FluoRender 2.34 Released

This year’s FluoRender release feels a little different to me. Maybe it’s because the past year pushed me to rethink what this project means and who I’m building it for. I applied for NIH funding, hoping it would give FluoRender a new chapter, but the proposal wasn’t funded. Oddly, that disappointment turned into something freeing. Without a grant dictating the roadmap, I could finally focus on the things biologists actually use every day — not the things that look good in a proposal.

Last summer, I sat down and asked myself a simple question: What would make FluoRender feel immediately better for its users? At the time, I was experimenting with holographic displays, trying to make volume data float in mid air. It was magical when it worked — and painfully slow when it didn’t. That frustration nudged me toward something I’d been putting off for years: improving how FluoRender handles meshes.

Some of you have asked for faster mesh generation, especially for 3D printing microscopy data. I always knew how to make it faster, but other obligations kept getting in the way. This time, I finally had the space to do it. I built a new GPU based system for generating meshes. While working on it, I kept thinking about Bill Lorensen, the creator of the marching cubes algorithm and an early supporter of FluoRender. I wish I could have shown him this update.

That one improvement set off a chain reaction. Once meshes were faster, I wanted them to look better too. FluoRender had been using the same shading style since the beginning — reliable, but a bit old fashioned. So I redesigned the lighting to give meshes a more modern, almost crystalline look. Softer shadows, subtle glow, hints of translucency… the kind of visuals that help you see structure rather than fight with it.

And once meshes looked better, I realized volumes deserved the same treatment. So they got it.

Along the way, I ended up rebuilding a big part of FluoRender’s internal machinery — not because I planned to, but because one improvement kept revealing the next. A major rewrite of the framebuffer system made the whole rendering pipeline more predictable and opened the door to something I’m really proud of: a complete overhaul of multichannel volume intermixing. Now, every combination of modes behaves sensibly. Even MIP plays nicely with color maps and depth mixing.

One of the longest requested features also finally made it in: clipping boxes for mesh objects. It sounds simple, but making clipping behave consistently across objects of different sizes turned out to be a puzzle. The solution ended up being surprisingly elegant — clipping boxes can now be shared, synced, or automatically matched to whatever you’re looking at.

Color mapping also got a major cleanup. Many of you have told me that FluoRender’s color modes are powerful but confusing when combined with brush selections or component generation. So I rewrote the system so selected and unselected regions can have their own color modes. You can now paint part of a volume and apply a color map to just that region while keeping the rest unchanged. Even MIP cooperates. It feels a bit like having foreground and background colors in Illustrator, but in 3D space.

I also added two new color mapping modes — radial and linear — inspired by feedback from users whose images ended up on journal covers. These new modes align with the data itself rather than the axes, making gradients feel more natural and more biologically meaningful.

With all the internal refactoring done, I circled back and added a handful of analysis tools that suddenly became easy to implement: mesh volume and surface area measurements, smoothing, simplification, 4D mesh playback, real time mesh generation for brush tools, and more. Brush selection and mesh generation now respect clipping boxes too. My hope is for FluoRender to grow into a segmentation tool that feels intuitive, flexible, and maybe even fun.

And finally — something many of you have asked for — FluoRender now reads DICOM files. I wrote the reader from scratch, in FluoRender’s own style. If you work with CT or microCT data, I hope this makes your life a little easier.

This release is the result of a year of wandering, experimenting, breaking things, fixing them, and rediscovering why I started building FluoRender in the first place. I hope you enjoy exploring these new features as much as I enjoyed creating them.

Mesh & Surface Improvements
• Much faster mesh generation using a new GPU based method.
• Cleaner, more modern mesh appearance with updated lighting and shadows.
• Clipping boxes now work for meshes, with options to sync across objects.
• New mesh analysis tools: volume, surface area, smoothing, simplification, welding.
• 4D mesh generation and playback for time series data.
• Real time mesh creation with paint and grow brushes.
• Color transfer from components to meshes for easier segmentation workflows.
Volume Rendering & Visualization
• Updated lighting for volumes (softer shadows, clearer structure).
• More reliable multichannel intermixing — all combinations now behave sensibly.
• Brush selected and unselected regions can use different color modes.
• New “no display” mode for hiding unselected regions.
• MIP now works with color maps and depth mixing.
Color Mapping Enhancements
• Two new color maps: radial and linear.
• Color maps now align with data, not just axes (PCA based by default).
• Color gradients can follow viewing direction using the probe tool.
Holographic Display Support
• Improved rendering pipeline enabling mesh and volume output to Looking Glass displays.
System & Performance Upgrades
• Major framebuffer redesign for more predictable rendering and faster effects.
• More intuitive shader system for both volumes and meshes.
• Smoother interaction when mixing render modes, effects, and color maps.
New Data Format Support
• DICOM reader.

