[Bf-committers] gsoc and Durian artists wishlists

Tom M letterrip at gmail.com
Wed Mar 30 02:05:15 CEST 2011

Hi all,

I emailed all of the main durian artists about their wishlists or
items that would have made their life easier - both in their core area
for Durian, but also in other tools they use in Blender.  Here are the
responses I've received back so far.  Ton had asked if we could have
items of gsoc that are a bunch of smaller items.  My thoughts are we
could place most of these as pick 15 or so of these to work on for a
gsoc project.

Anywho here are the items, some are stuff that needs to be brought
back from 2.49b and earlier.  (Note that I included Ben Simonds by
accident when i was sending out emails, but I figured he is similar
caliber so it doesn't hurt anything to have his wishlist in here

It might be worthwhile to contact the top 100 or so blender artists
and get similar lists from them.  I know we have the blenderstorm and
such, but I think feedback from hardcore users is a bit more valuable
in some respects.


Dolf Veenvliet

vector textures


Paint Texturing ( Brush dynamics : as random brush bitmap mask
rotation , best painting performances )
Video Editing ( adding text directly in VSE, or using *.svg template )

Beorn Leonard

- a waveform in the timeline. This would make lip sync animation
sooooo much easier. 3D Max has it, Maya has it. This is a really
important feature for us spline monkeys.

- Action groups. Allow putting separate armatures and objects into the
same action by adding them to an action group.  (I know this is
contentious, just putting it out there again)

- The ability to select by keying set. Makes it easier to quick select
and push/relax a pose in layered animation.

- "View by keying sets" in the dopesheet. The ability to group
channels in the dope sheet by their keying sets, with summaries
displayed for each. Duplication of channels across keying sets doesn't
matter. Channels that aren't in keying sets can sit below in an
"Ungrouped" group. This would keep the dope sheet cleaner when you're
doing layered animation.

- No ungrouped channels in the dope sheet and curve editor. I don't
quite know how this happens, but sometimes you end up with a bunch of
channels down the bottom of the dope sheet that are unattached to
their object. This happens not only with custom sliders, but also for
basic things like location and rotation. It gets messy and can slow
you down.

- the ability to bake NLA sequences to a single action for final
polish. Campbell added this feature for me during Durian
(spacebar>Bake Action), but it doesn't work properly when bones have
animation in more than one action strip.
- the ability to see the final result of the NLA stack while editing
an individual action. The lack of the ability to polish is why I very
rarely use the NLA; it's just about impossible to get professional
quality polished animation out of it.

- the warp modifier. This was added by Campbell for Durian, but
doesn't appear to be in trunk. I used it to animate Sintel's blanket
in the cabin scene. I know you can achieve the same results using
other methods, but I found it really quick and intuitive to use.

- the ability to do GL preview renders from the sequencer. Among other
things, this would complete the toolset for using the VSE for previz
of multi camera scenes. IMHO, much nicer than the camera switching
with markers method.

- the ability to render reduced quality RAM previews from the
sequencer and compositor. For quick tests, just like in After Effects.
The maximum length of the preview is determined by available RAM.
of course, this will only be useful for the compositor if we can
- finish and integrate Joeren's OpenCL compositor code.

...and my final beef...

Blender is still unable to output a video file with sound that can be
played in anything other than vlc. This is mainly an issue when
rendering GL previews of animation as most of us always do final
renders as image sequences, but doing quick GL previews is a standard
part of the animation workflow.

At the moment there are far too many options for codec/wrapper combos
and most of them don't even work in vlc, let alone other players. BTW,
vlc has the same problem - too many illegal options. Since Blender is
not really intended to be an encoding program I would suggest
restricting the options to a few that actually work and can be played
in vlc, quicktime and windows media player. If users want, we can use
a dedicated encoding program after rendering (e.g. avidemux, quicktime
pro, autodesk cleaner) to convert to whatever exotic format we want.
This has long been a noob confuser, and I reckon it's better to do a
couple of things well rather than a bunch of things badly.

Jeremy Davidson

Clear user transforms - a great little button in 2.49 that did a clear
locrotscale on every bone in the rig (even hidden ones). would love to
have that back.

Less armature dependency loops - i work around this stuff alot but the
two main ones that would be great to have are; seperating loc rot
scale channels (possibly too hard to do, too much coding) and spline
IK. Been banging my head against a brick wall with that one the last
week or so. trying to get around it with multi armatures and layers
and layers of bones copying all different channels but i can never get
it quite close enough.

