Pype global tools
Working with Pype in Maya
Pype is here to ease you the burden of working on project with lots of collaborators, worrying about naming, setting stuff, browsing through endless directories, loading and exporting and so on. To achieve that, Pype is using concept of being "data driven". This means that what happens when publishing is influenced by data in scene. This can by slightly confusing so let's get to it with few examples.
Publishing models in Maya is pretty straightforward. Create your model as you
need. You need to adhere to specifications of your studio that can be different
between studios and projects but by default your geometry has to be named properly.
cube1_GEO. Geometry needs to have freezed transformations
and must reside under one group, for example
sphere_GEO has frozen transformations.
Now create Model instance from it to let Pype know what in the scene you want to publish. Go Pype → Create... → Model
Asset field is a name of asset you are working on - it should be already filled
with correct name as you've started Maya or switched context to specific asset. You
can edit that field to change it to different asset (but that one must already exists).
Subset field is a name you can decide on. It should describe what kind of data you
have in the model. For example, you can name it
Proxy to indicate that this is
low resolution stuff. See Subset.
By changing subset name you can take advantage of LOD support in Pype. Your asset can contain various resolution defined by different subsets. You can then switch between them very easy using Inventory (Manage). There LODs are conveniently grouped so they don't clutter Inventory view.
Name your subset like
main_LOD1. Important part is that
_LOD1. You can have as many LODs as you need.
Read-only field just under it show final subset name, adding subset field to name of the group you have selected.
Use selection checkbox will use whatever you have selected in Outliner to be
wrapped in Model instance. This is usually what you want. Click on Create button.
You'll notice then after you've created new Model instance, there is new set
in Outliner called after your subset, in our case it is
And that's it, you have your first model ready to publish.
Now save your scene (if you didn't do it already). You will notice that path
in Save dialog is already set to place where scenes related to modeling task on
your asset should reside. As in our case we are working on asset called
Ben and on task modeling, path relative to your project directory will be
project_XY/assets/ben/work/modeling. Let's save our scene as
Now let's publish it. Go Pype → Publish.... You will be presented with following window:
Note that content of this window can differs by your pipeline configuration. For more detail see Publisher.
Items in left column are instances you will be publishing. You can disable them by clicking on square next to them. Green square indicate they are ready for publishing, red means something went wrong either during collection phase or publishing phase. Empty one with gray text is disabled.
See that in this case we are publishing from scene file
Maya model named
modelMain (ben) (next item). Publishing of workfile is
currenly disabled (last item).
Right column lists all tasks that are run during collection, validation, extraction and integration phase. White items are optional and you can disable them by clicking on them.
Lets do dry-run on publishing to see if we pass all validators. Click on flask icon at the bottom. Validators are run. Ideally you will end up with everything green in validator section.
To make things interesting, I intentionally forgot to freeze transformations
sphere_GEO as I know it will trigger validator designed to check just this.
You can see our model is now marked red in left column and in right we have
red box next to
Transform Zero (Freeze) validator.
You can click on arrow next to it to see more details:
From there you can see in Records entry that there is problem with
Some validators have option to fix problem for you or just select objects that
cause trouble. This is the case with our failed validator.
In main overview you can notice little up arrow in a circle next to validator
name. Right click on it and you can see menu item
select invalid. This
will select offending object in Maya.
Fix is easy. Without closing Publisher window we just freeze transformations. Then we need to reset it to make it notice changes we've made. Click on arrow circle button at the bottom and it will reset Publisher to initial state. Run validators again (flask icon) to see if everything is ok.
It should be now. Write some comment if you want and click play icon button when ready.
Publish process will now take its course. Depending on data you are publishing it can take a while. You should end up with everything green and message Finished successfully ... You can now close publisher window.
To check for yourself that model is published, open
Asset Loader - Pype → Load....
There you should see your model, named
Look development in Pype is easy. It helps you with versioning different kinds of shaders and easy switching between them.
Let se how it works.
In this example I have already published model of Buddha. To see how to publish model with Pype see Publishing Model.
First of lets start with empty scene. Now go Pype → Load...
Here I am loading
modelBuddha, its version 1 for asset foo. Just right-click
on it and select Reference (abc). This will load model into scene as alembic.
Now you can close Loader window.
Now you can create whatever look you want. Assign shaders, textures, etc. to model. In my case, I assigned simple Arnolds aiSurfaceShader and changed its color to red.
I am quite happy with it so I want to publish it as my first look.
Select your model in outliner and ho Pype → Create.... From there
select Look. Make sure
use selection checkbox is checked.
Mine subset name is
Main. This will create Look instance with a name lookMain.
Close Creator window.
Now save your scene, give it some sensible name. Next, go Pype → Publish. This process is almost identical as publishing models, only different Validators and other plugins will be used.
