tyFlow v2.000

[tyFlow]

tyFlow v2.000 is up!

https://pro.tyflow.com

Additions:

* Integrated new ZENITH solver for CUDA fluid simulations, as well as 20 new Inferno operators
* Updated tyFlow CUDA modules (package now also includes ML models for fluid upres/compression, as well as specific install location - see included README file for all info).
* Minimum supported CUDA version raised to 12.8 (please ensure your GPU drivers are up to date).
* added Global operator (see docs for more info)
* added MAXScript operator (see docs for more info)
* added tyMeshBlend modifier for blending normals where meshes intersect
* added SDF shape mode to PhysX Shape and PhysX Collision operators (for more accurate concave collisions)
* tyCache modifier can now cache spline data
* Export Particles operator now supports subframe export of tyCache data - tyCache objects now support loading subframe data (this enables support for tyCache subframe motion blur)
* added UVW seam weld options to multifracture, tyBoolean and tyWeld modifiers
* middle-clicking any tyFlow spinner UI control will bring up a Houdini-style value-stepper dialog, that can be used to incrementally modify spinner values by various orders of magnitude (thanks to John Martini for the reference implementation)
* added knot selection and direction controls to the tyPush modifier's spline push parameters
* added group filtering to the Fuse operator
* added neighbor tracking option to Property Test neighbor test, for tracking which neighbors were found by a particle
* added distance/falloff params to tyPush modifier's object push mode
* added interpolation settings to Restore operator
* Comboboxes, checkboxes, radiobuttons will now display non-default values in blue (previously only spinners did this)
* added custom vector magnitude/direction modes to Speed operator
* added PhysX bind connectivity display to tyCache object
* large dropdown comboboxes (ex: Custom Properties operator channel modes) now have a text entry at the top that you can use to search for items within the list (it's a permissive, fuzzy search, so text entered doesn't need to be an exact match to desired entry)
* added "Merge mat IDs" mode to tyMaterial modifier, for merging/concatenating all material IDs of objects that the modifier is instanced across (and also generating a compatible multi-material for the re-ordered material IDs)
* added "Merge mat IDs" option to tyMesher object, for merging/concatenating all material IDs of input objects (and also generating a compatible multi-material for the re-ordered material IDs)
* Export Particles operator now supports exporting ranges for tySwitcher indices and reseed values, allowing you to export permutations for both from the same operator in a single export operation. Export object names/layers now also support symbol replacements (ex: $tySwitcherIndex, $tySeedIndex, etc)
* updated OpenVDB SDK to version 13
* added "simulation age" option to operator timing rollouts, for controlling particles relative to the start of the simulation (the earliest start frame of any Birth operator), rather than an explicit frame value
* added options in Display Data operator to normalize/scale/colorize velocity vectors

Fixes:

* a warning will now be displayed during Max startup if a problematic version of vcomp140.dll is installed on the machine (v14.44.35208-v14.50.35719 can cause Max to crash if scheduled omp threading is used anywhere in Max, not just tyFlow)
* fixed a regression where VRay GPU wouldn't render particles with pivot offsets correctly (mainly multifractured particles)
* fixed a regression that prevented PhysX GPU from initializing properly
* optimized Set Target operator's early-exit condition (speeds up computations when no particles in its event meet its timing conditions)
* updated tyFlow's versions of CUDA cub/thrust libraries, which should fix some rare CUDA crashing issues during Multifracture operations
* fixed an issue where the Property Test operator's neighbor sweep mode was not working correctly (wrong velocity magnitude was used for sweep test)
* fixed an issue that prevented the Flow Update operator's timing controls from working when updating particles input from a tyCache
* fixed an issue where removing deleted items from a Script operator's accessible object list would cause the Script operator to recompile for each removed item
* fixed regressions which prevented tyFlow from interfacing with FumeFX grids properly
* fixed an issue that could cause flow indeterminism when using per-particle keyframe offset/variation
* fixed a race condition in the Custom Properties operator that could cause a crash when modifying particle target values
* fixed an issue that could cause editor lag when operator(s) had custom keyframe offsets enabled
* fixed an issue that prevented tyParticleSkin modifiers (using an underlying tyCache for PhysX tearing deformations) from properly welding untorn faces together
* fixed an issue where triangles were not properly skinned by tyParticleSkin modifiers when in PhysX tearing mode using polygons center as the search criteria
* tyParticleSkin auto-welding will no longer affect/change resulting mesh face count
* fixed an issue where Value operators' relative-to-property custom float mode wasn't working
* fixed an issue where operators could be pasted into incompatible events
* fixed an issue where certain startup tyFlow scripts could cause some Max/Windows configurations to crash during startup.
* fixed some visual glitches that could occur in the tyFlow editor if rollout columns hidden
* fixed an issue that could cause the Position Hair operator to crash in "closest strand" mode, and an issue that could prevent the Hair Bind operator from working
* fixed an issue where increasing offspring count in the Spawn operator (in travel mode) was not properly reseeding particle positions in parent surface spawn mode
* Birth Shape operator will now calculate object velocities directly, rather than from last particle positions (this allows it to compute proper velocities in history independent mode)
* Spawn operator (in travel mode) will now compute proper trail velocities in history independent mode
* fixed an issue where mass values created by the Mass operator were clamped by the binding min/max values even when binding mode not active
* fixed an issue where the parameters for certain listboxes (ex: tyActor - Animation rollout, tyProperties - custom floats) would be greyed-out after creating a new listbox entry, preventing their modification
* fixed an issue where tyCache scale overrides were transforming in world space instead of local space
* fixed an issue where certain Script operator mesh operations would fail when global mesh cache enabled
* fixed a few issues in the official tyFlow example scenes and added Value operator examples

