Documentation

Select Footage Panel

The Select Footage panel is the first panel you will encounter in MotionMaster 3D, and it is also where every new project begins. Here you decide which footage you want to work with. In Figure 1 you can see the full panel, where each numbered element is referenced in this documentation.

Image Sequence Selection

The first option [1] is Select Image Sequences. This lets you point the addon to the folder containing your image sequence. For example, if you select C:\ImageSequences\Sequence1, the addon will scan the folder, collect all images, and sort them by name.

You can either type in the path directly or use the folder icon [2] to browse for it.

Video or Sequence Converter

The Video or Sequence Converter is where you prepare your media. It can either transform a video into an image sequence or adjust an existing image sequence to a new resolution, color depth, or step rate.

When you select a file or folder [3], the addon automatically detects what you picked. If you select a video, the text "Video detected" will appear, while choosing a folder of images will show "Image sequence detected: [Number] frames".

From here you can adjust a few important settings:

• •
  Color depth [4] allows conversion to 8-bit or 10-bit images.
• •
  Step rate [5] determines how many frames are skipped. For example, a video with 100 frames and a step rate of 5 will result in 20 images. A higher step rate is useful when you only need a mesh, while a step rate of 1 is recommended for accurate camera tracking.
• •
  Target resolution [6] lets you downscale the footage for faster processing or keep the original resolution if you prefer maximum quality.

Once the settings are chosen, click Convert Video or Convert Image Sequence [7]. A new folder called Sequence will appear alongside your media, containing the processed images.

Batch Mode

The final subpanel is Batch Mode, enabled with the checkbox [8]. This feature is designed for larger projects where you want to process multiple sequences in one go. Instead of setting paths and running operations manually for each sequence, you can prepare everything at once and let the addon process them automatically.

First, you need to set an Output Folder [9]. Each processed sequence will create its own subfolder inside this directory. After that, you decide what type of results should be imported. This is controlled by the Import Options, and here's how they work:

• •
  With no boxes checked, the addon will import the standard Heavy camera tracking result.
• •
  Enabling Import Quick [10] switches it to the quick tracking result only (this disables all other import options).
• •
  The Smart Importer [11] still uses Heavy tracking but pays extra attention to aligning keyframes correctly, which helps avoid mismatches between the solved camera and the background footage.
• •
  Import Undistorted [12] brings in the result calculated from undistorted footage rather than the regular sequence.
• •
  Import Dense [13] includes the dense point cloud generated when High Quality Dense is enabled in the Heavy panel.
• •
  Import Mesh [14] adds the generated mesh to your scene.

Finally, you can add sequences to the batch using the Add button [18]. Each new entry lets you specify its path with the folder icon [15]. Make sure the sequences you actually want to use are enabled via their checkboxes [20].

To delete a footage entry, click the delete button [16]. After that is done, you can perform the operation you want, for example the Heavy camera tracking operator, and once that is finished you can click the Import button [17] with the correct import options in place as mentioned above.

Quick Section

The Quick Autotrack feature in MotionMaster 3D provides a fast way to track your footage using SLAM (Simultaneous Localization and Mapping). When using this method, you get a live preview of the camera solve combined with your footage, showing how features are extracted and how the camera's movement is calculated.

Keep in mind that this method requires knowledge of your camera intrinsics, such as sensor width, focal length, and resolution in both X and Y, in order to accurately solve the camera. It is also helpful if you know the distortion parameters of your lens, but this is not strictly required unless your footage has significant distortion.

With that said, let us go through the Quick panel in detail and explain what each parameter does.

Saved Optics Presets

The first section is the Saved Optics Presets tab. Here you can save your current configuration as a preset for later use by clicking Add preset [1]. To apply an existing preset, open the dropdown menu and select the one you want [2].

Optics Configuration

The main configuration box is where you define the core intrinsics. You can enter the Focal length (mm) [3] or the Focal length (px) [4], as well as adjust the Sensor width [5] and the footage resolution in X [6] and Y [7].

These values are linked: for example, if you change the sensor width or resolution, the focal length in pixels will automatically update because these parameters are constrained by a formula. Correct optics settings are crucial for a successful track. Most problems that arise are due to incorrect intrinsics, so make sure these values are accurate. If you notice that almost no features are being extracted, it is likely because the optics settings are incorrect, which results in the wrong focal length in pixels.

