Bifrost graph - Really trying but it's not happening

[Chris Marshall]

OK So is Bifrost graph years away from getting close to ICE or am I missing something?

I really am trying to figure it out, but either I'm doing something wrong or it's very very limited.

Anyone got anywhere with it?

Thanks

Chris

[Jonah Friedman]

Hi Chris, I'm the Product Owner of the Bifrost graph. I'd be happy to answer any questions about it. What specifically are you trying to do?

The graph is different from ICE in several ways, but it's very open and very powerful. 

------
Softimage Mailing List.
To unsubscribe, send a mail to softimag...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

[Ben Beckett]

Hi 

Why do you just make it work like Naiad that worked perfectly 

Please 

Something is defo wrong now

[Ben Beckett]

Houdini works why waste your time

[Chris Marshall]

Oh Hi Jonah,

Well ok as I’ve been quite a fan of ice, and it took me a while to warm to it but I can totally see it’s logic, all I’m seeing so far with Bifrost graph are limitations, compared to Ice. So putting all that to one side, I’m happy to start looking at it again afresh. What would help is a document that describes, with simple examples, what the structural differences are between the two systems and how bifrost graph is so open and powerful. ICE was / is a very comprehensive and what I’d describe as a mega flexible system, once you know what you’re doing. Some would probably argue it was too flexible for artists as it required some coding logic to really push it. 

To be honest, with Bifrost graph I’ve dipped in a few times and got nowhere. A basic fundamental description of how the system is supposed to used would help, and why this might be different to ICE. 

Thanks 

Chris

On Tuesday, 18 February 2020, Jonah Friedman <jon...@gmail.com> wrote:

Hi Chris, I'm the Product Owner of the Bifrost graph. I'd be happy to answer any questions about it. What specifically are you trying to do?

The graph is different from ICE in several ways, but it's very open and very powerful. 

On Tue, Feb 18, 2020 at 4:16 AM Chris Marshall <chrisma...@gmail.com> wrote:

OK So is Bifrost graph years away from getting close to ICE or am I missing something?

I really am trying to figure it out, but either I'm doing something wrong or it's very very limited.

Anyone got anywhere with it?

Thanks

Chris

------
Softimage Mailing List.

To unsubscribe, send a mail to softimage-request@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

--

Chris Marshall

Mint Motion Limited

029 20 37 27 57

07730 533 115

www.mintmotion.co.uk

www.dot3d.com

[Jonah Friedman]

Hi Chris,

I was also an avid ICE user for many years, and following that Fabric Engine. Being exposed to ICE was a turning point in my career in fact. 

I'm going to try to illustrate the difference with a simple graph that contains very familiar pieces that I can use to highlight the difference.This graph deforms a mesh with an f-curve on the x axis, applied to the Y position. The green grid is before, the white is after. 

 

The graph is doing familiar operations to ICE users with familiar names: get_point_position, converting the vector into scalars, applying an f-curve to one, and rebuilding the vector and setting the point positions all have essentially the same names and meanings in this graph. However the graph is different. (Note this is an internal build, if you notice discrepancies in the screenshot). 

To highlight the differences:

1. We can't just get point positions, we need to bring in a mesh first and get positions from that. Maya does not have the operator stack that XSI had, and so the graph is its own entity, and not a part of the mesh. Instead, the mesh is just an input. 

2. The part where we adjust the positions is operating on the per-point array of positions. However in the graph, this is just flowing data. Those three nodes in the center are processing the arrays. Ports shaped like hats instead of squares signify arrays. The graph is not operating in the context of per-point, but instead is just processing whole arrays in one shot. 

3. "set point position" needs a geometry to set point positions on. We can't set point positions on "this", we have to set them on data flowing in the graph, and there is no "this". The mesh is flowing through the graph on the purple connection. 

4. In the end we output the shape, and it appears in Maya as a "bif shape". This can be converted back to Maya geometry if you like, but is also renderable if needed. You could set up a Bifrost graph as a more-or-less traditional Maya deformer setup, but it'd require a bit more setup. (We're improving this too with improvements to the Maya integration). 

The fact that this mesh is flowing through the graph just like any other kind of data, buildable and modifiable, as pure data flow is the key difference from ICE to understand. Whole geometries can be created in the graph in multi-stage processes. Create_mesh_sphere outputs a sphere mesh and is programmed in the mesh, and that sphere can be used downstream for any purpose. 