[NaxosCG]

Inferno looks amazing and fast !

How do we render this (here Corona renderer is the engine i use) ?

Any special thing to check to get it rendrerd ?

[tyFlow]

For now you will need to export to VDB using the Inferno Export operator, and render using each renderer's preferred VDB container. All render devs have been sent the Inferno API, but it may take time for it to be implemented by them.

[NaxosCG]

Thanks, great one, again !

[d4rk3lf]

Incredible news.

And finally, I am not overwhelmed with work, as usually I am, when new hot stuff are released.
DOWNLOADIIIING!!!

[tyFlow]

Some display and crashing issues that occurred in my initial release of this build have been fixed - if you experienced anything like that please re-download v2.000 and try it again...

[instinct-vfx]

Looking great! For the CUDA DLLs and ML models: Is it possible to customize the location (to deploy tyFlow using rez for example)?

[tyFlow]

Currently the new location is hard-coded. I can potentially make it customizable, though...let me think about the best way to do that.

[instinct-vfx]

Just as a heads up, from a rez packaging perspective an environment variable would be easiest.

[tyFlow]

Sure, I could add that...

[chriss3d]

Is there any preset or recommended settings for a V-Ray Volume Grid to render an Inferno VDB export properly? I’m currently unable to get the fire to render. Smoke renders fine.... It looks like Temp/Fuel is saved in the same smoke channel.
Thanks

[tyFlow]

There's a bug in the latest build where if you have Inferno Recall enabled, and export VDB from the solver and also have "bake display blur" enabled, temperature won't export. You can workaround this for now by disabling "bake display blur" in the Inferno Export operator. This issue should be fixed in the next build.

[chriss3d]

There's a bug in the latest build where if you have Inferno Recall enabled, and export VDB from the solver and also have "bake display blur" enabled, temperature won't export. You can workaround this for now by disabling "bake display blur" in the Inferno Export operator. This issue should be fixed in the next build.

Thanks. It worked in Arnold.  In Vray still Smoke channel is combined with Temp/Fuel

[GreenDK]

Amazing job, Tyson.
Viewport inferno looks amazing and runs great.

[Moley]

Hi There

I'd like to start by saying I absolutely love Tyflow and have from the start.

However, I am feeling a little cheated today. I am working on a project that requires a lot of fire. I have been struggling through it with Phoenix. Slow, but functional.

So I get an email telling me about Tyflow 2.0 with this amazing super fast fire simulation. Amazing! Answer to my prayers. So I break out my credit card and drop 500 bucks on it.

Sure enough it does simulate fire really quickly, but I'm damned if I can figure out how to render it. The only information about rendering is a vague reference to the renderer's VDB container. I have been labouring to get a useful render now for several hours now with Arnold and Vray, but am working in the dark. So if I'm going to meet my deadline I'm going to have to write-off this time spent trying to figure it out and go back to Phoenix.

As far as I can tell there is zero documentation for inferno. This would be fine if this was a tech demo, but for a full release for which you charge $495 this is not acceptable.