Advanced Settings

The advanced settings include the optical centers in X [8] and Y [9]. By default, these are set to half of the resolution values [6][7], and they update automatically when the resolution changes. This is intentional, since most users do not need to adjust them manually. However, if you do want to set custom values, first define the resolution, then overwrite the optical centers.

The rest of the fields are for lens distortion values: K1 [10], K2 [11], K3 [12], P1 [13], and P2 [14]. In most cases, you can leave these at their defaults unless you know your camera's calibration.

Performance Settings

Performance settings affect how the tracker extracts and matches features. The Number of features [15] controls how many features to look for in each frame (default 1000). Levels [16] defines how many pyramid levels are used. A pyramid is a set of downscaled copies of each frame, which allows the tracker to find features at different scales (zoomed in or out).

Each level is scaled according to the Scale factor [17]. For example, with a scale factor of 1.20, level 0 is the original image, level 1 is scaled to 1/1.2 of the original size, level 2 to 1/(1.2²), and so on. A smaller scale factor gives finer steps between levels but requires more levels to cover large zoom differences, while a larger scale factor is faster but less precise.

The Init threshold fast [18] sets the primary threshold for the FAST corner detector. It looks for strong, obvious features first. If not enough are found, the tracker falls back to the Min threshold fast [19], which accepts weaker features. If you are not getting enough features, try increasing the feature count [15] and lowering the threshold values [18][19].

Viewer Settings

The Viewer settings control how the live preview window looks. While most users will not need to change these, they can be useful for customizing the display.

You can adjust the keyframe size [20] and keyframe line width [21] to change how keyframes appear, and the graph line width [22] to control the thickness of the motion path. The point size [23] changes how large feature points are displayed. The camera size [24] and camera line width [25] affect how the drawn camera rig looks. Finally, the viewpoint parameters [26–29] set the default viewer position and focal length, which define the angle and distance from which the tracking scene is shown.

Tracking Direction and Overlap

At the bottom of the panel are the Forward [31] and Backward [30] options, which control the tracking direction. Quick Autotrack is a SLAM-based tracker, meaning it only has access to past frames, not future ones. This allows real-time tracking but also requires an initialization period before it can begin solving the camera. As a result, the first few frames will not have a solved camera position.

For example, in a 100-frame sequence, the tracker might only start solving at frame 20, leaving frames 1–19 without a solution. To minimize this gap, try to include parallax movement (camera translation) early in the footage, as this helps the tracker initialize faster.

The Smart overlap [32] feature improves this process by limiting the backward pass to only the region where the forward tracker failed, plus a margin. The margin is controlled by the Overlap setting [33]: Low (factor 2), Medium (factor 3), or High (factor 4).

For example, if the forward pass starts solving at frame 20, the initialization period is 19 frames. With Smart overlap set to Medium, the backward pass will start at approximately frame 77 (20 + 19 × 3) and track backwards to fill the gap. In most cases, you can leave these settings at their defaults.

Running Autotrack

Once everything is configured, press Autotrack (Quick) [34] to begin. Two windows will appear: one showing the live camera solve and one showing the feature extraction. If both Forward and Backward tracking are enabled, these windows will reopen for the second pass. When the process finishes, you will see the extracted 3D points and a moving camera in your Blender scene.

Focal Length Estimation

The Quick panel also includes a Focal length estimation tool. Set a start frame [35] and an end frame [36], then press Estimate focal length [37]. The addon will run a short process using the Heavy Autotracker in the background to estimate the focal length from that section of footage.

Heavy Section

The Heavy panel is the primary processing panel in MotionMaster 3D. It combines COLMAP, a custom-compiled GLOMAP, and OpenMVS to deliver robust and high-quality results. In this panel, you can configure all major operations: feature extraction, matching, mapping, and mesh generation. The Heavy tracker is reliable and stable, offering a balance between processing time and accuracy. By adjusting the available settings, you can fine-tune performance and result quality to suit your project.

This panel includes two modes which are the Simple mode and the Custom mode. As you can tell from the names, one is intended to be much simpler than the other, which allows you to make custom changes and apply the exact settings you want. We will first go through how to use the Custom mode and then the Simple mode, starting with the custom mode:

Main Settings (Custom mode)

The first section is the Settings box, which defines the key operations and processing options.

