-6:00pm | Venue: Los Angeles Convention Center, Room 502B | Register Here: http://bit.ly/modosig19

If you're unable to join us at Siggraph you'll have to wait a little longer, sorry... But, we do plan to provide an early access release to customers with active Subscription or Maintenance. Look for this on the Foundry Community in the coming months...

For now and to help explain a bit more about what we’ve been up to I asked Allen a few questions that help clarify the differences and benefits of the current Default Modo Renderer vs the new Path Tracing Modo Renderer:

Q: Allen, can you explain a bit about the history of the Default Modo Renderer and why it was unique for its time?

The renderer in Modo was originally designed in 2003 in efforts to improve upon global illumination techniques and performance pains of existing production renderers in the market. At that time true global illumination was rarely used in production, and the key to high-performance rendering was reducing the number of shading evaluations on directly visible surfaces, while still maintaining high-quality antialiasing (in other words decoupling shading from antialiasing). PRMan's REYES algorithm accomplished this by shading grids of micropolygons at a low density before evaluating visibility at a higher density within each pixel.

The secret to the Default Modo Renderer’s performance benefits was in part by reversing the REYES order and sampling visibility at high density first, and then merging nearby similar samples before a deferred shading phase (an idea that was later patented by Luxology). Without the need to dice all surfaces into micropolygons, this method was more friendly to ray tracing, which proved beneficial as the popularity of physically based rendering grew over the following years.

Q: In recent years we have seen many commercial and in-house rendering solutions shift towards a path tracing paradigm. How is this technique different than the REYES approach that the Default Modo Render uses?

Path Tracing is an alternative approach with distinct advantages of its own. Shading is evaluated at every visibility sample, but typically only one secondary ray is spawned for each shading point. If this ray hits a surface, another ray may be spawned from the new hit point, until finally, a ray reaches the background or a light source, or until some bounce limit is reached, or until it is randomly terminated when its importance drops below a certain level. Each primary ray from the camera along with its chain of secondary rays can be thought of as a path, so this method is known as path tracing, and the rays and their hit points are often called path segments and path vertices. The method was first presented by Kajiya in 1986 as a numerical solution to the "Rendering Equation" but it was not practical for production use at that time.

Q: In your opinion what are some of the advantages that seem to be driving this trend towards path tracing?

Path tracing is easier to use for artists and requires fewer settings to learn, the main one being the number of samples (paths) per pixel. Much in the same way that point and shoot cameras changed photography the settings for a path tracer renderer could be simplified to the point of a single "quality knob" (although that's a bit of an exaggeration).

Path tracing is well suited for physically based rendering, making the lighting process more like real-world photography and thus more predictable for artists and more consistent from scene to scene.

Path tracing is usually unbiased, which means there is no systematic error which can cause blotches or flickering in animations. Instead, the error is in the form of random noise that diminishes as the number of samples increases.

Path tracing is a single pass method without the need for precomputed shadow maps or GI caches, simplifying studio pipelines.

Path tracing lends itself to progressive refinement, in which the entire image is quickly rendered with a small number of samples and additional samples are added over time, either uniformly or concentrated on particular pixels until the desired quality is achieved. This allows artists to see start seeing results right away, reducing their iteration time. A sample index can also be saved along with an incomplete image, enabling a render to be paused and resumed later.

Path tracing affords excellent scalability with multiple threads because each path is computed independently. The straightforward nature of the algorithm is also attractive for GPU implementations.

Path tracing supports rendering multiple output images (AOVs) containing specific types of lighting. Because radiance contributions at each path vertex are individually added to the frame buffer, this differentiation can be done in one pass with very little overhead. Render outputs can be defined that include only radiance from a particular light or group of lights, or light that has bounced a particular number of times, or reflected from a particular surface, etc. One way to specify these choices is through "light path expressions", a kind of pattern matching language.

Q: The Default Modo Renderer has certainly performed well in the process of creating high-quality photorealistic images. What are the main benefits an existing Modo user might find most useful?

In some ways, the shift to path tracing is not as revolutionary for Modo as it is for other renders using a REYES-type approach, since the current Modo renderer already supports physically based rendering, progressive refinement (in Preview), and has good scalability with more CPU cores. But updating the renderer is still worthwhile for the other benefits of path tracing, particularly the simplification of render settings, the light-specific render outputs (LPE), and opening up the possibility of GPU acceleration and various material description standards. Naturally, these features would be desirable in a variety of market use cases for the Modo Renderer.