---

Visualizing data: Watchpoints are not the same as the similar feature in ICE, the debug drawing. Instead right now to do visual diagnostics, you can use the "scope" nodes whose job it is to visualize data. For example, see the image below of "volume_scope" which visualizes volumes. This is working by generating strands in the graph to visualize the tile tree boxes (left) and flowlines in the volume (right). We're working on improving this as well.

To see the data in whole in numeric / text form, "dump_object" is your best bet which sends the data to a text file. In there you can see all the data that defines the geometries except volumes, which are still opaque for now.   

I'd also like to say it's a major concern of mine that Bifrost is not the easiest to learn and understand. It's not something we have a silver bullet for - some of the great things that ICE did are tricky for us and visual programming is not the easiest thing even in the case of ICE. We do however have multiple efforts underway to improve the situation. One nice example that's in right now that you may or may not have noticed - most nodes we ship with are documented in-app, in an info tab. 

Some other info: 

Docs: https://help.autodesk.com/view/BIFROST/ENU/

Release notes from our last release, released yesterday: https://help.autodesk.com/view/BIFROST/ENU/?guid=Bifrost_ReleaseNotes_release_notes_release_notes_2_0_5_0_html  

Tutorials by Matt Chan: https://area.autodesk.com/tutorials/series/getting-started-in-bifrost/

On Tue, Feb 18, 2020 at 3:29 PM Chris Marshall <chrisma...@gmail.com> wrote:

Oh Hi Jonah,

Well ok as I’ve been quite a fan of ice, and it took me a while to warm to it but I can totally see it’s logic, all I’m seeing so far with Bifrost graph are limitations, compared to Ice. So putting all that to one side, I’m happy to start looking at it again afresh. What would help is a document that describes, with simple examples, what the structural differences are between the two systems and how bifrost graph is so open and powerful. ICE was / is a very comprehensive and what I’d describe as a mega flexible system, once you know what you’re doing. Some would probably argue it was too flexible for artists as it required some coding logic to really push it. 

To be honest, with Bifrost graph I’ve dipped in a few times and got nowhere. A basic fundamental description of how the system is supposed to used would help, and why this might be different to ICE. 

Thanks 

Chris

On Tuesday, 18 February 2020, Jonah Friedman <jon...@gmail.com> wrote:

Hi Chris, I'm the Product Owner of the Bifrost graph. I'd be happy to answer any questions about it. What specifically are you trying to do?

The graph is different from ICE in several ways, but it's very open and very powerful. 

On Tue, Feb 18, 2020 at 4:16 AM Chris Marshall <chrisma...@gmail.com> wrote:

OK So is Bifrost graph years away from getting close to ICE or am I missing something?

I really am trying to figure it out, but either I'm doing something wrong or it's very very limited.

Anyone got anywhere with it?

Thanks

Chris

------
Softimage Mailing List.

To unsubscribe, send a mail to softimag...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

--

Chris Marshall

Mint Motion Limited

029 20 37 27 57

07730 533 115

www.mintmotion.co.uk

www.dot3d.com

------
Softimage Mailing List.

To unsubscribe, send a mail to softimag...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

[Busty kelp]

Hi Jonah, 

If you would like I’m happy to be paid By Autodesk to learn Bifrost in a public way!

I did that with ICE (although unpaid) and posted videos as I went, which became very popular and I’d consider myself an ICE expert now. 

People like watching other people learn as it’s more casual and less intimidating than watching an expert go through the motions.

Paul 

Sent from my iPad

On 19 Feb 2020, at 17:57, Jonah Friedman <jon...@gmail.com> wrote:



Hi Chris,

I was also an avid ICE user for many years, and following that Fabric Engine. Being exposed to ICE was a turning point in my career in fact. 

I'm going to try to illustrate the difference with a simple graph that contains very familiar pieces that I can use to highlight the difference.This graph deforms a mesh with an f-curve on the x axis, applied to the Y position. The green grid is before, the white is after. 

 

<image.png>

The graph is doing familiar operations to ICE users with familiar names: get_point_position, converting the vector into scalars, applying an f-curve to one, and rebuilding the vector and setting the point positions all have essentially the same names and meanings in this graph. However the graph is different. (Note this is an internal build, if you notice discrepancies in the screenshot). 

