This change adds the ability to export MaterialX networks into the resulting
USD layer.
Details:
A new export option has been added to the USD export to enable MaterialX
export. It is off by default currently due to reasons in the caveats
section.
When enabled, it exports the MaterialX shading network alongside the
UsdPreviewSurface network, on the same USD Material. This allows the same
material to be used by renderers that don't support MaterialX, using the
USDPreviewSurface as a fallback. This is similar to setups in other DCC
packages, and matches the format we've used in our Reality Composer Pro
asset library.
It uses the existing MaterialX framework used to generate MaterialX
documents for rendering, to act as the basis for the USD graph. In this
process it also re-uses the existing texture export code as well if provided
and necessary.
Once the MaterialX document is created, use usdMtlx to generate a USD
shading network. Unfortunately, usdMtlx generates a graph that is unlike
what other DCCs that support MaterialX-embedded-in-USD generates. It
generates several extra prim hierarchies, and externalizes all shader
inputs, making them difficult to edit in other MaterialX graph editors.
To workaround this, generate the MaterialX shading network onto a
temporary stage, where we then run various pre-processing steps to prevent
prim collisions and to reflow the paths once they're converted.
The PrimSpecs are then copied over to their new path. The resulting prim
hierarchy matches what many artists we've worked with prefer to work with.
Caveats:
The Export MaterialX check is off by default. When using the Principled
BSDF, the resulting graph is very usable. However, when using some of the
other BSDFs, the shading networks generated by the existing MaterialX
framework in Blender generate some shading graphs that are difficult for
usdview and other DCC's to understand. The graph is still correct, but
because we're trying to prioritize compatibility, the default is off.
In future PRs we can aim to make the graphs for those other BSDFs play
better with other DCCs.
Other Implementation Details:
As part of this commit we've also done the following:
* Place some of the materialx graphs inside a passthrough nodegraph to
avoid node conflicts.
* Better handle some shader output types , and better handle some
conflict cases.
* Moved the ExportTextureFunction to materials.h due to some difficult
to resolve header ordering issues. This has no effect on any runtime code.
* There is a test for the MaterialX export that does some basic checking to
make sure we get an export out the other end that matches our expectations
Authored by Apple: Dhruv Govil
This PR is based on an earlier implementation by Brecht van Lommel , as well
as Brian Savery and his teams' work at AMD to implement the general
MaterialX framework within Blender.
Pull Request: https://projects.blender.org/blender/blender/pulls/122575
This patch implements blue-noise dithered sampling as described by Nathan Vegdahl (https://psychopath.io/post/2022_07_24_owen_scrambling_based_dithered_blue_noise_sampling), which in turn is based on "Screen-Space Blue-Noise Diffusion of Monte Carlo Sampling Error via Hierarchical Ordering of Pixels"(https://repository.kaust.edu.sa/items/1269ae24-2596-400b-a839-e54486033a93).
The basic idea is simple: Instead of generating independent sequences for each pixel by scrambling them, we use a single sequence for the entire image, with each pixel getting one chunk of the samples. The ordering across pixels is determined by hierarchical scrambling of the pixel's position along a space-filling curve, which ends up being pretty much the same operation as already used for the underlying sequence.
This results in a more high-frequency noise distribution, which appears smoother despite not being less noisy overall.
The main limitation at the moment is that the improvement is only clear if the full sample amount is used per pixel, so interactive preview rendering and adaptive sampling will not receive the benefit. One exception to this is that when using the new "Automatic" setting, the first sample in interactive rendering will also be blue-noise-distributed.
The sampling mode option is now exposed in the UI, with the three options being Blue Noise (the new mode), Classic (the previous Tabulated Sobol method) and the new default, Automatic (blue noise, with the additional property of ensuring the first sample is also blue-noise-distributed in interactive rendering). When debug mode is enabled, additional options appear, such as Sobol-Burley.
Note that the scrambling distance option is not compatible with the blue-noise pattern.
Pull Request: https://projects.blender.org/blender/blender/pulls/118479
The Glare node shifts the color of the highlights when the threshold is
high. That's because the thresholding algorithm simply subtracts the
threshold from the RGB data, which is not expected to retain the same hue
of the color.
To fix this, we do the thresholding only on the luminance of the color
in HSV color space. This eliminates the color shifting and also helps to
smooth the edges of the highlights.
This is a breaking change, but it is more of a fix rather than a change
of behavior.
Pull Request: https://projects.blender.org/blender/blender/pulls/122570
Split `bpy.types.LightProbe` into specialized subclasses.
- Rename all grid_* to remove the prefix in the RNA and limit access to
volume probe.
- Parallax and other sphere probe only properties should be limited to
sphere probe.