This should be painless and cause no trouble so go ahead, click play icon button at the bottom and it will publish your look.
publishing multiple looks
You can reference same model into scene multiple times, change materials on every instance with what you need. Then on every model create Look instance. When
publishing all those Look instances will be published at same time.
Loading looks into models
Now lets see how look are applied. Start new empty scene, load your published model there as before (using Reference (abc)). If you didn't notice until now, there are few yellow icons in left shelf:
Those can be found even in top menu, but that depends on your studio setup.
You are interested now in Look Manager - first item with brush icon. Select your Buddha model and open Look Manager.
This is Look Manager window. Yours would be empty until you click Get All Assets or Get Assets From Selection. You can use later to quick assign looks if you have multiple assets loaded in scene. Click on one of those button now.
You should now see all assets and their subsets loaded in scene, and on right side all applicable published looks.
Select you asset and on the right side right click on
Main look. Apply it.
You notice that Buddha model is now red, materials you've published are now applied to it.
That way you can create looks as you want and version them using Pype.
Setting scene data
Maya settings concerning framerate, resolution and frame range are handled by Pype. If set correctly in Ftrack, Maya will validate you have correct fps on scene save and publishing offering way to fix it for you.
For resolution and frame range, use Pype → Reset Frame Range and Pype → Reset Resolution
Creating rigs with Pype
Creating and publishing rigs with Pype follows similar workflow as with other data types. Create your rig and mark parts of your hierarchy in sets to help Pype validators and extractors to check it and publish it.
Preparing rig for publish
When creating rigs, it is recommended (and it is in fact enforced by validators) to separate bones or driving objects, their controllers and geometry so they are easily managed. Currently Pype doesn't allow to publish model at the same time as its rig so for demonstration purposes, I'll first create simple model for robotic arm, just made out of simple boxes and I'll publish it.
For more information about publishing models, see Publishing models.
Now lets start with empty scene. Load your model - Pype → Load..., right click on it and select Reference (abc).
I've created few bones and their controllers in two separate
controls_GRP. Naming is not important - just adhere to
your naming conventions.
Then I've put everything into
When you've prepared your hierarchy, it's time to create Rig instance in Pype.
Select your whole rig hierarchy and go Pype → Create.... Select Rig.
Set is created in your scene to mark rig parts for export. Notice that it has
two subsets -
out_SET. Put your controls into
and geometry to
out_SET. You should end up with something like this:
Publishing rig is done in same way as publishing everything else. Save your scene and go Pype → Publish. When you run validation you'll mostly run at first into few issues. Although number of them will seem to be intimidating at first, you'll find out they are mostly minor things easily fixed.
Non Duplicate Instance Members (ID) - This will most likely fail because when creating rigs, we usually duplicate few parts of it to reuse them. But duplication will duplicate also ID of original object and Pype needs every object to have unique ID. This is easily fixed by Repair action next to validator name. click on little up arrow on right side of validator name and select Repair form menu.
Joints Hidden - This is enforcing joints (bones) to be hidden for user as animator usually doesn't need to see them and they clutter his viewports. So well behaving rig should have them hidden. Repair action will help here also.
Rig Controllers will check if there are no transforms on unlocked attributes of controllers. This is needed because animator should have ease way to reset rig to it's default position. It also check that those attributes doesn't have any incoming connections from other parts of scene to ensure that published rig doesn't have any missing dependencies.
You can load rig with Loader. Go Pype → Load..., select your rig, right click on it and Reference it.
Set dressing in Maya
Set dressing is term for easily populate complex scenes with individual parts. Pype allows to version and manage those sets.
Publishing Set dress / Layout
Working with Set dresses is very easy. Just load your assets into scene with Loader (Pype → Load...). Populate your scene as you wish, translate each piece to fit your need. When ready, select all imported stuff and go Pype → Create... and select Set Dress or Layout. This will create set containing your selection and marking it for publishing.
set dress vs layout
Currently set dress and layout are functionally identical
Now you can publish is with Pype → Publish.
Loading Set dress / Layout
You can load Set dress / Layout using Loader (Pype → Load...). Select you layout or set dress, right click on it and select Reference Maya Ascii (ma). This will populate your scene with all those models you've put into layout.
Rendering with Pype
Muster is now configured so every user must log in to get authentication support. If Pype founds out this token is missing or expired, it will ask again for credentials.
Creating basic render setup
If you want to submit your render to farm, just follow these simple steps.
Lets start with empty scene. First I'll pull in my favorite Buddha model. Pype → Load..., select model and right+click to pop up context menu. From there just click on Reference (abc).
Next, I want to be sure that I have same frame range as is set on shot I am working on. To do this just Pype → Reset Frame Range. This should set Maya timeline to same values as they are set on shot in Ftrack for example.