To highlight the differences:

1. We can't just get point positions, we need to bring in a mesh first and get positions from that. Maya does not have the operator stack that XSI had, and so the graph is its own entity, and not a part of the mesh. Instead, the mesh is just an input. 

2. The part where we adjust the positions is operating on the per-point array of positions. However in the graph, this is just flowing data. Those three nodes in the center are processing the arrays. Ports shaped like hats instead of squares signify arrays. The graph is not operating in the context of per-point, but instead is just processing whole arrays in one shot. 

3. "set point position" needs a geometry to set point positions on. We can't set point positions on "this", we have to set them on data flowing in the graph, and there is no "this". The mesh is flowing through the graph on the purple connection. 

4. In the end we output the shape, and it appears in Maya as a "bif shape". This can be converted back to Maya geometry if you like, but is also renderable if needed. You could set up a Bifrost graph as a more-or-less traditional Maya deformer setup, but it'd require a bit more setup. (We're improving this too with improvements to the Maya integration). 

The fact that this mesh is flowing through the graph just like any other kind of data, buildable and modifiable, as pure data flow is the key difference from ICE to understand. Whole geometries can be created in the graph in multi-stage processes. Create_mesh_sphere outputs a sphere mesh and is programmed in the mesh, and that sphere can be used downstream for any purpose. 

---

Visualizing data: Watchpoints are not the same as the similar feature in ICE, the debug drawing. Instead right now to do visual diagnostics, you can use the "scope" nodes whose job it is to visualize data. For example, see the image below of "volume_scope" which visualizes volumes. This is working by generating strands in the graph to visualize the tile tree boxes (left) and flowlines in the volume (right). We're working on improving this as well.

<image.png>

To see the data in whole in numeric / text form, "dump_object" is your best bet which sends the data to a text file. In there you can see all the data that defines the geometries except volumes, which are still opaque for now.   

I'd also like to say it's a major concern of mine that Bifrost is not the easiest to learn and understand. It's not something we have a silver bullet for - some of the great things that ICE did are tricky for us and visual programming is not the easiest thing even in the case of ICE. We do however have multiple efforts underway to improve the situation. One nice example that's in right now that you may or may not have noticed - most nodes we ship with are documented in-app, in an info tab. 

[Gerbrand Nel]

I get the feeling ,allot of those videos will be kinda like the one where you tried ICE kinematics for the first time.

I still think that is one of the funniest things on the interwebs to date.

[Thomas Volkmann]

Videos from Paul might actually convince me to take a look at Maya. I loved the ICE videos!

[pa...@bustykelp.com]

Ha.. Yes probably.

 

As I recall, I was a bit Gung-Ho about how easy a 2bone IK chain would be, so didn’t even think to check before videoing myself going through explaining it. So I’ll not be doing that again.

I really miss doing those ICE videos though.

 

Paul

------

Softimage Mailing List.
To unsubscribe, send a mail to softimag...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

--

Chris Marshall

Mint Motion Limited

029 20 37 27 57

07730 533 115

www.mintmotion.co.uk

www.dot3d.com

 

------

Softimage Mailing List.
To unsubscribe, send a mail to softimag...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

------

Softimage Mailing List.
To unsubscribe, send a mail to softimag...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

------

Softimage Mailing List.

To unsubscribe, send a mail to softimag...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

[pa...@bustykelp.com]

Thanks Thomas. I loved doing them

 

From: softimag...@listproc.autodesk.com <softimag...@listproc.autodesk.com> On Behalf Of Thomas Volkmann
Sent: 20 February 2020 11:40
To: Official Softimage Users Mailing List. https://groups.google.com/forum/#!forum/xsi_list <soft...@listproc.autodesk.com>
Subject: Re: Bifrost graph - Really trying but it's not happening

 

Videos from Paul might actually convince me to take a look at Maya. I loved the ICE videos!

 

------

Softimage Mailing List.
To unsubscribe, send a mail to softimag...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

--

Chris Marshall

Mint Motion Limited

029 20 37 27 57

07730 533 115

www.mintmotion.co.uk

www.dot3d.com

 

------

Softimage Mailing List.
To unsubscribe, send a mail to softimag...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

------

Softimage Mailing List.
To unsubscribe, send a mail to softimag...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

------

Softimage Mailing List.