FluoRender 2.33 Released

Thank you for choosing FluoRender! To get started, download the installer package for your operating system from the list of assets below. You can also find videos demonstrating the latest features of FluoRender on YouTube.

This release marks a major leap forward in both functionality and usability. With a focus on clarity, speed, and intuitive interaction, FluoRender now offers a richer toolkit for analyzing volume data from fluorescence microscopy. Users can dive into tasks like painting, filtering, and color mapping with minimal setup — or no manual fine-tuning at all — and still achieve precise, insightful results. Key updates streamline workflows, reduce the learning curve, and make advanced features more accessible. Whether you're visualizing time-dependent properties, customizing your color map, or leveraging smart automation, FluoRender adapts to your data and goals with surprising agility. The goal remains constant: to empower researchers with interactive tools that feel natural to use, yet remain powerful under the hood.

• Looking Glass Support

  FluoRender now seamlessly supports the latest Looking Glass holographic displays, expanding your visualization toolkit with intuitive controls and immersive clarity.
  The 16" Looking Glass displays can be purchased here: https://lookingglassfactory.com/looking-glass-16-lightfield
  The 27" Looking Glass displays can be purchased here: https://lookingglassfactory.com/looking-glass-27
  And a portable holographic display, the Looking Glass Go can be purchased here: https://lookingglassfactory.com/lkg-go

  Hologram Modes: Choose from three distinct modes to configure camera and lens shifts—whether for depth emphasis, parallax refinement, or artistic rotation.
  Projection Flexibility: Easily toggle between orthographic and perspective projections to match your analytical or illustrative needs.
  Camera Control Options: Navigate with precision using either Globe Mode for orbital rotation or Flight Mode for dynamic pathing.
  Auto Focusing: Automatically focus on the scene’s center or snap to a user-specified point with a simple click.
  Hologram Snapshot: Capture and share high-resolution holographic snapshots—perfect for documentation, collaboration, or presentation.

• Visualization & Volume Properties

  Overhauled volume property settings, with reordered layout
  New UI displays:
  Intensity distribution
  Color map range
  Updated multi-function buttons and introduced min–max/boundary high controls
  Filter updates: Lanczos-bicubic scaling with zoom-aware window sizing
  Added 4D color maps featuring time, intensity delta, and speed options
  Keyframe animation:
  Now supports volume property changes
  Enabled only when keyframes are present

• Interactive Tools & Painting

  New brush tools: Segment and Isolate
  Brushes support fine-grained grow rate
  Automatic threshold estimation for paint and component generation
  Isolate brush works with ruler tools to locate center points

• Automation & Scripting

  OpenCL filter script now runs last-used parameters from UI when left empty
  Added script stop command for one-time scripts
  Introduced automation options in configuration dialog, supporting:
  Histogram generation
  Paint selected size computing
  Component generation
  Colocalization
  Ruler relaxation

• Volume Filtering

  OpenCL editor renamed to Volume Filter
  Added deconvolution filters, including Richardson-Lucy and Wiener filters
  Updated Gaussian filters for smoother results
  Added other commonly used filters

• UI & Usability Enhancements

  Refreshed icons and text throughout UI
  Adopted notebook-style tabs for improved dialog navigation
  Dialog layouts can now be saved and restored
  Added dark mode support on Windows
  Reordered settings to prioritize frequent tasks
  Frequently used interactive tools now available in the workspace panel
  Added dialog buttons to the project panel
  Interactive tool states now sync across dialogs

• File & Data Management

  Added support for INI, XML, and JSON config file formats
  Refactored memory handling using smart pointers
  Updated volume cache system for time-sequenced data
  Introduced movie playback caching for smoother experience
  Enhanced capture capabilities: Support for JPEG and PNG
  Ability to read JPEG/PNG sequences

• Core System & Build Improvements

  Reorganized CMake structure and third-party libraries
  Switched from wxString and legacy path utilities to std::string and std::filesystem
  Resolved type cast warnings
  Updated FFmpeg integration to use current API
  Forward-declared third-party types for cleaner compilation

FluoRender 2.32 Released

Thank you for choosing FluoRender! To get started, download the installer package for your operating system from the list of assets below. You can also find videos demonstrating the latest features of FluoRender on YouTube.