I have my time set, so lets create some animation. We'll turn Buddha model around for 50 frames (this is length of my timeline).
Select model, go to first frame, key Y axis rotation, go to last frame, enter 360 to Channel Editor Y rotation, key it and its done. If you are not sure how to do it, you are probably reading wrong documentation.
Now let set up lights, ground and camera. I am lazy so I create Arnolds Skydome light:
Arnold → Lights → Skydome Light. As ground simple Plane will suffice and I'll set
my perspective view as I like and create new camera from it (
CTRL+SHIFT+C) and rename
One last thing, I'll assign basic aiSurfaceShader to my Buddha and do some little tweaks on it.
Prepare scene for submission
Ok, now we have working simple scene. We need to do few more things to be able to submit it. First, we need to create RenderGlobals instance. This is special kind of instance that is not designed to hold other objects or mark them as publishable. This instance holds data needed for render farm submission.
Lets create it. Go Pype → Create.... There select Render Globals from list. Now, no matter if you use Deadline or Muster, Pype will try to connect to render farm and fetch machine pool list.
This might fail on Muster in the event that you have expired authentication token. In that case, you'll be presented with login window. Nothing will be created in the scene until
you log in again and do create Render Globals again.
So my scene now looks like this:
My scene looks like this:
renderglobalsMain in Attributes Editor:
Few options that needs explaining:
Primary Pool- here is list of pool fetched from server you can select from.
Suspend publish Job- job sent to farm will not start render automatically but is in waiting state.
Extend Frames- if checked it will add new frames to previous render, so you can extend previous image sequence.
Override Existing Frame- will overwrite file in destination if they exists
Use Legacy Render Layers- this currently does nothing as Legacy Render Layers are not supported and will probably never be.
Priorityis priority of job on farm
Frames Per Taskis number of sequence division between individual tasks (chunks) making one job on farm.
If you now try to publish render, you notice that Publish seems to be totally ignorant about the fact you want to submit something to render on farm.
This is because our scene is missing defined render layers. Their presence is signal for publish to consider it for submission.
As legacy layers are not supported, lets open Render Setup window.
I've created new layer, named it
Main and simple Collection including all in the scene.
Now if you run publish, you notice there is in right column new item called
Render Layers and in it there is our new layer
Main (foo) [1-10]. First part is
layer name, second (
(foo)) is asset name and rest is frame range.
You see I already tried to run publish but was stopped by few errors. Lets go through them one by one just to see what we need to set up further in scene for successful publish.
Render Single Camera and No Default Cameras Renderable can be resolved by
simple change in Main layer render settings. As I created new camera from
persp, Maya added it to renderable cameras for that layer. All I have to do is
just remove the
persp camera from render settings.
This leaves me only with Render Settings error. If I click on it to see details, I see it has problem with animation not being enabled and with file name prefix.
Go to Render Settings, select your render layer and in Common tab change
in File Output
Frame/Animation ext to whatever you want, just not Single Frame.
Set Frame Range
Start frame and
End frame according your needs and at the top in
File Output set
File name prefix to:
for default Arnold, and:
Doing Pype → Reset Resolution will set correct resolution on camera.
Scene is now ready for submission and should publish without errors.
what happens when I publish my render scene
When publishing is finished, job is created on farm. This job has one more dependent job connected to itself.
When render is finished, this other job triggers in and run publish again, but this time it is publishing rendered image sequence and creating quicktime movie for preview from it.
Publishing Render Setups
Pype can publish whole Render Settings setup. You can then version in and load it to any Maya scene. This helps TDs to distribute per asset/shots render settings for Maya.
To publish render settings, go Pype → Create... and select Render Setup Preset.
In your scene will appear set
rendersetup<subset>. This one has no settings, only its presence
in scene will trigger publishing of render settings.
When you publish scene, current settings in Render Settings will be serialized to json file.
Loading Render Setups
In any scene, you can load published render settings with Pype → Load.... Select your published render setup settings, right+click on it and select Load RenderSetup template.
This will load and parse json file and apply all setting there to your Render Setting.
This will overwrite all setting you already have.
Pype supports creating review video for almost any type of data you want to publish. What we call review video is actually playblast or capture (depending on terminology you are familiar with) made from pre-defined camera in scene. This is very useful in cases where you want to add turntable preview of your model for example. But it can be used to generate preview for animation, simulations, and so on.
Setting scene for review extraction
Lets see how review publishing works on simple scene. We will publish model with turntable preview video.
I'll be using Stanford University dragon model. Start with empty scene. Create your model, import it or load from Pype. I'll just import model as OBJ file.
After we have our model in, we need to set everything to be able to publish it as model - for detail see Publishing models.
To recap - freeze transforms, rename it to
dragon_GEO and put it into group
dragon_GRP. Then select this group and Pype → Create... and choose Model.