- `visibility` properties are deprecated (to be removed in a future
Blender version)
Ref #113976
Pull Request: https://projects.blender.org/blender/blender/pulls/122353
This handles the transition to EEVEE-Next (now EEVEE).
This removes some things that make no sense to keep
even for compatibility.
- Scene.eevee.light_cache_data
- Scene Light cache operators
- Scene Light cache RNA properties
The remaining legacy properties will be removed later
on to avoid python API breakage.
We keep the identifier of EEVEE-Next as `BLENDER_EEVEE_NEXT`
to avoid addons being incorrectly silently made compatible
with the EEVEE-Next where the Python API is different.
This renaming should be done in 5.0 release.
Thank you EEVEE-Legacy, you served us well.
Pull Request: https://projects.blender.org/blender/blender/pulls/122433
This commit adds logic to convert between USD dome lights and Blender
world materials.
The USD dome light rotation is represented in a mapping node input to the
environment texture. If the dome light has a color specified in addition to
the texture map, the color will be converted to a vector multiply on the
the environment texture output.
I the imported USD has multiple dome lights, only the first dome light will
be converted to a world material.
Co-authored-by: kiki <charles@skeletalstudios.com>
Co-authored-by: Michael Kowalski <makowalski@nvidia.com>
Pull Request: https://projects.blender.org/blender/blender/pulls/121800
This patch also fixes a crash when image input of `Stabilize2DNode` is unconnected and interpolation is set to bicubic.
Differences to GPU compositor:
- ~1px difference is observed due to different rounding of domain size vs. canvas size. This difference is constant for all image sizes.
- If image input is unconnected but other inputs are, CPU compositor doesn't consider the operation to be constant, whereas GPU compositor still outputs a constant result.
Pull Request: https://projects.blender.org/blender/blender/pulls/122288
Add guards to the RNA code to prevent the creation of legacy data on a
layered Action, and vice versa.
On a legacy Action, it is now impossible to create new layers or bindings.
On a layered Action, it's now impossible to create legacy F-Curves or
groups.
Refactor: Anim: rename bl_animation_id.py to bl_animation_action.py
The `Animation` datablock is no more, and this file tests `Action`.
No functional changes.
Pull Request: https://projects.blender.org/blender/blender/pulls/122483
EEVEE-next has removed the MA_BM_CLIP / alpha_threshold material
properties in favor of using nodes for equivalent functionality. This
changes USD to build and traverse node graphs during import and export
accordingly. Indirectly this allows Cycles to correctly render such
materials now too.
A complicating factor is that the UsdPreviewSurface defines its opacity
threshold using greater-than-equals[1], which Blender does not support
(and for which was technically already incorrect as EEVEE-legacy only
used greater-than for its shaders). Due to this we actually need to use
2 nodes: A less-than, followed by a one-minus invert, to arrive at the
proper value. We'll translate UsdPreviewSurface to this form on Import.
For Export we will look for either this 2-node pattern or a Round
node plugged into Alpha. Looking for Round is a result of the glTF
documentation which recommended the use of this node for thresholds of
0.5[2]. It's a tiny addition that seems reasonable to accommodate.
[1] https://openusd.org/release/spec_usdpreviewsurface.html (search for "opacityThreshold")
[2] https://docs.blender.org/manual/en/4.2/addons/import_export/scene_gltf2.html#alpha-modes
See PR for example images
Pull Request: https://projects.blender.org/blender/blender/pulls/122025
Add support for reading and writing custom `Curves` attributes.
This allows us to roundtrip Blender's Hair grooms containing UVs and
other attribute data. Note that animated attribute values are not
supported with this change.
This will also address #120042
Pull Request: https://projects.blender.org/blender/blender/pulls/121928
This patch implements the Fog Glow Glare node by porting the CPU
implementation, so it is not GPU accelerated and is not expected to be
realtime. However, after d4bf23771d, it is now fast enough to be usable,
see that commit for more information on the implementation.
The only difference is that the kernel part of the convolution is cached
in the realtime compositor, so it should be about 30% faster than CPU
for interactive editing.
In the future, this implementation will be replaced by a proper GPU
implementation, likely based on VkFFT.
This patch explicitly sets the compositing device in the compositor test
scripts. This is done to make the tests more robust, since a developer
might accordantly save a file with a different device.
Effectively, make GPU compositor available without need to enable
an experimental feature set.
The compositor device is now exposed in the Performance panel of
Render Buttons. It is also still available in the compositor's
N-panel, together with some other options which are more about how
editing works, and not exactly related to render performance.
Pull Request: https://projects.blender.org/blender/blender/pulls/121398
The new/experimental, layered `Animation` data-block is merged with the
existing `bAction` data-block.
The `Animation` data-block is considerably newer than `bAction`, so the
supporting code that was written for it is also more modern. When moving
that code into `bAction`, I chose to keep the modernity where possible,
and thus some of the old code has been updated as well. Things like
preferring references over pointers.
