Matte Assist ML
Generates evolving mattes over time using machine learning-based object segmentation and propagation. You can start with any combination of inputs: Mask ML, Roto, EZ Mask, or Input Alpha--whether from an alpha-embedded source or a node that generates alpha.
Note: The Create Mask ML, EZ Mask, and Roto Node buttons automatically insert a Mask ML, EZ Mask, or Roto node before Matte Assist ML, with their Data output connected to the Data input. For Matte Assist ML tutorials, you’ll perform these steps manually so you can get familiar with the correct settings and workflow.
Mask ML
1 Add a Mask ML node from the Nodes > Key tab and connect it to the source node.
The Add tool is selected by default.
2 Click within the object where you want to create a mask.
With each click, a green point appears, a mask is generated, an alpha overlay appears and an object is automatically created in the Object List.
Note: Auto DOD is automatically enabled which crops the source before matting so the model works only on the active image region, improving matte accuracy.
3 Add additional points as needed.
4 If you see areas of alpha outside of the selected object, right-click or select the Subtract tool and click on those areas.
To mask an object without using the Add and Subtract tools, use the Object tool.
5 Select the Object tool and drag out a box around the object you want to mask.
6 Connect the Mask ML output into the Matte Assist ML input.
7 In the Matte Assist ML node, make sure the Mask Source is set to Input Alpha.
8 Click Auto DOD to analyze the sequence which inserts a DOD (Domain of Definition) node after the source and automatically crops the image prior to matting.
This tighter region produces a more accurate matte by limiting processing to the active image area.
9 Play the sequence to generate the matte.
If the matte is not accurate enough, create additional Mask ML keyframes.
10 Select the Mask ML node and create keyframes at frames where the matte was not accurate enough.
11 To take advantage of the keyframes, connect the Mask ML Data output into the Matte Assist ML Data input.
12 Set the Matte Assist ML > Mask Source to Cryptomatte.
The Cryptomatte option not only transfers keyframe data, but treats each Mask ML object as an independent entity, tracking them separately.
13 Play the sequence again to generate the revised matte using the additional keyframes.
Matte Assist ML generates a binary matte. To create a more natural edge with partial opacity, you can use Matte Assist ML in conjunction with Matte Refine ML.
14 Add a Matte Refine ML node after Matte Assist ML and connect the data ports together to retain separate objects as Cryptomatte data.
Matte Refine ML produces fine hair detail with more natural edges.
15 If you see frame-to-frame jitter, add a Matte Deflicker ML node after Matte Refine ML and connect the data ports so individual objects remain separated as Cryptomatte data.
Matte Deflicker ML evaluates consecutive frames for luminance, edge, and density fluctuations, then applies a learned temporal smoothing model that preserves fine detail while eliminating jitter across the sequence.
16 To improve matte quality at the cost of longer render times, use the Matte Assist ML (Better) model and set Matte Refine ML > Quality and Matte Deflicker ML > Quality to Enhanced.
Roto
1 Add a Roto node, connect it to the source, and create a shape around the object where you want to generate a matte. To start with, there only needs to be one keyframe.
Note: Each top level item (shape or layer) in the Object List is considered a separate object. For best performance when using multiple shapes for one object, place the shapes within a layer.
2 Add a Matte Assist ML node from the Nodes > Key tab and connect the Roto output into the Matte Assist ML input.
3 In the Matte Assist ML node, make sure the Mask Source is set to Input Alpha.
4 Click Auto DOD to analyze the sequence which inserts a DOD (Domain of Definition) node after the source and automatically crops the image prior to matting.
This tighter region produces a more accurate matte by limiting processing to the active image area.
5 Play the sequence to generate the matte.
6 If the matte is not accurate enough, select the Roto node and go to the frames where it is not correct and create more shape keyframes.
7 To take advantage of the keyframes, connect the Roto Data output into the Matte Assist ML Data input, and set the Mask Source to Roto.
Setting the Mask Source to Roto generates a matte throughout the sequence using Roto > Object (Shapes, Layers, Trackers) keyframes when connected to the Data input port.
8 Play the sequence again to generate the revised matte.
By default, the Matte Propagation > Auto mode starts matte propagation with the first keyframe. You may get a better result when using Custom mode and defining a particular frame.
9 If necessary, change Matte Propagation > Auto to Custom and set the Propagation Start Frame.
Matte Assist ML generates a binary matte. To create a more natural edge with partial opacity, you can use Matte Assist ML in conjunction with Matte Refine ML.
10 Add a Matte Refine ML node after Matte Assist ML and connect the data ports together to retain separate objects as Cryptomatte data.