• •
  Advanced Feature Detection [1] enables domain size pooling during feature extraction. This makes the system more robust on difficult footage with motion blur or large perspective changes. Note that enabling this forces the extraction step to run on the CPU, even if CUDA [9] is enabled.
• •
  Zoom [2] allows focal length to vary across frames. Use this if your footage contains zooming, as the solver will no longer assume a constant focal length.
• •
  Undistort [3] processes each image to remove lens distortion and saves undistorted versions in a subfolder.
• •
  Max Size [4] sets a limit on image resolution. For example, if your footage is 6000 × 4000 pixels, and you set Max Size to 3000, the images will be downscaled to 3000 × 2000.
• •
  Features [5] controls the maximum number of features extracted per image. More features increase detail and accuracy but also processing time.
• •
  High Quality Dense [6] is available only when Undistort [3] is enabled. It generates a dense, colorized point cloud from your footage, allowing highly detailed 3D scene reconstruction.
• •
  Intelligent Matching [7] uses a smarter strategy for feature matching between images, improving efficiency and robustness.
• •
  Upright Footage [8] assumes the footage has no roll rotation, greatly speeding up feature extraction.
• •
  CUDA [9] enables GPU acceleration for supported operations.
• •
  Overlap [10] sets how many neighboring images are compared during matching. Lower values are faster but less accurate, while higher values improve reconstruction quality at the cost of speed.

Advanced Settings

The Advanced settings allow further control over tracking and reconstruction. For most cases, simply select the Medium preset [11], which balances speed and quality.

• •
  Fast preset reduces processing time significantly but may lower tracking accuracy.
• •
  Heavy preset is slower but produces higher quality than Medium.
• •
  Extreme preset switches to exhaustive matching and additional refinements, producing the highest quality results but with very long processing times.

All presets adjust the values in the advanced settings, but you can also override them manually:

• •
  Feature Extractor [12] improves extraction on footage with strong perspective changes.
• •
  Confidence [13] sets the geometric verification threshold. Higher values are stricter, filtering out more bad matches but also potentially discarding good ones.
• •
  Max Ratio [14] controls how selective matching is. Lower values give fewer but more reliable matches; higher values keep more matches but risk false positives.
• •
  Min Inlier Ratio [15] defines the minimum ratio of inliers required for a match to be accepted.
• •
  Min Inliers [16] is the absolute minimum number of inliers needed between two images for them to be linked.
• •
  Max Matches [17] limits the number of matches per image pair to avoid memory overload and excessive processing time.
• •
  Max Trials [18] sets the maximum number of attempts to fit a geometric model. Higher values improve reliability but slow things down.
• •
  Max Error [19] sets the maximum pixel error allowed for a point match. Lower values enforce stricter accuracy; higher values are more forgiving but less precise.

Loop Closure

Enable Loop Detection [20] activates loop closure detection. This helps when the camera revisits the same area, such as walking in a circle or orbiting around an object, ensuring consistent alignment.

Track Establishment

These settings control how many observations are required for reliable tracks:

• •
  Min Views Per Track [21] is the minimum number of images that must observe a 3D point for it to be accepted. More views increase reliability but may discard useful points if set too high.
• •
  Min Tracks Per View [22] defines the minimum number of points required for an image to be included in the reconstruction.

Optimization

Optimization settings refine the final reconstruction:

• •
  Bundle Adjustment [23] sets how strongly the system refines camera poses and 3D points. Higher values improve accuracy but require more computation.
• •
  Global Positioning [24] controls how strictly global position constraints are enforced. Lower values allow flexibility, while higher values enforce tighter alignment.
• •
  Min Triangulation Matches [25] sets the minimum matches needed to triangulate a new 3D point. Lower values create more points but may add noise, while higher values give cleaner geometry but fewer points.

Running the Heavy Tracker

To start the Heavy Autotrack, press Autotrack (Heavy) [26]. Make sure the system console is open so you can follow progress in real time. To stop the process, click inside the console window and press Ctrl+C.

Now that you have gotten a good understanding of the different settings in the custom mode let's go through the simple mode:

Main Settings (Simple mode)

Now nearly all the different settings are gone and left are some presets and a few options. Currently the simple view offers 9 different premade presets [1] which are each intended to be used in different scenarios which we will go through one by one:

• •
  Fast preview, use for fast first pass to see if clip reconstructs.
• •
  General Purpose, this is the default preset and is intended to be used in most scenarios such as handheld footage with mixed textures. This is a balance between speed and performance.
• •
  Repetitive textures, this is good for scenes with repetitive textures such as grids, bricks, architecture, floors and more.
• •
  Forward motion, when walking straight, car dash, head-mounted and so on.
• •
  Orbit / Turntable, when rotating around object, tripod pan, turntable.
• •
  Extreme robust, good for more challenging scenarios which requires extreme reliability.
• •
  Moving subjects, use for dynamic scenes with people/objects moving in frame.
• •
  Sparse features, good for deserts, walls, uniform surfaces, low-texture scenes.
• •
  Last Resort, use when all other presets fail, last attempt.