To unsubscribe, send a mail to softimag...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

------
Softimage Mailing List.
To unsubscribe, send a mail to softimag...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

[Chris Marshall]

Thanks Jonah

I can’t see the images on my phone and I’m out of the office for a couple of days but I’ll definitely follow this through and provide feedback on what I find. 

Thanks for putting this together.

Chris

To unsubscribe, send a mail to softimage-request@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

--

Chris Marshall

Mint Motion Limited

029 20 37 27 57

07730 533 115

www.mintmotion.co.uk

www.dot3d.com

------
Softimage Mailing List.

To unsubscribe, send a mail to softimage-request@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

[adrian wyer]

i second that, i'd be more inclined to dip into bifrost if i could watch Paul learning it on my behalf!

 

a

 

Adrian Wyer
Fluid Pictures
4th Floor
4 Bath Place
Rivington Street
London
EC2A 3DR 
++44(0) 207 684 5575 

adria...@fluid-pictures.com

www.fluid-pictures.com

 

Fluid Pictures Limited is registered in England and Wales.
Company number:5657815
VAT number: 872 6893 71

---

From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Thomas Volkmann
Sent: 20 February 2020 11:41
To: Official Softimage Users Mailing List. https://groups.google.com/forum/#!forum/xsi_list
Subject: Re: Bifrost graph - Really trying but it's not happening

 

Videos from Paul might actually convince me to take a look at Maya. I loved the ICE videos!

[Olivier Jeannel]

Nobody of you tried Houdini ?

[Thomas Volkmann]

I love Houdini, but unfortunately my Boss doesn't want to spend money on it and we have an existing Maya Pipeline...  atm I'm only working in SublimeText anyway....

[Olivier Jeannel]

Then quit that job from the past ;)

[Chris Marshall]

I’d love to learn Houdini but I can’t justify the price unfortunately. 

On Thursday, 20 February 2020, Olivier Jeannel <facial...@gmail.com> wrote:

Then quit that job from the past ;)

Le jeu. 20 févr. 2020 à 17:54, Thomas Volkmann <listm...@gmail.com> a écrit :

I love Houdini, but unfortunately my Boss doesn't want to spend money on it and we have an existing Maya Pipeline...  atm I'm only working in SublimeText anyway....

On Thu, Feb 20, 2020 at 5:43 PM Olivier Jeannel <facial...@gmail.com> wrote:

Nobody of you tried Houdini ?

To unsubscribe, send a mail to softimage-request@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

--

Chris Marshall

Mint Motion Limited

029 20 37 27 57

07730 533 115

www.mintmotion.co.uk

www.dot3d.com

 

------
Softimage Mailing List.

To unsubscribe, send a mail to softimage-request@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

------
Softimage Mailing List.

To unsubscribe, send a mail to softimage-request@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

------

Softimage Mailing List.

To unsubscribe, send a mail to softimage-request@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

------
Softimage Mailing List.
To unsubscribe, send a mail to softimage-request@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

------
Softimage Mailing List.
To unsubscribe, send a mail to softimage-request@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

------
Softimage Mailing List.

To unsubscribe, send a mail to softimage-request@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

------
Softimage Mailing List.

To unsubscribe, send a mail to softimage-request@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

[Jordi Bares]

Houdini is actually CHEAPER than Maya, check these

HOUDINI CORE

Maya Subscription       = £1,782.00

Houdini Core Rental     = £1,544.00

Now let’s add the “extras” for any Maya operation so we can compare apples with apples.

Krakatoa Standar bundle = £1,220

Ziva VFX Studio 1 year  = £1,394

Atoms Crowds for Maya   = £2,800  (This is the cheapest I have found -Golaem, Massive for Maya or Miarmy)

And now let’s look at the totals versus HoudiniFX

HOUDINI FX

Maya + Plugins          = £7,196

Houdini FX Rental       = £3,867

So, Maya may be the right tool for the job, it may be artist driven decision, etc... but please let’s look at the facts.

I hope it makes sense

jb

PS. BTW, learning Houdini is free, just download the Apprentice version.