New features:

• Support for a Wide Range of VR and AR Headsets: OpenXR API and its various extensions are included to support different headsets.
• Support for Microsoft Hololens 2: Use Holographic Remoting API to connect to the Hololens remotely.
• Support for Windows Mixed Reality Headsets: Use OpenXR Direct3D binding to display 3D contents.
• Meta Quest Headsets: Use OpenXR OpenGL binding to display 3D contents.
• HTC Vive Headsets: Both OpenXR and OpenVR can be used.
• Support for Various Controls for Viewing 3D Data: Head tracking, hand gestures, controller thumbsticks, controller poses, and voice commands can be used to control 3D data viewing when a supported headset is connected.
• Dynamic Gradient Background: The gradient background changes with the viewing direction.

Fixed issues:

• Dual Purpose Sliders: Resolved an issue that the thumb of the slider changes its position incorrectly in the "jog" control mode.

FluoRender 2.31 Released

Thank you for choosing FluoRender! To get started, download the installer package for your operating system from the list of assets below. You can also find videos demonstrating the latest features of FluoRender on YouTube.

New features:

• Looking Glass Holographic Display Support: Experience your data in a whole new dimension.
• Consolidated Progress Bar: Keep track of your tasks with a unified progress bar located at the bottom of the main window.
• Deep Neural Network Solver: Utilize machine-learning-generated volume property settings with our new solver. An example neural network trained by the developer is included.
• Ruler Display Options: Easily show or hide ruler points and names as needed.
• UI Code Reorganization: A more streamlined notification and update mechanism.

Fixed issues:

• Volume Baking: Resolved an issue where volume baking failed when cropping by clipping planes was enabled.
• MIP Mode: Fixed a display issue where MIP mode did not show correctly with a white background.
• Pencil Ruler: Corrected the problem where all points from multiple drawings were appended to a single ruler.
• Ubuntu Icon: Added an application icon for Ubuntu users.

v2.30

Thank you for using FluoRender. Download the installer package for your operating system from the list of assets below. Videos demonstrating the latest functions of FluoRender can be found on YouTube.

FluoRender Version 2.30 improved the user interface and added new functions. The highlights include:

• Slider control. A custom slider replaced the system's default control. It supports custom range and thumb colors, dual thumbs, and history record keeping for undo and redo operations. The orientation of the vertical slider was updated to have low values at the bottom. This behavior can be configured in the settings.
• Adjustable panel tabs. Settings in a panel of the main user interface are further organized into tabs. These tabs can be dragged by the mouse for reordering or changing layout.
• Undo/Redo. A history of changes made within the FluoRender user interface is saved for undo and redo operations. Operations for visualization adjustments can be undone or redone by the buttons in the main toolbar.
• UI layout restore. The previously adjusted UI layout will be restored when FluoRender is closed and relaunched. The UI layout is also saved in a project file.
• Display scaling support. The user interface is adjusted with the display scaling setting of the operating system. FluoRender needs to be restarted when the system scaling setting is changed. The UI layout saved before the scaling change will not be used.
• Multifunction buttons. The names of settings are shown as multifunction buttons in the output adjustment panel, the clipping plane panel, and the volume property panel. The function of a multifunction button is configured in the settings. The options include synchronizing settings of multiple channels, focused scroll, reverting the setting value to default, adjusting the setting value by machine-learning prediction, undoing the changes to the setting value, and disabling or enabling the adjustment of the setting value.
• Movie playback controls. Movie playback controls in the movie-making panel now support backward playback, looped playback, playback range in a clip, rewind/fast forward, and frame stepping. The scrollbar for movie playback can be configured to a jog mode for fast movie preview. The time point number of a time sequence is separated from the frame number of a movie clip. Therefore, time points of a time sequence can be remapped to arbitrary movie frame numbers in both the basic and keyframe animations.
• Keyframe animation templates. Commonly used keyframe animations can be quickly made by applying a keyframe template in the movie-making panel.
• Crop frame settings. The crop frame for exporting a screen capture or movie can be adjusted by dragging in the render view. The location of the scalebar can be also adjusted under the crop tab in the movie-making panel.
• Volume property settings. All volume properties with adjustable sliders can be individually enabled or disabled with checkboxes. Properties with low and high values can link the two settings.
• Display settings. Settings for the displays of a computer are organized under their own tab in the configuration window. New display settings include the aspect ratio of stereo rendering on a 3D TV, display ID for full-screen mode, and display color depth.

