The Flame Learning Channel

In the Flame 2020.1 update, another type of machine learning has been introduced into the Flame products. So in part 4 of the Machine Learning Series, you’ll learn how to extract the sky out of a shot. So Flame can look at an image and using Machine Learning identify what parts of that image make up the sky. You can then use that extraction as part of a selective or composite. For example, you may want to grade the sky or perhaps even do a sky replacement. We’ll run through two examples and discuss some of the finer details as the video progresses.

When it comes to animated masks in the Flame products, one question that always gets asked is how do you offset a vertex that already contains animation? For instance, the animation may have been tracked or key-framed but now you’d like to move the same vertices and keep the original animation. This video will cover the vertex offsetting technique for both the Gmask Tracer and the older Gmask Toolset.

In this video, you'll learn how to use the refractive properties of a surface to distort an image. This is different to previous examples where you have a semi-transparent object and light is refracting through it. Instead, you’ll use refraction to create localised distortion and not see any 3D objects. This workflow video will cover extended bicubics, refraction, masks and tracking. Now Flame already has other tools that can perform distortions and warping and this technique is another option for you to choose from.

Refraction in Action has introduced some really creative opportunities but did you know that its not just limited to 3D models? In the new video example added to added to the Flame Learning Channel, you learn how refraction can work with a flat object such as a surface object or a flat 3d plane geometry. You'll also see how extended bicubics can affect light refracting through a surface. Finally we'll discuss what happens when you have multiple overlapping objects in composite and they have refraction!

This video shows you how to add custom menu actions in the context menus. This requires NO coding knowledge to enable them in Flame. A few versions ago, the Flame products were given python scripting capabilities to automate a variety of tasks. This can help with production pipelines as well as various functionality within the Flame products. These tools have grown in popularity hence the new ability to easily add custom menu actions. For those of you creating python code, you can easily share them with other Flame Artists. And for those of you who are not Python savvy, you can use these scripts simply by copying them into a shared directory. The ideology is very similar to the way the Flame community shares Matchbox shaders and hopefully this will also apply to Python scripts. We’ll cover all you need to know in this video!

In this video, you’ll learn about using the expanded user interface with the MasterGrade. This will give you more controls up front and less digging down into multiple menus. We’ll cover some of interface options as well as the new layout of the MasterGrade when it is fully expanded. Please note that the expanded user interface is also available for other tools within the image toolset, Gmask Tracer and the Action 3D compositor. These can be either in the editorial or compositing contexts.

The Selective toolset in the Image tools and Action is very powerful, allowing you to create a matte or alpha based on a key, gmask and 3D data. This matte or alpha is normally confined to the Selective due to the image processing pipeline. But in this video, you'll learn about a new SelectiveFX shader known as "Selective Matte Extraction". This allows you to push the matte or alpha from the Selective to the main output result of Action or the Image node. You can take this output and use it further upstream in your flow graphs.

This video introduces the 3D selective which uses 3D information to isolate a part of your image. This works alongside the 2D selectives such as keyword and masks when performing any grading, compositing and final finishing. So you can take 3D information such as a Z-depth data pass and a Normals data pass to break up your shots. This is available in the Timeline Effects Environment, BatchFX and Batch. This video will look at the general 3D Selective workflow as well as how to apply a Z-depth data pass to the Selective. The next video will cover the usage of a Normals data pass with the 3D Selectives.

In part 6 of the Image Toolset series, you’ll carry on exploring the 3D Selectives. But this time, you’ll use a normals pass to create the isolation Selectives and these isolations are based on the X, Y and Z coordinates of the surface’s normals in 3D space. And finally, to conclude this video, you will go over how to perform 3D relighting using a combination of the Z-depth map, normals map and the Action 3D lights.

With the release of Flame 2020, a whole range of new technology has been integrated into the Flame products. This is in the form of machine learning which uses trained artificial intelligence to analyse an image and compute a variety data that you could use for VFX and look development. So you could use this data to isolate a part of your image when grading through a 3D selective or further enhance a 2D or 3D composite when working in Action. There are quite a variety of use cases and this is where you start learning about it. In part 1, you’ll be introduced to the Z-depth Generator when working in a sequence within the Effects Environment. This machine learning is also available in Batch and BatchFX.

In part 2 of the Machine Learning Series, you'll continue looking at Machine Learning in Batch as well as another generator which can detect the human face and then extract a normals map. This can be used to isolate and enhance facial features whether you’re performing look development using 3D Selectives or even 3D lighting in the Action 3D compositor.

In part 3, we continue with the “Face Maps” generator to extract a normals and depth map of the human face. You learn how to deal with shots other than close ups using constraints and the region of interest. Finally, you learn how to uses these generated maps with 3D lights in the Action 3D compositor to relight the human face.

In this video, you’ll learn how to apply refraction to 3D Objects in Action. The concept of refraction refers to the way that light distorts as it passes through an object that has transparent or semi-transparent properties. This includes objects such as glass, plastic, ice, diamonds and the list is goes on. Each material has its own refractive properties and you can now emulate this in the Flame 2020 products. It’s incredibly fast and works with a range of GPUs and not just the latest and greatest. This is available on the Linux and Mac platforms of the Flame products and it can be used on 3D Models, 3D text, 3D shapes as well as regular Image objects.

In this video you’ll learn how to use the new Cryptomatte node in Flame 2020. The Cryptomatte data pass allows you to automatically generate a variety of mattes even after the CGI had been rendered. It’s not an image pass that is filled with colours like an RGBA matte pass. Instead it contains actual metadata for each object in the CG render. The other big advantage is that the Cryptomatte takes into account transparency, motion blur and depth of field. This makes the matte generation even more practical when it comes to compositing, look development and grading.

In this video, you’ll cover grading using the new Manager view in Flame 2020. The Manager is a much more simplified way of working with some of Flame’s tools. And whether you’re a beginner or an experienced Flame artist, the Manager will allow you to access a whole range of functionality without digging deep into the node flow graph. Other videos cover the Manager in the context of compositing and rotoscoping.