Just released JWildfire V2.56.

This version comes with the following major new features:

• improved render quality: spatial oversampling and a few of other option to significantly improve render-quality (see the official Wiki for details: http://wiki.jwildfire.org/index.php?title=Manual_RenderQualityFlameFractals)
• visualisation of filter-kernels to make it more transparent how filtering works and how different settings work
• reduced memory-usage (needs about 50% of memory with identical settings as before)
• simplified install: no more installer, just unzip, additional package for Windows 64 Bit with pre-installed Java and Leap-motion-dll’s

Complete list of changes:

•  new spatial-oversampling-option for improved render-quality. The default value is 2, this may be set in the Preferences
• visualisation of filter-kernels for better understanding/easier customizing of the spatial filtering process, both a nice-looking 3d-mode and a fast 2d-mode are available
• new sample-jittering- and color-oversampling-option for improved render-quality. The default value is off, this may be set in the Preferences.
• new post-noise-filtering-option for redusing noise of the final noise by using an adaptive gaussian filter. The default value is off, this may be set in the Preferences.
• added a brief description of all of this options in the official Wiki: http://wiki.jwildfire.org/index.php?title=Manual_RenderQualityFlameFractals
• changed the memory management/layout of the iteration buffer (rendering should now take significantly less memory, and initialisation of rendering runs substantly faster, comes to a cost of only slightly degraded render time)
• the final image-generating-process now works fully multi-threaded and should be faster
• added a new experimental option “Opacity” to control the opacity of the foreground when rendering an image with background (often those renders caused the loss of too many details)
• new “Pseudo3D”-random-flame-generator
• improved the “Xenomorph”-random-flame-generator
•  improved the “Cross”-random-flame-generator
• some more useful statistics in the interactive renderer, e.g. absolute render-speed-indicator to make it easier to compare different settings (render speed means the quality level the renderer can reach in one hour)
• the windows-launcher (JWildfire.exe) should now be much smarter to find a valid Java-install
• completely removed the izpack-installer (until now there were always two files: a *.zip for users who just wanted to unzip, or an izpack-based installer which caused too many problems)
• there are now two packages available: platform-independant without jdk and any dll’s, and windows64 with jdk and certain dll’s
  – removed the applet from the distribution
• the Chaotica-integration is now disabled by default (in order to hide the export-button which may occupy important space for some users)
• changed the animation-playback in the main-editor to full-window-size
• made the detached renderer and the interactive renderer to behave more smooth
• progressbar-update should be much smoother now, i.e. use less ressources
• negative “perspective”-values are allowed now (again)
• when using post-symmetry, the maximum of symmetry-order of point-symmetry was increased from 18 to 36
• new ouputchannel to change param-values of plugins via the Leap Motion controller
• added a new option tinaIntegrationChaoticaAnimationExport to the preferences. This option controls if the flames which are exchanged over the Chaotica-bridge are animated flames or static flames. In case you export them as static flames and have not a Studio licence for Chaotica you may render in higher resolutions.
• fixed a bug in the crop-like variations
• fixed a bug regarding autoimatically saving the current flame in the iteractive renderer
• some refactoring of the structure synthesizer
• completely removed all experimental opencl-code

Just click the “Random flames” button to start playing around and have fun 🙂

Released the BETA 2 which adresses several performance issues:

• default rendering is faster now
• rendering in the interactive renderer/detached view is substantly faster
• render-display-update in the interactive renderer/detached view is changed to non-buffered by default (you may change this in the Preferences, property tinaOptimizedRenderingIR)
• faster and smoother update of the progress-bar while rendering

Download

Hi all,
in the last weeks I was working on some major changes which I should have done much earlier, but never found the time before: increasing the render-quality! The new version actually works quiet good, but there were so many changes under the hood that I decided to release a BETA-version to receive your feedback before the final release.

 

So what has changed?

The previously used cheap antialiasing-mode was completely removed.

To increase image quality and reduce noise there are now the following options (at the Antialiasing/Filter-tab):

• spatial oversampling
• apply one of 14 supplied filtering kernels (depending on the fractal, to increase sharpness or reduce noise)
• in order to help you to choose the right filter, there is a visualisation of the filter-kernel, either in flat or 3d-mode. As a rule of thumb: sharpening filters have negative parts (indicated red in 2d-mode)
• color oversampling
• sample jittering
• post noise reduction

I will describe those options later at the official Wiki, just have fun to play around with them.

 

You will notice that the renderer appears to be slower. But, it isn’t, in fact it is faster. It just renders at spatial-oversampling-level of 2 per default (i. e. internally rendering an image which is larger by factor 4 in size).

So you should now get better quality in nearly the same time, maybe in some more time. But, more importantly, there is no need to render at insane quality levels to render an image in decent quality.

 

You may change all the new default settings in the Preferences.

 

Any feedback is welcome

 

Download