Jonathan Williamson

Mesh modeling/Retopology: pen tool to quickly draw polygons without
the need to fill faces manually. An example can be seen in Modo's Pen
tool here: http://www.luxology.com/common/qt/LuxQT.aspx?u=http://video.luxology.com/modo/301/video/PenTool.mov&w=1280&h=780
and another in Topogun via the SimpleCreate tool:

Mesh modeling: Bevel tool:
Mesh modeling: Edge bevel tool with sharp corners to preserve
topology: http://www.luxology.com/common/qt/LuxQT.aspx?u=http://video.luxology.com/modo/301/video/EdgeBevelSharpCorner.mov&w=1280&h=780

Mesh modeling: Native Bridge tool: Crouch has implemented this through
his Loop Tools add-on here:
or Modo's tool can be seen here:

Mesh modeling: Better Rip tool. Namely need to be able to rip multiple
selections at a time and more consistent behavior when ripping. An
example of multiple selection ripping can be seen here:
You can almost think of it as a combination of the Rip and Split

Mesh modeling: Better Loop Cut tool that allows cutting based on
selection, even cutting edges that are perpendicular to each other:

Retopology: Slice tool for quickly adding loops:

Retopology: Paint Stroke tool for adding faces based on intersecting
strokes: http://www.3d-coat.com/uploads/pics/rtp_strokes.gif

Interface: Proper Units system input. Currently when using the
imperial system all input is converted in normal Blender Units. For
example, if I switch to Imperial units and then translate my selection
along the X-axis by pressing X > 1 to move it one unit, it should move
1 yd (or ft or inch) but instead it moves 1.0936 yds. This makes the
Imperial system in Blender next to worthless. The Metric system works
no problem since the number of minor units per major units is the

Interface: Prevent unit conversion during input. Currently if you
switch to Imperial and input 3 ft it will display as 1 yd. Normally
this is not an issue but many times you'd like to always display units
as inches or feet. This also applies to Metric.

Ben Simonds

Bevel tool in edit mode. I gather this will come back with bmesh, but
I use it all the time, and at the moment that means I do most of my
modelling in 2.49 before importing to 2.5.

Object ID Masks: Are terrifically useful, but it is a pain going
through all the objects in a scene and assigning an ID value to them.
It would be really nice if there were a way to automatically assign
object ID's, for example one per object, per group, or per layer. Also
if ID's could be assigned to materials within an object, that would be
brilliant. It would make composting much faster if that kind of stuff
could be done automatically, with the user only having to go back if
there was a specific object they wanted to set the ID for.

On the same note, if user could assign a range (1 to 5) or a list (1,
2, 5) in the ID mask node, that would be great. I know it could be
done with multiple nodes, but it would keep the node layout much
Render improvements: I'd love to see the tile cache for raytraced
AO/Bounce/Environment light from the render branch integrated into
trunk. Currently though it produces very artifact-ey reults (at least
last time I checked) unless samples are set very high. Also enabling
more/smaller tiles than currently possible would be good to help with
Any other improvements to blenders raytracing/GI would be fantastic.

Compositing: The Depth Blur node currently give incorrect objects for
parts of the scene in front of the focal plane. Within the outline of
the object they blur correctly, but the object still has a sharp
outline, where it should be blurred.
Layers in texture paint mode, maybe with support for .psd .xcf or .ora
files or some other format that supports layers. This would be great
for more non-linear, non-destructive texturing.


Tooltips for tablet users - when using a drawing tablet, it is
difficult to keep the cursor perfectly still for any length of time,
in order to bring up the tooltip for a button. In zbrush this is
solved quite elegantly by having the ctrl key bring up a tooltip when
the user moves the cursor over a button. Something similar in blender
would be great.
A Surface-Based mesh deform modifier. I'm not sure if this one is
possible, but it would be brilliant. The idea would be to use the
surface of one mesh to deform another. This could be used for example
to rig a high-density sculpted mesh, using a low density mesh
retopologised over it's surface, or to deform an object  in a similar
way to using a planar lattice modifier (only one point thick), but
with arbitrary topolgy.


1. HUV sampling
When you look at a render result in the uv/image editor, it is
possible to sample a pixel with your mouse, showing RGB and other
stuff. It would be great to additionally see HUV and being able to
choose a sampling area. Within that area blender would average the
different color channels and shows the average result and not the
result of a single pixel. Kinda like the pipette within photoshop.

2. A lens flare system for lamps.
The current lens flare system is very inflexibel. Besides being not
for lamps, it's not even possible to control the color or size of a
single flare. There are methods, but they always eat time.

3. programable shaders

4. Texturable specular highlights
Being able to asign textures to specular highlights and I don't mean
specularity maps. At the moment specular highlights are circles and
instead of that having a texture would enable to make it for example
square (like a reflection of a window).
The perfect solution would of course be to only use raytraced, gloss
reflections but that's insanely slow and a bit noisy.

5. Split the material option 'traceable' up into: cast ao, mirrorable
etc. This would be mostly useful for optimizing render performance.
For example, sometimes you want an object to cast raytraced shadows
but not getting reflected by mirrors

6. First and second derivative for f-curve editor
Being able to see speed and acceleration of a f-curve is something that I miss..

More information about the Bf-committers mailing list