The `Animation` data-block is now gone from DNA, the main database, etc.
As this was still an experimental feature, there is no versioning code
to convert any of that to Actions.
The DNA struct `bAction` now has a C++ wrapper `animrig::Action`, that
can be obtained via `some_action->wrap()`.
`animrig::Action` has functions `is_empty()`, `is_action_legacy()`, and
`is_action_layered()`. They **all** return `true` when the Action is
empty, as in that case none of the data that makes an action either
'legacy' or 'layered' is there.
The 'animation filtering' code (for showing things in the dope sheet,
graph editor, etc) that I wrote for `Animation` is intentionally kept
around. These types now target 'layered actions' and the
already-existing ones 'legacy actions'. A future PR may merge these two
together, but given how much work it was to add something new there, I'd
rather wait until the dust has settled on this commit.
There are plenty of variables (and some comments) named `anim` or
`animation` that now are of type `animrig::Action`. I haven't renamed
them all, to keep the noise level low in this commit (it's already big
enough). This can be done in a followup, non-functional PR.
Related task: #121355
Pull Request: https://projects.blender.org/blender/blender/pulls/121357
This adds a light parameter to avoid near camera pixels
allocating too much shadow resolution.
This is more intuitive than the scale shadow setting and
allows reducing the precision without changing the look
of distant shadows.
For sun lights, the property sets the minimum pixel
size the shadow can contains. This allows to
remove a lot tilemap close up.
For local lights, there is another additional property
to set the maximum resolution in shadow space (like
EEVEE-Legacy) instead of in world space (like sun
lights). This allows making older files lighter and
allow a more conservative approach to resolution.
This add versionning code to handle EEVEE-Legacy files
that had fixed resolution per light type.
The resolution setting is always in world space distance
as the maximum shadow resolution distribution might change
in the future or be dependent on other parameters
(like beam angle). This ensure that there is no
dependency on these parameters, but make the
setting use very small units. But this is more of
a UI problem.
Pull Request: https://projects.blender.org/blender/blender/pulls/121701
If a curve was not using U parameter range of 0..1, the importer
did not properly detect whether the "U endpoint" flag should
be set on it. Fixed that, along with test coverage for the case.
This allows to expose these settings in the Performance panel in the
render buttons. Also moves compositor-specific options away from the
generic node tree structure.
For the backwards-compatibility the options are still present in the
DNA for the bNodeTree. This is to minimize the impact on the Studio
which has used the GPU compositor for a while now. They can be
removed in a future release.
There is no functional changes expected.
Pull Request: https://projects.blender.org/blender/blender/pulls/121583
These are really basic testing as well, merely linking or appending the
first collection or object in all processed blendfiles. But it should
help catching 'rare' issues in linking/appending from older blendfiles.
The `io_blendfile_versioning` test is currently one of the slowest
(excluding Cycles ones) in debug builds, it can easily take several
minutes to complete.
This commit split it into several instances, each processing a subset of
all the blendfiles.
This gives a strong speed-up when only running that specific test.
As expected, speedup is neglectable when running the whole test suite
though.
| instances | debug | release | debug all* | release all |
| --------- | ------ | ------- | ---------- | ----------- |
| 1 | 190.95 | 19.39 | 439.54 | 63.51 |
| 4 | 61.80 | 6.81 | N/A | N/A |
| 8 | 38.33 | 5.14 | 435.00 | 58.93 |
| 16 | 33.97 | 4.16 | N/A | N/A |
| 32 | 46.54 | 5.14 | N/A | N/A |
Times are in seconds.
`instances` are the number of tests generated (1 is same as before this
commit).
The first two columns are timings for running the versioning test only,
the last two are timings for the full test suite (excluding Cycles tests
in the debug build case).
The node uses multiple transform operations, each using its own sampling. Using bicubic sampling in the translate node causes undesired offsets.
These changes were intentional (see test updates in `c4e1be73`), but the offsets were thought to be too small for users to notice.
Pull Request: https://projects.blender.org/blender/blender/pulls/121495
This updates OIDN to 2.3.0-beta. The final version is planned to be
released in time for Blender 4.2 Beta. The most relevant changes:
1. Much higher quality when denoising with accurate prefiltering in *high*
quality mode, but at the cost of lower performance (use *balanced*
quality mode to revert to previous *high* quality mode)
2. Added new *fast* quality mode for 1.5-2x higher performance viewport
denoising
3. Lazy device module loading to avoid potential stability issues caused by
drivers of unused devices
4. Release CUDA primary context as soon as the OIDN device gets destroyed
to prevent potential memory leaks
To enable 2 and 3, code changes are needed in Blender, to be committed
separately.
