Created attachment 50005 [details] comparison of fading at 0% in different apps Version: svn trunk (using KDE 4.4.5) OS: Linux The softness of generated brush tips looks wrong when compared to brush with similar settings in photoshop and gimp. Krita's brush mask seems non-linear and too opaque when the hardness is set to 0%. I think gimp's fading is the one that is the best as it has a bigger range of softness. At 50% of fading, it looks like photoshop's 0% fading. That could be useful for importing ABR brushes. Reproducible: Always
Can you check whether this is still the case? We had lots of changes there.
I think we still need to implement besides WASH/BUILD UP new GIMP-compatibile mode. I remember that also Deevad requested this. Nothing was changed in this area. Also we need to implement PS's flow, but that is probably different bug report.
We have a different bug for flow already.
*** This bug has been marked as a duplicate of bug 281364 ***
I've been working on a new pair of autobrushes that use Gaussian fading functions. They are somewhat slower than the default autobrushes (about 25% to 70% the speed for valueAt) but are much softer and nicer looking, in my opinion. The patch adds a third type of autobrush after "default" and "soft brush" so it is fairly noninvasive. I'll push it when I get home in a few hours.
do you use a precalculated lookup table for the values of the gaussian curve or a formula?
Git commit ce6ff0b1c61a5929aea5b545b32c15df96ce45c4 by Geoffry Song. Committed on 16/10/2011 at 22:32. Pushed by geoffrysong into branch 'master'. New autobrush type: Gaussian The Gaussian autobrushes use a Gaussian blur for their fade, rather than the linear function used by the default autobrushes. The result is that the brushes look softer and have less of an edge, as well as having fewer aliasing issues. CCBUG:247363 M +2 -0 krita/image/CMakeLists.txt M +17 -7 krita/image/kis_base_mask_generator.cpp M +1 -0 krita/image/kis_base_mask_generator.h A +84 -0 krita/image/kis_gauss_circle_mask_generator.cpp [License: GPL (v2+)] A +56 -0 krita/image/kis_gauss_circle_mask_generator.h [License: GPL (v2+)] A +83 -0 krita/image/kis_gauss_rect_mask_generator.cpp [License: GPL (v2+)] A +56 -0 krita/image/kis_gauss_rect_mask_generator.h [License: GPL (v2+)] M +3 -0 krita/image/kis_mask_generator.h M +2 -2 krita/plugins/paintops/libbrush/kis_auto_brush.cpp M +22 -6 krita/plugins/paintops/libpaintop/kis_auto_brush_widget.cpp M +1 -0 krita/plugins/paintops/libpaintop/kis_auto_brush_widget.h http://commits.kde.org/calligra/ce6ff0b1c61a5929aea5b545b32c15df96ce45c4
a formula, with a few expensive math calls. Using a lookup table would probably improve performance.
Created attachment 64655 [details] comparison with new gaussian mask I think the new gaussian mask is good enough to consider this bug fixed.
I agree, we can close the bug. Two points though: using a lookup table for performance would be very nice, as Goffrie notes. And as Lukas said, it might be worth it to add a new mode that's compatible with Gimp's VBR brushes, which also uses gaussian masks, but in a different way (as shown by your screenshot). But we shouldn't replace the new gaussian mode by the gimp mode, since this code looks very compatible to photoshop. All in all, great work. Goffrie, you have the honor of closing the bug :-)
Okay :)