Hardware acceleration for all compositing operations.
You can interactively zoom and pan the view, also while you're editing
or transforming selected layers, in order to see how the changes affect
different details of your work.
IRIX-native, with Silicon Graphics User Interface look and feel. Icons
for the IRIX Interactive Desktop, with drag-and-drop support, Alt+doubleclick
support for entering command line arguments, etc. And, of course, minimized
Unlimited Undo/Redo for all actions (except viewing actions, since they
don't modify actual data).
Optional "Low Resolution mode", for avoiding TRAM swapping in systems with
limited texture memory.
Optional "texture LOD optimization", for avoiding TRAM swapping in systems with
limited texture memory.
"Tile Culling" optimization, for faster frame rate when some textures fall
outside the window area, so that these textures don't be made current if
Online Manual (the complete documentation is online, in the 'Help' menu).
Insert any number of image files in any format supported by the SGI IFL
for RGBA, RGB, grayscale-alpha, and grayscale color models.
Both 8-bit and 16-bit per component images are supported (although the
internal image precision will be transparently converted to the bits supported
by your hardware, which may be 8bit per component or 12bit per component,
depending on your machine).
Work with very large images. Image size is not limited by the application
(although your hardware may impose a limitation, since all images are managed
as tiled textures, and texture memory is usually a limited resource except
in UMA systems).
Create any number of Bézier layers. They can be blended and processed
just like any other layer.
Interactive vertex handles in Bézier layers.
Bézier layers support variable-width fading (so the external fading
doesn't need to have constant width).
The external fading in Bézier layers has an adjustable fade exponent.
Insert any number of TexFont files for creating single-line text layers.
They can be blended and processed just like any other layer.
Create any number of 2D CAD layers by inserting DXF files (in release 12
format), either ASCII or binary. Supported entities are: Lines, polylines,
arcs, circles, 3Dfaces, meshes, and blocks. The Z coordinate of all entities
is ignored, since the application is targetted at image compositing and
lacks the notion of a "3D camera", which would be necessary if 3D objects
had to be inserted.
Customize line patterns and line width for your CAD layers.
Optimized DXF rendering (lines, polylines and circles are automatically
grouped into line strips and loops if possible, and 3Dfaces and meshes
are grouped into triangle strips if possible).
Real time antialiasing for DXF line-based entities.
As an option, DXF layers can be automatically converted in contours, and
"Tiler" tool: repeat any layer (image, TexFont or DXF) into an NxN array
of contiguous tiles, with optional mirroring.
Reorder the layers stack.
Create layers that share the same image file, TexFont file, or DXF file,
in order to save memory and loading time. The data will be loaded and stored
just once for all layers that share it.
Reload all inserted files at any time, and the composition will be updated
to the new contents of the files (for example, you can modify an image
with an external painting application, and then update your composition).
Select layers with either direct mouse click or with the "Layer Stack".
Two selection modes: Independent and Multiple.
Define "Selection Groups" for convenient selection of multiple layers.
Use a pivot point (with sub-pixel coordinates) for rotation, scaling, and
positioning of either images or DXFs.
Distortion transformation (independently drag each image corner). The maths
used are those of quad to quad perspective transformation. This tool is
available for images only (not DXFs, nor Bézier, nor TexFonts).
Snapping to vertices, lines, and midpoints of any layer (including all the
supported entities in DXF files). Snapping lets you accurately place
layers with the mouse.
Work without modifying image data: No matter how many transformations you
apply, the images are always stored with their original file data, and
transformations are performed "on the fly" using texture hardware. You
can make any number of tests without loss of the original pixels.
Use the pivot point to align layers positions.
Layers can be transformed/placed with either the mouse (optionally with
snapping), or also with numerical keyboard control for greater accuracy.
Sub-pixel accuracy for all transformations.
Several blending modes for each layer: Mix, Add, Sub, Rsub, Min, Max, Mult,
AND, OR, XOR, AND Inverted, OR Inverted, Negate.
Per-layer opacity and color modulation.
Turn any layer (image, Bézier, TexFont or DXF) into a matte, in
order to define the opaque and transparent areas of other layers above
the matte layer. Matte layers are invisible (their only visible feature
is the alpha channel when above layers are viewed through it).
Combine several matte layers (using any of the available blending modes)
in order to define complex mattes (for example, you can add or subtract
several mattes, make holes in a matte, etc... ). Note that matte layers
keep their independence from each other, so you can transform each matte
layer independently, and the combined matte is automatically updated.
Each layer's transparency can be configured so that it's defined by its
own alpha channel, or by matte layers behind.
Per-layer brightness/contrast (hardware accelerated).
Per-layer hue/saturation/lightness/monochrome (hardware accelerated).
Per-layer Convolution filters: Box Blur, Gaussian Blur, Motion Blur,
Emboss, Box Sharpen, Gaussian Sharpen, and Edge Detect (all of
them hardware accelerated).
Per-layer Look-Up Table (LUT) filters: Levels In/Out (luminance or per-channel),
Solarize, Gamma (luminance or per-channel), and Posterization
(all of them hardware accelerated).
All image processing can be masked or not with the current matte, and can
be applied either before or after blending.
Warping Operators: Spherize/Pinch, Whirl, Sine Wave (all of
them hardware accelerated through textured meshes).
Warping Volumetric Operators, for considering refraction through an imaginary
"water surface" over each layer. Supported operators are Spherize Positive/Negative,
and Top Sine Wave (all of them hardware accelerated through textured
Each layer can have any number of warping operators: Their effects are
properly combined together. The "normal" operators are just added (just
a sum of X/Y distortions). The "volumetric" ones add their contribution
to the Z "water coordinate", and then refraction is computed through the
resulting water surface after all volumetric operators are applied.
All warping operators have handles for interactive positioning and adjustments.
ASCII description file
Save (and load/merge) the composition as a (more or less) human-readable
ASCII description file, which defines the files to insert, the transformations
to apply, the layers stack order, blending settings, and every piece of
information needed to define your composition.
Since these ASCII files don't contain actual images, TexFonts, nor
DXF data, they're usually small, and you can save many different compositions
that use the same image, Texfont, and DXF files without wasting disk space.
The composition ASCII files are independent of image size: you can replace
the images with lower (or higher) resolution ones. This is useful if, for
example, you want to email a low resolution version of your composition
to a colleague.
Names of files to be inserted are stored without path when you save an
ASCII file. This allows porting of compositions across different machines
and directories. If the inserted files are not in the same directory as
the ASCII file, you can also save a "paths file".
Render the composition into an image file of any size, with optional supersampling
for antialiasing and subpixel-accurate rendering. This final rendering
also benefits from hardware acceleration.
Manage an arbitrary number of render areas per composition. This lets you define
several "output windows" in the same composition.
Save the rendered image as either 24bit RGB (8+8+8) or 48bit RGB (16+16+16).
NOTE: although you can save with 16bit per component precision, the effective
bits are limited by the graphics hardware, which in the best case will