Matte Refine ML produces fine hair detail with more natural edges.
11 If you see frame-to-frame jitter, add a Matte Deflicker ML node after Matte Refine ML and connect the data ports so individual objects remain separated as Cryptomatte data.
Matte Deflicker ML evaluates consecutive frames for luminance, edge, and density fluctuations, then applies a learned temporal smoothing model that preserves fine detail while eliminating jitter across the sequence.
12 To improve matte quality at the cost of longer render times, use the Matte Assist ML (Better) model and set Matte Refine ML > Quality and Matte Deflicker ML > Quality to Enhanced.
EZ Mask
1 Connect an image into an EZ Mask node.
The Foreground brush is selected by default.
2 Go to the first frame of the sequence and with the left mouse button, click and drag green foreground strokes inside the area where you want to create a matte.
3 Next, select the Background brush and draw a stroke around both the outer edge of the object and the rest of the background. You can also draw background strokes by right-clicking and dragging.
As you finish drawing background strokes, a matte is generated.
4 Press the A key to view the alpha overlaid on the image.
5 Add additional foreground and background strokes as needed to refine the matte.
6 Connect the EZ Mask output into the Matte Assist ML input.
7 In the Matte Assist ML node, make sure the Mask Source is set to Input Alpha.
8 Click Auto DOD to analyze the sequence which inserts a DOD (Domain of Definition) node after the source and automatically crops the image prior to matting.
This tighter region produces a more accurate matte by limiting processing to the active image area.
9 Play the sequence to generate the matte.
If the matte is not accurate enough, create additional EZ Mask keyframes.
10 Select the EZ Mask node and create EZ Mask keyframes at frames where the matte was not accurate enough.
11 To take advantage of the keyframes, connect the EZ Mask Data output into the Matte Assist ML Data input.
12 Set the Matte Assist ML > Reference Frames to Keyframes.
13 Play the sequence again to generate the revised matte using the additional keyframes.
Matte Assist ML generates a binary matte. To create a more natural edge with partial opacity, you can use Matte Assist ML in conjunction with Matte Refine ML.
14 Add a Matte Refine ML node after Matte Assist ML.
Matte Refine ML produces fine hair detail with more natural edges.
15 If you see frame-to-frame jitter, add a Matte Deflicker ML node after Matte Refine ML and connect the data ports so individual objects remain separated as Cryptomatte data.
Matte Deflicker ML evaluates consecutive frames for luminance, edge, and density fluctuations, then applies a learned temporal smoothing model that preserves fine detail while eliminating jitter across the sequence.
16 To improve matte quality at the cost of longer render times, use the Matte Assist ML (Better) model and set Matte Refine ML > Quality and Matte Deflicker ML > Quality to Enhanced.
Input Alpha
Generates a matte throughout the sequence using single or multiple input alpha reference frames. Input Alpha is meant to be used for matting a single object. For matting multiple objects, use the Matte Assist ML > Mask Source > Roto option.
1 Connect an image with alpha or a node that generates alpha into a Matte Assist ML node.
2 Make sure the Mask Source is set to Input Alpha.
3 Set the Reference Frames to Frame and enter the frame you want to propagate the matte from in the Frame field.
4 Click Auto DOD to analyze the sequence which inserts a DOD (Domain of Definition) node after the source and automatically crops the image prior to matting.
This tighter region produces a more accurate matte by limiting processing to the active image area.
5 Play the sequence to generate the matte.
If the matte is not accurate enough, set Reference Frames to Markers and create markers at frames where the matte was not accurate enough.
6 To create a marker, click the Marker > + icon in the Node > Options tab.
Matte Assist ML will now use the marker frames as keyframes while generating the matte.
7 Play the sequence again to generate the revised matte.
Matte Assist ML generates a binary matte. To create a more natural edge with partial opacity, you can use Matte Assist ML in conjunction with Matte Refine ML.
8 Add a Matte Refine ML node after Matte Assist ML.
Matte Refine ML produces fine hair detail with more natural edges.
9 If you see frame-to-frame jitter, add a Matte Deflicker ML node after Matte Refine ML and connect the data ports so individual objects remain separated as Cryptomatte data.
Matte Deflicker ML evaluates consecutive frames for luminance, edge, and density fluctuations, then applies a learned temporal smoothing model that preserves fine detail while eliminating jitter across the sequence.
10 To improve matte quality at the cost of longer render times, use the Matte Assist ML (Better) model and set Matte Refine ML > Quality and Matte Deflicker ML > Quality to Enhanced.