Ref #118455
Pull Request: https://projects.blender.org/blender/blender/pulls/121359
Adds a test to validate the import of Mesh attributes for USD.
It reuses the recently added export attribute test file to generate the
USD file to load.
The test validates the current behavior in `main`. Missing data and bugs
are noted in comments and can be addressed afterwards.
Pull Request: https://projects.blender.org/blender/blender/pulls/121327
This patch implements the Fast Gaussian blur mode for the Realtime
Compositor. This is a faster but less accurate implementation of
Gaussian blur.
This is implemented as a recursive Gaussian blur algorithm based on the
general method outlined in the following paper:
Hale, Dave. "Recursive gaussian filters." CWP-546 (2006).
In particular, based on the table in Section 5 Conclusion, for very low
radius blur, we use a direct separable Gaussian convolution. For medium
blur radius, we use the fourth order IIR Deriche filter based on the
following paper:
Deriche, Rachid. Recursively implementating the Gaussian and its
derivatives. Diss. INRIA, 1993.
For high radius blur, we use the fourth order IIR Van Vliet filter based
on the following paper:
Van Vliet, Lucas J., Ian T. Young, and Piet W. Verbeek. "Recursive
Gaussian derivative filters." Proceedings. Fourteenth International
Conference on Pattern Recognition (Cat. No. 98EX170). Vol. 1. IEEE,
1998.
That's because direct convolution is faster and more accurate for very
low radius, while the Deriche filter is more accurate for medium blur
radius, while Van Vliet is more accurate for high blur radius. The
criteria suggested by the paper is a sigma value threshold of 3 and 32
for the Deriche and Van Vliet filters respectively, which we apply on
the larger of the two dimensions.
Both the Deriche and Van Vliet filters are numerically unstable for high
blur radius. So we decompose the Van Vliet filter into a parallel bank
of smaller second order filters based on the method of partial fractions
discussed in the book:
Oppenheim, Alan V. Discrete-time signal processing. Pearson Education
India, 1999.
We leave the Deriche filter as is since it is only used for low radii
anyways.
Compared to the CPU implementation, this implementation is more
accurate, but less numerically stable, since CPU uses doubles, which is
not feasible for the GPU.
The only change of behavior between CPU and this implementation is that
this implementation uses the same radius, so Fast Gaussian will match
normal Gaussian, while the CPU implementation has a radius that is 1.5x
the size of normal Gaussian. A patch to change the CPU behavior #121211.
Pull Request: https://projects.blender.org/blender/blender/pulls/120431
This patch matches the size of the Fast Gaussian mode of blur with the
standard Gaussian mode. The sigma value was computed as half the radius,
while it should be third of the radius, since Blender's Gaussian
function is truncated at 3 of the standard deviation of the unit
Gaussian. The patch include versioning to adjust the size of existing
files.
Pull Request: https://projects.blender.org/blender/blender/pulls/121211
Implement the design discussed in #120384.
This adds two parameters. One for changing the approximation
method, and another to use the thickness from shadow map.
We pack the former in the gbuffer by dividing the 16bits
used for thickness by two and use one bit to store the
method.
The thickness from shadow map is now decoupled from the
light evaluation shader. This makes it more performant and
compatible with ray-tracing. This commit also uses the
same biases as shadow mapping to avoid aliasing artifacts
(fix#119339).
This refactors the light evaluation quite a bit to
remove unused bits bits and make the whole transmission
light evaluation without too much complexity.
Pull Request: https://projects.blender.org/blender/blender/pulls/121171
Add an RNA enum property `AnimData.animation_binding` that lists all the
bindings available in `AnimData.animation`.
The list of bindings is filtered to only contain the bindings suitable
for the animated ID. This prevents assigning a 'camera' binding to a
mesh.
Un-assigning is done via an operator, represented as an 'X' button in
the interface.
The enum property contains up to two special items:
- "New" to create a new binding for the ID.
- "(none/legacy)" to indicate that this ID doesn't have a binding
assigned. This one is conditional, and only appears when it is
necessary.
These two special items are experimental, and mostly exist because we're
still evaluating things and building a better UI. It is intended that
the binding selector will become as close to the ID selector as
possible.
-----------
Note that this PR also contains #121268 as it builds up from that one, and I didn't want to wait with testing on the buildbot until that one lands.
The new Baklava panel:
Pull Request: https://projects.blender.org/blender/blender/pulls/121269
Refuse to set the RNA property `AnimData.animation_binding_handle` when
`AnimData.animation` is nil.
When in a hypothetical future `AnimData.animation` would be set, the
binding handle would be ignored anyway, as it's only meaningful within
the context of a known animation. Refusing to set the binding handle in
cases where it is known to be meaningless may help to find bugs in
Python scripts.
Pull Request: https://projects.blender.org/blender/blender/pulls/121268
