Twist brush giving crazy results
Closed, ResolvedPublic

Description

The twist brush seems to behave with too much strength... in any case, not usable.

r42337 under Ubuntu 10.10 64 bit

Details

Type
Bug

I suppose you've got issues with twist brush in sculpt mode, right? Can't see anything weird here.
Can you be more specific and describe what exactly going wrong in your case? Also, please always attach sample .blend files which demonstrates issue and do not forget to specify exact steps of reproducing issue.

Hi Sergey, steps are nothing special, simply I tried to rotate a part of the model by the twist brush and discovered it behave that way.
I attached a simplified blend showing the issue (this happened also several time ago and got fixed, the name of the brush was "rotate" then, not sure it was the same code).

Forgot to attach new images to show the blend issue.

The only thing i can see changed in twist brush is that in 2.53 rotation angle used to be scaled down by 4 times, now there's no such scale. Don't think it's actually a bug, but probably having slider to control angle scaling can be useful here.

Made quick patch which adds strength slider for rotation brushes. Are you familiar with applying patches? If not, just say and i'll prepare test build for you to see if it's really useful feature.

I'm not that much familiar, sorry. And thanks.

EnV

Pardon for the delay, here's build with patch applied: http://download.blender.org/ftp/incoming/linux-glibc27-x86_64-twist-brush.tar.bz2

No problem, thanks for the build. :)
The strength is much better now.
Btw, it doesn't seems to work properly to me... I've attached a short avi screen-cast showing the brush behavior: besides the fact that there is a "step" in the deformation where the affected vertices go back to their original position and then restart to move, the brush doesn't seems to rotate the mesh part, but to twist it on a perpendicular ax to the screen (weird); unless there is something I missed about the brush? Maybe it wasn't actually meant to replace the rotate brush but exactly to twist (as the name seems to indicate, after all)?

That flipping is just how angle is calculating -- the whole circle is 360 degrees, it's getting multiplied by strength, so it's less that 360 degrees which leads to "jumping" to original position. Not sure it's easy to solve -- we'll need to switch brush to be more context-related counting full circles around point and so =\
What's indeed confusing is an axis around which rotation happened. Rotation supposed that rotation happens around axis coming from scene origin. IMO, much better results giving rotation around axis coming from point of initial click. Here's updated build: http://download.blender.org/ftp/incoming/linux-glibc27-x86_64-twist-brush-2.tar.bz2
Now about direction of this axis. By default, it's calculating as average normal of affected area, but it can indeed be switched to axis defining by viewplane (see that Area Plane combobox behind the brush icon).

Ah, it works ok now. :) Just a little bit slow, but not really an issue (depending also on a dense mesh, maybe).
The flipping is a little bit annoying, but not a real issue as well, just using more strength it can be bypassed.
Thanks a lot for the patches and the builds.

EnV

Cool to hear, thanks for testing. Will check with others from team if it's fine to change now or better to wait for after release..
Is that slider for strength is needed after recent change for rotation axis or strength=1 can always be used (if'll eliminate that jumping) ?

Well, IMO it can be useful anyway, to add rotation with more or less strength (= speed, in this case). :)

Commited patch to svn rev42555, so now rotation happens around correct axis and rotation can be effected by strength (default strength is 1.0)
Thanks for the report, testing and feedback :) Closing report now.

Sergey Sharybin (sergey) closed this task as "Resolved".Dec 10 2011, 3:46 PM