FluoRender 2.30 fixed these issues:

• The render view port failed to zoom correctly when the micro blending was enabled.
• Clipping of texts and icons.
• Movie playback became slower than the target FPS.
• The paint mask did not show when colormap was enabled.
• Game controller became unresponsive.
• Updating stopped prematurely for a large data in streamed rendering.
• A volume channel did not update correctly when MIP and large data streaming were enabled.
• Coloring of shadows when large data streaming was enabled.
• A very short movie failed to export.
• A volume channel failed to bake when clipping planes were enabled.
• Left and right eye offsets computed incorrectly in stereo rendering.
• Other user interface issues.

FluoRender 2.29.3 Released

Thank you for using FluoRender. Download the installer package for your operating system from the list of assets down below. Videos demonstrating the latest functions of FluoRender can be found on YouTube.

FluoRender Version 2.29.3 improved usability and fixed issues in the previous versions.

• A new script is added to FluoRender to reconstruct a 3D scene from rulers generated by tracking feature points in a pair of stereo videos.
• New functions for saving and managing project files improve usability.
• The Magnet function is added to the Edit tool for moving ruler points.
• A new script is added for generating synthetic walk animation from an average walk cycle, which can be extracted using the ruler tracking functions.
• A new script is added for converting the components in a time-dependent data set to animated rulers, which can then be processed using the scripts for rulers.

FluoRender 2.29.3 fixed these issues:

• An issue that the total number of training frames for DeepLabCut was limited to 128.
• An issue that the current time point was not loaded from a project file.
• An issue that previous movie frames were not cleared before rendering a next frame when large data streaming was on and rendering order was set to front to back.
• An issue in reading czi file that the volume was shifted.
• An issue that FluoRender crashed when editing component ID.
• Various user interface issues

FluoRender 2.29.2 Released

Thank you for using FluoRender. Download the installer package for your operating system from the list of assets down below. Videos demonstrating the latest functions of FluoRender can be found on YouTube.

FluoRender 2.29.2 adds support to run third-party or custom Python scripts. Specifically, the DeepLabCut is integrated into FluoRender, which is typically used for analyzing animal postures from video files. Typical machine-learning operations including model training and application can be accomplished only from FluoRender's user interface without direct interactions with Python.

• A new script for markerless training of video files. Use FluoRender's ruler tools to place points to be tracked on frames of a video file. Run the script to launch training using the DeepLabCut. Both Python and DeepLabCut need to be installed correctly by user.
• A new script for analyzing animal postures from video files. This script applies the trained model from the DeepLabCut and generate rulers, which can then be analyzed using existing scripts for ruler information extraction.
• Nikon ND2 format is supported on macOS.
• A smoothing setting is added to the script for image registration. It filters the motion vectors of registered results to make typical drift movement smooth.
• Volume properties can be automatically generated at loading by machine learning. The Auto Apply setting can be switched on in the Machine Learning Manager.
• A new script for analyzing the fluorescent intensity change over time on segmented components.

FluoRender 2.29.2 fixed these issues:

• The script for image registration failed to generate results when data had low signal-to-noise ratio.
• Laser wavelength and color information was not read correctly from the Bruker/Prairie View format.

FluoRender 2.29.1 Released

Thank you for using FluoRender. Download the installer package for your operating system from the list of assets down below. Videos demonstrating the latest functions of FluoRender can be found on YouTube.

FluoRender 2.29.1 included several improvements over the measurement functions and fixed issues. Additionally, the code directories and CMake file were reorganized to enable building on Linux systems. An executable file built on Ubuntu 22.04 is provided for download. This is an experimental build for Linux. We can make improvements per user request. Please provide detailed information on your OS version and hardware configuration when contacting us. Furthermore, a brief instruction for building on Linux systems is provided. Note that hardware support of OpenGL and OpenCL is needed in order to run FluoRender. Make sure that a specific Linux distribution and hardware configuration can support GPU computing. Please reference the readme file or use manual for further information.

• The ellipse ruler can be used as an ROI to sample intensity values of volume data. The region is defined by the projection of the ellipse from the viewing angle.
• We added a script to compute the fluorescence intensity changes. The changes are computed based on the resting state of an ROI, which in turn is estimated from the background intensity levels.
• We provide three methods of interpolation for ruler points: step, linear, and smooth.

FluoRender 2.29.1 fixed this issue:

• An issue that intensity values were not sampled correctly by rulers.

FluoRender 2.29 Released

Thank you for using FluoRender. Download the installer package for your operating system from the list of assets down below. Videos demonstrating the latest functions of FluoRender can be found on YouTube.
FluoRender Version 2.29 added a machine-learning module for training and applying various settings in FluoRender. The current version implemented machine learning for localized component generation (for volume segmentation) and volume property settings (for visualization). We also improved the ruler tools to enable measurement over time. Combined with the ability to directly load video files, the improved ruler tools can be used for analyzing movements and gestures in videos. Several issues in previous versions were also fixed.

