noizfactory
Hey all,

Its been a while!

Started playing with bifrost and thought I'll start converting some of my really old SOuP examples to bifrost. Took a stab at "_passNormals_soupVolumeScatter.ma".

Attached the result snapshot and the bifrost graph. Will continue to post examples to this thread as I find some time to convert more examples.

SOuP_attributeTransfer_surfaceToVolumePoints.png 
bifrost_sample_property_graph.png 
Cheers,
Sachin
Quote 0 0
pshipkov
Good stuff.
Keep it coming.
How do you like it so far, the Bifrost i mean ?
Quote 0 0
noizfactory
I'm liking it so far. The examples are a bit sparse and the node names are not immediately obvious. All the nodes that get data seem to be prefixed get_* as well as sample_*. Took me some time to realize that sample_property was what I needed. Also, the nodes seem quite atomic which is great but I think as the product grows we will see more useful compounds straight out of the box.
Quote 0 0
samadoni

might be a silly question but i guess we can have custom icons on the created compound by modifying the json file no?

Cause i don't see myself working with dozens of compounds with all the same icons it would get quickly messy 🙂 

Quote 0 0
noizfactory
Yes, you can. The rebel pack which I’ve used in the above setup can be seen in the screenshot with its own icon.
Quote 0 0
noizfactory
Raycast node attempt. This is with a single point right now. Hit point turns red if its outside the mesh (i.e. it actually didn't hit the mesh). WIP graph shared here.

bifrost_raycast_single_point_hit_test.gif 
Quote 1 0
pshipkov
A good first step.
It sounds like you are back to Maya or something ?
🙂
Quote 0 0
noizfactory
pshipkov wrote:
A good first step.
It sounds like you are back to Maya or something ?
🙂

Just noodling around with biforst ;-)
Quote 0 0
noizfactory
Some more simple tests.

Closest point without using the built-in node (eventual goal is to replicate something like SOuP's attributeTransfer):
bifrost_closest_point.gif
A simple ray line intersection test using strands.
bifrost_ray_line_intersection.gif
Cheers,
Sachin
Quote 1 0
Bruce Lee
Very powerful, can you upload your scene file? I want to learn them. Thanks!
Quote 0 0
pshipkov
Thiswill be a goodexcersise.
I will be looking for your feedback about how you like building more complex logic using visual programming.
Quote 0 0
noizfactory
@Bruce Lee  - closest point test file shared here.

@pshipkov - So far, its kind of fun building the logic but frankly speaking it does feel slower than simply writing code. So, for more complex logic its going to be a bit of a pain. Its also quite easy to get lost in a sea of nodes and graphs than simply go through code. I think we are going to need a lot of compounds to make this less tedious eventually. 
Quote 0 0
pshipkov
I think we are aligned on this one.
VisProg is a great "glue", but not convinced if it is a great fit for building the main components.
Maybe once we have a public API, we can program in the traditional way the big/monolitic/complex modules and use the standard toolkit as the "glue" between them - to combine and augment their functionality for our needs.

On the same note - i am not crazy about where Houdini is going these days either - breaking down established stuff into smaller and smaller chunks in the name of flexibility, despite the fact that the needs of VFX/MoGraph/CG have been, more or less, well defined already. Things like - make smoke, make explosion, make destruction, combine simple objects and move them around for motion graphics, etc. I know that this brief message is oversimplifying it, but in reality the cookbook  has been written.
Once upon a time we were in the exploration phase - how to do things, but today is all about how quickly we can do the same things over and over again.

We obviously need the right granularity of operators that process the data. If too small chunks - it takes longer to string them together and people tend to noodle too much. From project management stand point this is not good.
Too much encapsulation and we loose control and start "fighting the system". Not good either.
To me the golden middle ground was the level of granularity around Houdini 9ish. Since then i have much less joy dealing with it.

Bifrost visprog graph is early stage, so let's see if developers will find the right balance.
I spend 6-8 months with Autodesk and Marcus working on the original designs back then. So far i am quitepleased with the results. Hope they stay on target with the upcoming releases.
Quote 1 0
noizfactory
@pshipkov - Yes, I see what you are saying. This becomes even more evident when you are working with loops and iterations and want to remove/insert something. Its just so much quicker to write that logic in code than noodle with the iteration nodes which are frankly not as straight-forward to use right now.

But a good and much needed step in the right direction for the future of Maya as these days people become more technical and are able to problem solve without relying on a pipeline or software resource. Interesting times ahead.

EDIT - Same goes for writing re-usable modules/functions versus first wiring together reusable compounds.
Quote 0 0
pshipkov
Time is coming where computing will be more or less generalized and made accessible on a whole different level.
We are already experienced that with electronic consumer products - smart phones, e-office suites, etc. - the common stuff, but we are entering the next stage - high-level interfaces providing acceptable level of access to things that used to be for the very smart only.
Bifrost is the latest tech to join the club. There is a tremendous potential for it.
But also, remote (cloud) computing is become a reality and all traditional frameworks that we are familiar with (including Bifrost now) will be (already are) at disadvantage.
Let's see what's ahead. I am very optimistic about the future. 🙂
Quote 0 0