(If you notice that any specific configuration works great in any particular situation and you would like to share it with the community feel free to contact me and I might add it as a preset available for everyone in the next update)

Options

Here you have the option to turn on or off two features, the Variable Focal Length [2] and the Angled/Rotated Camera [3]. The first option should be enabled if your footage includes a zoom, that means that your focal length changes throughout the sequence. The next option is the Angled/Rotated Camera [3] which should be enabled if you rotate your camera a lot in the sequence you are trying to track.

Essential Settings

In this box you have three settings to decide. The first one being the CUDA (GPU) [4] and when this is enabled parts of the processing will happen on the GPU, note that you need to have a Nvidia GPU for this to work. Next up is the Undistort [5] setting and when this is enabled it will undistort your footage and try to find the distortion intrinsics of your footage. Lastly is the High Quality Dense [6] setting which requires the Undistort [5] option to be enabled for it to work therefore it is greyed out in the right image. When the High Quality Dense [6] setting is enabled it also makes the heavy tracker produce a dense point cloud of the scene. Each point in the point cloud will receive its own color from the footage making it accurately replicate the captured scene. Note that the High Quality Dense [6] feature needs to be enabled in order for you to later be able to generate a mesh of the tracked scene.

Next up is just to press the Autotrack (Heavy) [7] button and the heavy autotracker will begin and use the settings you have chosen. Note that you can see the progress of the operator live in the progress panel at the top or inside Blenders system console. You can also cancel/stop the current process by going into the system console and pressing ctrl + c and it will stop the current process.

Mesh Generation

If Undistort C[3]/S[5] and High Quality Dense C[6]/S[6] are enabled, you can also generate a mesh. The Mesh Quality C[27]/S[8] setting offers four presets: Low, Medium, High, and Ultra. Higher settings produce finer meshes but require longer processing times. Then you can press Generate Mesh C[28]/S[9] and it will use the current dense point cloud and the mesh quality setting to generate a mesh.

Conclusion

The Heavy panel gives you complete control over the tracking and reconstruction process. In most cases, using the simple view and the general purpose preset provide a good balance between speed and quality. From here, the next step in the workflow is the Point Cloud Cleanup panel, which helps clean up your newly created point cloud.

Point Cloud Cleanup

The point cloud cleanup panel is a relatively simple panel, it is only meant to provide a simple way for you to filter out unwanted vertices in your point cloud. It uses a statistical outlier removal (SOR) method which generally works well.

Selecting Your Point Cloud

The first thing you do is to reference your point cloud in the Target object box [1].

Cleanup Parameters

Next is for you to decide the cleanup parameters. These two settings control how aggressively the algorithm removes outlier points from your point cloud.

Running the Cleanup

After you have played around with the settings, press the Cleanup Point Cloud [4] button which will perform the filtration on your point cloud. The process should complete quickly, removing unwanted noise points from your cloud.

Next Steps

From here, the next step in the workflow is the Aligner panel, which helps orient and align your tracked solution.

Aligner Section

The Aligner panel is much simpler than the previous panels. It provides two main tools to quickly orient your tracked solution.

The Floor [1] feature allows you to select a set of vertices and align them with the XY plane. When you click Floor, the parent object of the result is adjusted so that the selected points lie flat, effectively snapping your ground plane into place.

If the orientation is inverted after applying Floor (for example, the ground points end up facing the wrong way), you can use the Flip [2] button. This flips the Z axis of the parent object, switching the orientation from -Z up to +Z up.

Reset Section

The Reset panel helps you clear out old tracking data from a specific folder while keeping your image sequence intact. Every time you use Quick or Heavy Autotrack in MotionMaster 3D, files containing 3D data from previous runs are generated. Over time, these can clutter your project folder and make it harder to iterate with new settings. The Reset panel makes it easy to remove this data so you can quickly start fresh.

The first option is Also delete subfolders [1]. When enabled, the reset operation will not only remove files but also delete any subfolders within the current working folder.

The second option is the Reset [2] button. Clicking this will remove all files in the current working folder except for your image sequence. If Also delete subfolders [1] is enabled, subfolders will be removed as well, effectively erasing everything in the folder except your source images.