Now, lets create camera we need to generate turntable video. I prefer to animate camera itself and not model because all animation keys will be associated with camera and not model we want to publish.
I've created camera, named it
reviewCamera and parent it under
locator. I set my timeline to 50 frames, key
reviewRotation_LOC Y axis on frame
1 to 0 and on frame 50 to 360. I've also set animation curve between those two keys
To mark camera to be used for review, select camera
reviewCamera and go Pype → Create...
and choose Review.
This will create set
review<subset> including selected camera. You can set few options
on this set to control review video generation:
Active- control on/off state
Frame Start- starting frame for review
Frame End- end frame for review
Handles- number of handle frame before and after
Step- number of steps
This is my scene:
* note that I had to fix UVs and normals on Stanford dragon model as it wouldn't pass model validators
Publishing model with review
You can now publish your model and generate review video. Go Pype → Publish..., validate if you will, and publish it. During publishing, Maya will create playblast for whole frame range you've specified, then it will pass those frames to ffmpeg. That will create video file, pass it to another extractor creating burnins in it and finally uploading this video to ftrack with your model (or other type) published version. All parts of this process - like what burnins, what type of video file, settings for Maya playblast - can be customized by your TDs. For more information about customizing review process refer to admin section.
Working with Yeti in Pype
Pype can work with Yeti in two data modes. It can handle Yeti caches and Yeti rigs.
Creating and publishing Yeti caches
Let start by creating simple Yeti setup, just one object and Yeti node. Open new empty scene in Maya and create sphere. Then select sphere and go Yeti → Create Yeti Node on Mesh Open Yeti node graph Yeti → Open Graph Editor and create setup like this:
It doesn't matter what setting you use now, just select proper shape in first
Import node. Select your Yeti node and create Yeti Cache instance - Pype → Create...
and select Yeti Cache. Leave
Use selection checked. You should end up with this setup:
You can see there is
yeticacheDefault set. Instead of Default it could be named with
whatever name you've entered in
subset field during instance creation.
We are almost ready for publishing cache. You can check basic settings by selecting Yeti cache set and opening Extra attributes in Maya Attribute Editor.
Those attributes there are self-explanatory, but:
Prerollis number of frames simulation will run before cache frames are stored. This is usefull to "steady" simulation for example.
Frame Startfrom what frame we start to store cache files
Frame Endto what frame we are storing cache files
Sampleshow many time samples we take during caching
You can now publish Yeti cache as any other types. Pype → Publish. It will
create sequence of
.fur files and
.fursettings metadata file with Yeti node
Loading Yeti caches
You can load Yeti cache by Pype → Load .... Select your cache, right+click on it and select Load Yeti cache. This will create Yeti node in scene and set its cache path to point to your published cache files. Note that this Yeti node will be named with same name as the one you've used to publish cache. Also notice that when you open graph on this Yeti node, all nodes are as they were in publishing node.
Creating and publishing Yeti Rig
Yeti Rigs are working in similar way as caches, but are more complex and they deal with other data used by Yeti, like geometry and textures.
Let's start by loading into new scene some model. I've loaded my Buddha model.
Create select model mesh, create Yeti node - Yeti → Create Yeti Node on Mesh and setup similar Yeti graph as in cache example above.
Then select this Yeti node (mine is called with default name
create Yeti Rig instance - Pype → Create... and select Yeti Cache.
Use selection checked.
Last step is to add our model geometry to rig instance, so middle+drag its
yetiRigDefault set representing rig instance.
Note that its name can differ and is based on your subset name.
Save your scene and ready for publishing our new simple Yeti Rig!
Go to publish Pype → Publish and run. This will publish rig with its geometry
.ma scene, save Yeti node settings and export one frame of Yeti cache from
the beginning of your timeline. It will also collect all textures used in Yeti
node, copy them to publish folder
resource directory and set Image search path
of published node to this location.
Collect Yeti Cache failure
If you encounter Collect Yeti Cache failure during collecting phase, and the error is like
No object matches name: pgYetiMaya1Shape.cbId
then it is probably caused by scene not being saved before publishing.
Loading Yeti Rig
You can load published Yeti Rigs as any other thing in Pype - Pype → Load ..., select you Yeti rig and right+click on it. In context menu you should see Load Yeti Cache and Load Yeti Rig items (among others). First one will load that one frame cache. The other one will load whole rig.
Notice that although we put only geometry into
input_SET, whole hierarchy was
pulled inside also. This allows you to store complex scene element along Yeti
auto-connecting rig mesh to existing one
If you select some objects before loading rig it will try to find shapes
under selected hierarchies and match them with shapes loaded with rig (published
input_SET). This mechanism uses cbId attribute on those shapes.
If match is found shapes are connected using their
outMesh. Thus you can easily connect existing animation to loaded rig.