• A unified user interface, called the machine learning manager, is added for the management of all machine-learning related operations. These operations are organized into categories, each having a similar workflow to manage training tables and start/stop training.
• Machine learning results can be applied to generate components at local level for more accurate and easier volume segmentation.
• Machine learning results can be applied to generate volume property settings.
• The locations of ruler points can be modified over time. Rulers can be animated over time to track the movements of single points or complex behaviors.
• A Magnet tool allows drawing a 3D curve and attracting closest ruler points to the curve. Rulers with multiple points can be modified easily with the magnet tool, especially for tracking over time.
• A Redraw tool allows drawing a 3D curve and redistributing all points of the closest ruler on the curve. Rulers with multiple points can be redrawn quickly without changing their topology.
• A setting is added to adjust the distance between two ruler points in the screen space when the pencil tool or magnet tool is used for drawing a curve.
• A script task for exporting ruler coordinates over time. Rulers can be used to track complex movements over time. Use this script to export the analysis results.
• A script task for replacing a volume channel. Use this script to change volume data upon finishing and rewinding of a script.
• A script task for switching the current script. Use this script to load a new script upon finishing and rewinding of a script.
• A script task for loading a project. Use this script to load a project upon finishing and rewinding of a script.
• Image registration can now be constrained using a paint mask. Only the data within the mask are considered, which can improve efficiency and accuracy.
• A DPI setting is added to the screen capture options. It can be used to automatically enlarge the captured image when a high DPI value is set. A base value of 72DPI is used. The DPI value is also saved in TIFF metadata.
• Commonly used video files can be directly loaded into FluoRender as 3-channel time-dependent volumes of single Z slices. Supported video containers include MP4, M4V, MOV, AVI, WMV, etc.
• When loading TIFF images of the 32-bit floating point format, the intensity values are converted to 16-bit integers.

FluoRender Version 2.29 made these improvements or fixed issues:

• An option to customize the behavior of the fixation operation in component generation. It determines whether the fixated regions can continue growing into other regions or they need to be constrained.
• A memory allocation issue when a time-dependent data set was processed using scripts.
• An issue that FluoRender crashed when an external program, such as a web browser or spreadsheet editor, was called to display analysis results using a script.
• Other user interface issues.

Notes:

• The Windows installer was updated on Jan-17-2023 to grant access permissions to the machine-learning tables in the installed directory. This is necessary for the machine-learning module to work. Otherwise, copy the FluoRender folder to a user directory.

FluoRender 2.28 Released

Thank you for using FluoRender. Download the installer package for your operating system from the list of assets down below. Videos demonstrating the latest functions of FluoRender can be found on the channel "FluoRender One" on YouTube.
FluoRender Version 2.28 added useful and practical functions for typical workflows to analyze fluorescence images. Several issues in previous versions were also fixed.

• The keyframe animation system supports setting keyframes on the sequence numbers of time-dependent data.
• The primary color of a volume channel is included in a keyframe. The animation of the volume color provides a visual indicator for the time progression.
• Improvements to the performance and robustness of the tracking of brush-selected data.
• A new script for rigid-body registration of images in a time sequence. The registration results can be saved as a new sequence and/or directly applied to render view transformations for real-time analysis.
• Fluorescence intensity sampling with the ruler tools. Intensity values are interactively updated.
• A ΔF/F mode that automatically computes the background intensity and normalizes fluorescence intensity values.
• Real-time fluorescence intensity sampling on registered time sequence data with the ruler tools.
• New analysis result templates supporting offline viewing.
• The ridgeline template supports line colors from ID values.
• Width adjustment to the borders of clipping planes.
• A new clipping plane display mode to show borders of only front facing clipping planes.
• New settings to the ruler profile script provide the maximum intensity and its location along a ruler.

FluoRender Version 2.28 fixed these issues:

• OpenCL code stability and performance issues on various platforms.
• A render view scaling issue on macOS computers.
• An issue that the default browser failed to launch when a script with analysis results was executed on macOS computers.
• An issue that the component generation did not produce valid results when both distance field and density field were enabled.
• An issue that the paint brush operation crashed FluoRender on macOS computers.
• An issue that the paint brush position did not match the render view on macOS computers.
• Flicker of UI widgets when a keyframe animation was played on Windows computers.
• An issue that exported Tiff volumes were not readable by programs such as Photoshop.