Enables motion blur. For convenience, duplicates the switch on the tab Quality Options. Other settings on the tab Shutter.
Depth of Field
Enables DOF. For convenience, duplicates the switch on the tab Quality Options. Other settings on the tab Aperture.
Adaptive Image sampling
Enables adaptive image-space sampling, consisting of several algorithms. For convenience, duplicates the switch on the tab Quality Options.
Adaptation level
Determines the degree of aggressiveness during adaptation. 0 - low aggressiveness, 1 - high.
Recompute on n-th pass
Each pixel in the final image gets some priority, depending on noise and other factors. These priorities will not be recalculated at every pass, but only at every n-th pass.
Variance max estimation diameter
Intermediate renderings contain noise. This noise has a certain spectrum, therefore, in order to evaluate it, it is necessary to take into account not only the pixel for which the estimation is made, but also the neighboring ones. The larger the evaluation radius, the lower the noise frequencies that will be detected. Low-frequency noise is much more unpleasant for the human eye than high-frequency noise, so it must be disposed of.
Variance blur diameter
After the noise for a particular pixel is estimated, its estimate can be blurred into neighboring pixels so that they also receive their share of priority. This allows you to transfer part of the priority from very noisy pixels to neighbors.
Adaptive Image sampling. Importance
Enables pixel luminance noise adaptation. Dali Renderer keeps two copies of the rendered image, for odd and even passes. Comparing these instances allows a fairly accurate estimate of the current noise. This requires more memory, but it is more accurate than estimating the noise level when compared with neighboring pixels.
Prefer lights
Brighter pixels will have more priority.
Tail sampling
As stated in the description for Quality Options, there are "golden" values for the number of passes at which the picture becomes the least noisy. This rule conflicts with adaptive rendering, where different pixels will be sampled a different number of times. When this option is enabled, the pixel will continue to be sampled until the next "golden" value, even if its noise already meets the stop signal-to-noise criterion.
Adaptive Image sampling. Time
If the rendering is done with Motion Blur turned on, then very fast moving objects are possible, which for certain pixels on the screen are present only for a short period of time. In this case, it is very difficult for this pixel to be "hit" by the ray. This option enables an algorithm that, for each area of the screen, accumulates statistics on the location of objects in time. Thereafter, the time for these areas is deformed and this can greatly reduce the noise. The current implementation adapts constantly and this introduces subtle luminance distortions (bias).
Adaptive Image sampling. Spectral
The option is currently locked.
Emitter resampling
Enables a special mathematical algorithm that can significantly reduce noise in some scenes with a large number of local light sources.
Emitter adaptation
Enables adaptation of light sources depending on their actual effect on the picture. For convenience, it duplicates the switch on the Quality Options tab. During several training passes, statistics are accumulated, then a three-dimensional database is created on the effect of emitters on the scene. In further rendering passes, it is used for optimal selection of light sources. The presence of training passes eliminates the introduction of distortions (unbiased). The parameter is the size of the database in gigabytes.
Path guiding
Enables adaptation of secondary ray bounce directions. For convenience, it duplicates the switch on the Quality Options tab. During the training passes, statistics are accumulated on the most profitable directions, then a three-dimensional database is created and further used for more efficient sampling. Each part of the scene "knows" where the light is mainly coming from and directs the rays from the camera there. The presence of training passes eliminates the introduction of distortions (unbiased). The parameter is the size of the database in gigabytes.
Multipass adaptation
In the current implementation, this option means that there will be several blocks of training passes. This option is required for volumetric caustics. For surface caustics, it is desirable but not necessary.
Path guiding. Volumetric
Enables the accumulation of statistics for Emitter Adaptation and Path guiding in volumetric materials: fog, water, etc. The parameter is the size of the database in gigabytes.
Path guiding. Choice
The option is currently locked.
Reconstruction filter
The filter that will be applied when adding a sample to the final picture.
Filter diameter
Reconstruction filter diameter. The larger it is, the softer the picture is.
AA Spread diameter
Antialiasing width in pixels. For each pass, the ray does not start through the center of a pixel, but at a random location within the boundaries of that pixel. This parameter sets the width of such a zone.
Ray diameter
The initial diameter of the ray from the camera, in pixels. Affects the sharpness of textures. This is what is called "Ray Differential".
Pure randomness
Includes complete randomness in everything. Disables all sampling mechanisms that are designed to reduce noise. This option is purely for debugging purposes if you suspect something is wrong. May be removed from the GUI.
Blue noise samplers
Includes optimized blue noise matrices for sampling. Should always be on.
Reconstruction sub-pixel kernels
Without this option, the pixel value is added to the final buffer as if the ray were exiting the center of the pixel. With this option, the correct effect of the ray on the surrounding pixels is calculated. In practice, the difference is hardly noticeable.
Add chaos for better AA
This option adds a bit of chaos to the artificially constructed pseudo-random numbers. Useful if there is a suspicion of parasitic rendering dependencies. In practice, it was only needed in synthetic tests.