Maybe $495 dollars isn't a lot of money to you, but for me it is a really significant investment at a time where I'm just about managing to keep my business afloat.

Can I either have a refund or some actual, meaningful, practical information about how this is supposed to work?

[tyFlow]

Hi Moley,

There is no documentation yet for Inferno, that's correct. It will be released in the coming weeks as I write it.

To render with VRay is pretty simple: use an Export Inferno operator to export your simulation to VDB. Make sure you enable "kelvin" units in the Inferno Export operator, under temperature, because VRay expects units in kelvin. Note: there is a bug in v2.000 where temperature may not expect when "bake temperature blur" is enabled - so you can disable that for now. This bug is fixed and the next build should be out soon, but for now that's a workaround.

Then to render with VRay you can load the sequence into a VRayVolumeGrid container (same as you would for rendering Phoenix when not rendering directly from the Phoenix sim container). Once loaded, ensure you see your channels listed in the VRayVolumeGrid container (likely you'll want at least smoke/temperature, and velocity too if you want to render with motion blur).

From there, you can also click the little [>>] menu icons in your Inferno Display operator next to your color gradient and opacity curve to copy the data into a VRay-friendly format. This gradient/curve can then be pasted into the VRayVolumeGrid container's Volumetric Options window. You may need to tweak things like fire opacity multiplier and such, within the VRayVolumeGrid to get the look you want, but these steps at least will give you a good starting point.

[Moley]

Thank you for getting back to me so quickly. I have managed to get a useable render.

I understand the rush and enthusiasm to drop a release, and what an amazing release it is, but it is worth remembering that a release without meaningful documentation is, for your users, at best frustrating and at worst completely useless.

[SassyMotion]

Hi there,

A quick question:
Is the "instant cache recall" disabled in the free version?
I'm using a floating license, and whenever I don't have it for a short time (server not yet started f.e.), I quickly notice that Inferno only displays the last cached frame.
(Basic Inferno Simulation. Just a box as an emitter)
Is that intended, or do I might have done something wrong with my setup?
With the license it is working just fine.

Thanks in advance

[tyFlow]

It's not only for PRO users, no....

You're saying that cache recall works only once your license activates? That's weird...seems like maybe something else is going on...

[SassyMotion]

It's not only for PRO users, no....

You're saying that cache recall works only once your license activates? That's weird...seems like maybe something else is going on...

Yes. Is the incident reproducible? I saw the following line in the listener:

"Zenith: VRAM usage: [Solver: 53,88 MB | Display 16,05 MB | Total: 69,93 MB] | RAM usage [Cache: 0,00 Bytes] | Steps: 1 | Sparse cubes: 42 | Sparse voxels: 1.38m | Domain voxels: 0.59m"

and noticed that the RAM usage for the Cache is always at 0,00 Bytes.
So a workaround for me was just disabling the playback cache for the moment. The loading might take some time but this way I can see the simulation at every frame and not just the last simulated one in the timeline.
(Just trying some fun stuff for the weekend and can't use PRO because I am currently not at the office to start the server)

[tyFlow]

What GPU does your machine have?

[SassyMotion]

What GPU does your machine have?

Here it's a 1080ti (probably older driver. I should update to the latest one or is this gpu too old? I know that there is no support by Nvidia anymore)
At the office it's a 4060ti (latest driver)

[tyFlow]

Ah, that's why - cache compression isn't supported on 1080 cards because it requires compute capability 7.0 or higher (ie, 2XXX cards or higher). The next build shows a warning in the UI about this.

Your option is to either set compression mode to uncompressed for cache recall, or upgrade your GPU.

[SassyMotion]

Ah, that's why - cache compression isn't supported on 1080 cards because it requires compute capability 7.0 or higher (ie, 2XXX cards or higher). The next build shows a warning in the UI about this.

Your option is to either set compression mode to uncompressed for cache recall, or upgrade your GPU.

Ah thank you very much  That's great to know. I already had my eye on an "affordable" 5080. Alright finally it's time to upgrade.
And thank you for the upcoming warning in the UI for those who, like me, are stuck in time.

Message to exactly those people for the moment: You will find the compression mode in the Inferno Recall Operator -> RAM cache -> All channels: Uncompressed 
It works like a charm while you wait for the gpu delivery to get ready for the next decade 

[BasLeopold]

So is it safe to assume, you cannot render inferno in vray with a tyflow free version at all?.. I only see it in viepowrt IPR, but thats it. (rtx 5090)