To unsubscribe, send a mail to softimag...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

[J R Sanchez]

To be fair there is an indie license now which I have. Just $250 for Maya for a year.

https://area.autodesk.com/maya-indie/?fbclid=IwAR11Fa-dzpLmuBLfkeAzu5NmZbo5G7hq2AdwXhVL7DAt2CBvG-UPA91cxxM

https://sanchezfilms.wixsite.com/rebel

[Steven Caron]

Just trimming... don't mind me.

On Fri, Feb 21, 2020 at 11:18 AM J R Sanchez <youngup...@gmail.com> wrote:

To be fair there is an indie license now which I have. Just $250 for Maya for a year.

https://area.autodesk.com/maya-indie/?fbclid=IwAR11Fa-dzpLmuBLfkeAzu5NmZbo5G7hq2AdwXhVL7DAt2CBvG-UPA91cxxM

https://sanchezfilms.wixsite.com/rebel

On Fri, Feb 21, 2020 at 5:32 AM Jordi Bares <jordi...@gmail.com> wrote:

Houdini is actually CHEAPER than Maya, check these

HOUDINI CORE

Maya Subscription       = £1,782.00

Houdini Core Rental     = £1,544.00

Now let’s add the “extras” for any Maya operation so we can compare apples with apples.

Krakatoa Standar bundle = £1,220

Ziva VFX Studio 1 year  = £1,394

Atoms Crowds for Maya   = £2,800  (This is the cheapest I have found -Golaem, Massive for Maya or Miarmy)

And now let’s look at the totals versus HoudiniFX

HOUDINI FX

Maya + Plugins          = £7,196

Houdini FX Rental       = £3,867

So, Maya may be the right tool for the job, it may be artist driven decision, etc... but please let’s look at the facts.

I hope it makes sense

jb

PS. BTW, learning Houdini is free, just download the Apprentice version.

[Jordi Bares]

Considering the limitations of the indie license in Maya I would not put that as an example… for example, with Maya you can only have 1 license per studio, with Houdini you have 3 AND these are the full toolset.

Not only that, the Maya indie program seems to be only available for Australia, Canada, New Zealand, the UK or the US… which is the reason I didn’t even mentioned on my post… it is plain  ridiculous.

jb

------
Softimage Mailing List.
To unsubscribe, send a mail to softimag...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.

[Chris Marshall]

I just knew you were going to do that!

--

Chris Marshall

Mint Motion Limited

029 20 37 27 57

07730 533 115

www.mintmotion.co.uk

www.dot3d.com

[Matt Lind]

Meanwhile with Houdini the files get saved in a proprietary file format
which are not compatible with other Houdini licenses. You only get a single
opportunity (lifetime) to convert your files, and that opportunity requires
sending your data to side FX to have it done manually by their staff. If
you're in that position, then it's likely you'll still be generating new
content while Side FX works on the transition making for a headache in file
management.

Not exactly practical, Jordi.

Matt



Date: Sat, 22 Feb 2020 19:22:33 +0000
From: Jordi Bares <jordi...@gmail.com>

Subject: Re: Bifrost graph - Really trying but it's not happening

To: "Official Softimage Users Mailing List.

Considering the limitations of the indie license in Maya I would not put

that as an example? for example, with Maya you can only have 1 license per

studio, with Houdini you have 3 AND these are the full toolset.

Not only that, the Maya indie program seems to be only available for

Australia, Canada, New Zealand, the UK or the US? which is the reason I
didn?t even mentioned on my post? it is plain ridiculous.

jb

[Olivier Jeannel]

Hahaha these "limitations" are easily circumvented...

[J R Sanchez]

Since I live in the USA and I am only one person. I am happy. (happy is not really the right word when having to use Maya)  However I will use ICE before I use Bifrost..... at least for now.

https://sanchezfilms.wixsite.com/rebel

[Ben Barker]

Thanks for that write up Jonah.
Right now my biggest issues with bifrost are the plug-in load time and the graph compile time.
Compile time I could eat since it only affects me, but for now I disable bifrost plugin auto load just because it speeds up maya boot for our animators.
I have a similar issue with Arnold but maya didn't like disabling that.
Since it's rigging for games it's not a gigantic loss but I would like to have it as an option to get compiled plug-ins out of the pipeline if need be, and speeding up those things would really help.