Compositing

Create new images and generally create new planes.

Generators create new planes. They all have the same parameters for specifying the image and sequence information. Generators define the following information:

• Image information

• Sequence information

If a generator has an input, it can either add a plane to the sequence (“inline generation”) or affect an existing plane using a simple operation (“quick composite”).

Planes are generated “inline” when a generator adds a plane to an existing sequence. A generated plane can have its own data type and composition, it is not restricted to the input types.

A generator performing a “quick composite” generates the plane and uses a simple operation to combine the input plane (I) with the generated plane (G). The simple operations are Add, Subtract, Multiply, Minimum (Min), Maximum (Max), and Average. This is useful for creating masks or slates, or adding noise to images.

These techniques have the following advantages:

• The frame range, resolution, interlace options and extend conditions are inherited from the input. This makes matching the resolution and frame range of another sequence much easier; if these are changed in the input, they automatically update.

• The number of nodes needed is reduced; no merge or composite node is needed. This also boosts performance and memory efficiency.

Inline Generation or Quick Composite mode overrides most of the operator’s Image and Sequence parameters. Use the “Add Plane” parameter on the Image tab to specify the plane to create or modify, and the operations to perform. If the plane specified does not exist, or the operation is set to Rename or Replace, the generator creates a new plane.

Filter operations which may be applied to only specified planes, channels or frames. Certain operations cannot be scoped, such as a Scale or Crop COP (which changes the resolution).

This family of operators allows an operation to be applied to selected planes and frames. By default, all planes and frames are selected.

Planes and channels that are not scoped are passed through as if the node was bypassed.

You specify scoping with a space-separated list of planes. You can specify specific channels with a dot followed by the channel name. YOu can also use wildcards. For example:

Scoped filters also have options for scoping frames:

Affect portions of an image through the use of a mask. This mask is another image which is similar to an alpha matte. Masked filters are a type of scoped filter.

Many operators let you apply them per-pixel using a mask image. The mask can be any plane or channel within the input sequence, or any plane or channel in the sequence of the mask input.

The mask acts as a stencil - wherever the mask is 1 (white), the operation is performed. Wherever the mask is 0, the input pixel is not affected. In between, the input pixel is linearly blended with the output pixel. The mask can be inverted so that all pixels outside the mask are affected instead.

Use the Alpha plane as a mask as a very efficient way of masking color correction operations to the actual image without affecting outside black levels.

Highly optimized operations that can be combined into a single operation if they are grouped together in a chain. Pixel filters are a type of masked filter.

Houdini “collapses” multiple consecutive pixel filters in a network chain into a single cooking operation.

Pixel filter node tiles have a light blue background to distinguish them from other types of nodes.

• Reduces quantization and cutoff errors by doing quantization once, at the end of the pixel filter chain, rather than at every node.

• Uses less memory. Only 1 image is cooked, rather than making a copy for each pixel filter.

• Calculates in floating point, preventing clipping and quantization error of fixed formats (8- and 16-bit integer). Can handle pixel values well above white and below black during the pixel chain without being clamped, even if the image’s native data format would clamp it. The values are clamped at the end of the chain.

• Reduces network and quantization overhead, resulting in faster cooking.

If you use a mask, it must be the same for all pixel filters in that chain. Pixel filters with different masks will not collapse together. Houdini will log a warning to notify you that quantization is occurring at that node (to get rid of this warning, set the Quantize parameter to “At This Node”).

If a pixel filter chain branches into two pixel filter outputs, each output pixel filter will cook the chain independently of the other.

At any point in the chain, you can tell a pixel filter node to quantize using the Quantize parameter. This will terminate the chain at that point and start collapsing COPs again at the next Pixel Filter COP. Setting the display flag on a pixel filter COP has the same effect.

Move frames to different frame numbers, modify the frame range, or shift the sequence in time. Timing modifiers don’t generally cook image information, and so are very efficient in both memory and cooking performance.

Timing Modifiers perform a variety of editing operations and allow you to cut, splice, shift and sequence frames easily.

The following are timing modifier COPs:

• Trim

• Extend

• Shuffle

• Time Warp

• Shift

• Reverse

• Time Scale

Most Timing Modifiers do not need to cook images, so they are very quick and memory efficient. Some Timing COPs allow you to blend frames together; in this case, new images are produced and this efficiency advantage is lost.

Timing modifier node tiles have a beige background when they are passing through data without cooking.

Timing Modifiers are best viewed in the Timeline view mode . Most timing COPs have handles in the timeline view.

Take two image inputs and composite or blend them together in different ways. All compositing operators can be masked.

Compositing operators are the heart of the compositor. They take two or more image inputs and combine them in different ways in the output.

Compositing operators have parameters that let you choose the behavior of the operator in cases where the inputs have different frame ranges, resolution, and/or planes. You can transform the foreground elements as part of the operation, and apply motion blur to animated transforms.

See how to blend and layer images .

All compositing operations can be masked using the mask input.

Create, swap, or merge planes.

You will sometimes need to swap or create new planes to store information during compositing.

See working with planes.

Some COPs are implemented in VEX .

You can create custom compositing operators using the VEX language or VOPs. Some of the compositing operators shipped with Houdini may also be defined this way.

See how to create a custom COP with VOPs.