aQtree

aQtree is a dedicated non-photorealistic environment. It can render 3D
geometry with expressive shaders, or it can use geometry to interpolate (tween)
strokes drawn manually by the artist (but not yet).

vision of enhancing the expansive possibilities of 3D with
haptic, non-photorealistic image generation and artistic control. Our goal is
a hybrid 2D/3D workflow in which artists are supported by a software that more
intelligently interprets their input, interpolating frames with controllable
coherence and accepting corrections intuitively.

[screen grab, interface]

It is a stand-alone application written from the ground up according to the
non-physical needs of stylized rendering. It will support a range of working
methods:

It will eventually be suited for:

There are three main innovations which we aim to combine:

Everything in aQtree is simultaneously 3D and 2D. A manually
drawn stroke is defined in space (free-floating or along a
3D surface), and an imported 3D mesh is at the same time a
flat, rendered layer. As a result, spatial relations such as
proximity values, overlap and incidence angles provide valuable
data with which visuals can be interpolated, while the resulting
layers can be layered or processed against each other to create
convincing “non-real” contours and surfaces.

A beautiful looking image will fail as an animation if its
temporal qualities aren't convincing. A paused element can
'die' if it is too still. A contour can be disruptive if it
pops, slides and jitters. What's more, the viewer will
expect a line that looks like charcoal to move like charcoal,
to smear like charcoal and may reject the illusion of the line's
materiality if it is too clean - ie. if imperfections like
an artist's thumbprint are omitted.

The greatest machine is only as useful as as the operator who
knows how to use it. To access the power and efficiency that
we hope aQtree will offer, artists will need to communicate with
the computer, informing it how to interpret the strokes and color
values as they draw. Only so can they create lushly detailed or
convincingly simple looks and artist-driven stories.


The application is docked upon the Frapper framework developed at the
Filmakademie Baden-Wuerttemberg, which in turn uses the ogre 3D engine.
This open framework supports the integration of additional renderers as well
as proprietary input/output formats.

aQtree is a cooperation between the University of Konstanz, the
Animation Institut at the Filmakademie Baden-Wuerttemberg and brainpets GbR.
The driving members are Thomas Luft, Stefan Habel and David Maas.
Valuable support comes from Volker Helzle and Oliver Deussen. See team.

aQtree will become a community-driven application.

The aQtree project strives to bring graphics research and artists
together in a meaningful exchange. On the one hand, we pursue the
vision of making detailed visual looks feasibly efficient for
animation production. On the other, we seek to better define and
understand the mechanisms within which the human visual system
perceives temporally successive image sequences - stylized animation.
By providing a platform with which artists can pursue previously
unfeasible looks, researchers gain feedback into the perception of
temporal qualities in animation that will in turn drive further
software and graphics research.

We have attempted to structure our efforts accordingly, so that the
application might sustainably engage a passionate community.

To quote a John Maeda haiku:


Theorie ist Wissen, das nicht funktioniert. Praxis ist, wenn alles
funktioniert und man nicht weiß warum.
(Hermann Hesse, 1877-1962)

Why make another NPRenderer?
There are numerous commercial products which are capable of NPR
for still imagery in convincing quality. Choices lessen when the
artist demands acceptable temporal coherence and lessen even
further when a technically accessible interface is required.
Nearly all NPR applications are built within shading languages
based on the physics of light - or their efficient simulation.
aQtree takes the liberty to approach the problem without that
preconception. This brings with it drawbacks, but ultimately allows
the analysis of artistic media to be the sole consideration.

Why isn't aQtree a plug-in?
Developing aQtree as a stand-alone application makes it possible
to pursue a shading model committed to freely exploring a model of
representation based on media and tool (ie: pigment, paper and brush).
This causes a good deal of effort replicating basis features that
pre-exist in other programs, but it also enables access to a larger,
software independent community.

Is making an NPRenderer really that difficult?
Its easy to take the temporal qualities of NPR for granted. Everyone
has seen a painting or a watercolor and can imagine a moving image in
that style. But no such coherent movement exists in that style.
We anticipate a process simlar to the maturing of photorealistic
techniques and the viewing habits that accompanied them.