Maya2017 whatWorks/whatNot


After 5 years, the nodeEditor still does not support containers.
This is such a big joke, I don`t know whats going on with Autodesk.

The hypergraph still does not remember node positions inside containers
(well, he did until back 2 years)

So, dear Autodesk, how am I supposed to graphically work with my node networks?



Hotkey Editor reverts hotkeys to Maya defaults on each restart, so you have to open it each time and choose ‘User’ to get them back.

Auto Frame in Graph Editor is not working.

I’ve submitted bugs for both of these last night.


A workaround for mantain the hotkey is assign they to a workspace.


line 1: No module named PySide #



SCREAMING ABOUT IT is probably not going to help anything.


Yep, Maya2017 upgraded to PySide2 and Qt5.6.1. Please “import PySide2”.
It means you need to upgrade your python scripts in case you are using pyside.
Find more details here:

Let me know if you have more questions about PySide2 :slight_smile:




is it compatable with

import MayaQWidgetDockableMixin


File “C:\Program Files\Autodesk\Maya2017\Python\lib\site-packages\maya\app\general\”, line 307, in setDockableParameters

workspaceControlName = cmds.workspaceControl(workspaceControlName, label=self.windowTitle(), retain=retain, loadImmediately=True, floating=True, initialWidth=width, initialHeight=height, requiredPlugin=plugins, requiredControl=controls)

RuntimeError: Object’s name ‘WorkspaceControl’ is not unique.


I think so.I can import it:

from import MayaQWidgetDockableMixin


The bad window mode on the sculpt dialog, etc. is fixed (in 2016.5 you couldn’t maximize, minimize, etc. them), which is good since they’ve applied this window type to a lot more of the UI as part of the panel feature. HW2.0 rendering is working for me again, too.

The panel UI is definitely a big help. Maya’s UI was ancient and a real headache on multiple monitors. Now, you can maximize a panel on each monitor and then arrange different panels in each. You can also place less-used windows in tabs and then minimize the panel of tabs. (For example, I often don’t use the graph editor much, but I still want it quickly available when I need it, so I can just keep it in a tab and it’s available at one click without taking much screen space.) It seems reasonably sane as far as getting panels to drop into the place you want them to go (a lot of panel UIs are really bad at this). Side panels (eg. the AE/CB windows) behave a bit oddly and there are some other small issues, but it’s a big step forward.

I need to check whether minimized tabs take resources (eg. having a UV editor open tends to cause heavier texture loads and can stop GPU override from working–not sure if it knows to disable the tab when it isn’t visible). It would be nice to just have everything I need loaded as quick tabs that I can open or close. I don’t want the wholesale layout changes that “workspaces” cause.

Not a ton of bugfixes for 2016.5 issues so far. The channel box dragging bugs and weird outliner selection after filtering are still there. The shape editor doesn’t show keyframes for blendshape deformers that are in character sets. Undo is still laggy due to problems with MASH (unload it if you aren’t using it). I still haven’t found the bugfix list for this release…

If you’re using “include hierarchy” in render setups, note that it’s gone in 2017 and replaced with collection hierarchies, but the auto-conversion isn’t perfect. Check and double-check to make sure that your render setups are including the nodes you expect them to be. For example, 2016.5 include hierarchy would walk through character sets, which was usually annoying and it’s much better that it doesn’t now, but if your scene expected it for some reason it’ll be broken now.

shadingEngine nodes now appear in the outliner. I don’t know if this is just some setting that got reset or an actual change, but this is good–I always had to jump hoops to access them before since they would never show up in the outliner.

I’m not thrilled about them suddenly changing pyside and giving everyone a bunch of new work to do. They should really include both and give people a release to transition, not just suddenly break compatibility. Hoping the transition is nothing more than changing import statements, but it’s QT, so headaches are what I’m expecting.


Poking at the time editor:

It seems to only let me place keys on a layer on top of the track. Animation layers are important, but other times you’ll just want to edit the clip, and it seems like you can’t do that yet. You can edit keys in the graph editor, but not normally in the viewport. You also can’t drag keyframes on clips around in the time editor, only keyframes on layers, and I can’t find any way to add more attributes to an existing clip.

(My guess is that this was intended to handle editing clips, but was limited to importing clips from other files and only editing with layers for this release. Keyframing on clips is actually in the scripts but commented out, so hopefully it’ll get better at this in the future. If you’re working with mocap data maybe it’s okay to not edit the base data, but for keyframe animation you’ll really want to be able to edit the clips.)

I can’t seem to get pose creation on a track to work: if I move something in the middle of a track and create pose, a pose marker appears in the track, but it doesn’t save and the changes are lost when I move the time slider. I also can’t seem to see inside poses at all: they can’t be expanded like clips and their attributes and key values don’t seem to appear anywhere. I have no idea how to edit a pose after creating it. From the documentation it seems like it should interpolate between poses, but this doesn’t seem to happen and it just steps from one to the next.

The weight field seems to randomly become enabled and disabled. Just soloing and unsoloing a layer will make it become greyed out when it wasn’t before. Not sure what’s happening there.

Ghosting doesn’t really seem to work: if I create a pose track and ghost that it’ll work, but ghosting a regular track won’t, even when just keying a simple mesh. I’ve never had much luck with ghosting in Maya, but it would be really nice if it would work.

You can solo a track, but I haven’t found a way to solo a single clip inside a track, eg. to see a single clip without it being blended into another clip.

Overall, it’s looking pretty useful, and hopefully it’ll get better at editing clips. It’s definitely looking like an improvement from the trax editor. I’ll keep experimenting with it but probably won’t try to use it for much until I see how it irons out through a SP or two, and I’m curious to hear other people’s thoughts.


hypergraph thing was filed as bug for 2016.5


Maya 2017 won’t load my Hotkeys when i start maya, i have to manually load them every time i start maya .Work around or fix Please.


see Post #4


Repost from the 2017 thread.

  • Hotbox doesn’t remember pane specific menus between maya sessions (if they’re on or off)
  • Viewport2 still has major issues, currently with a simple 4 polygon cube, the framerate can go above 100fps (on a K-series quadro) to .2fps for some reason just by tumbling around. It drags really badly. Using latest WQHL drivers. Maya 2016.5/2016 works fine.
  • Hotbox spawns top left of the screen and jumps to the middle, minor but noticeable
  • Hotkeys keep reverting back to the maya default set instead of remembering user hotkeys (apparently known bug)
  • Major selection issues unless “use camera based selection” is turned on in preferences. With scenes that are set inside most of the time when trying to select an object, Maya will select the biggest object instead (ie floor or ceiling). Have been using Maya since 6 and this never used to happen :frowning:
  • In wireframe mode, “camera based selection” should have no effect but it does.
  • Maya Render View always pops up when using 3rd party render framebuffer…although this time instead of being able to minimise it and leave it there (2015,2016,2016.5etc…) it floats and becomes a giant blank window.
  • Workspaces aren’t remembered between Maya sessions (I mean seriously???)
  • Tear off menus dont work with expanded menus. Tear off a menu so when the menu floats, expanded menus don’t work (ie the options with the arrow > indicator) They appear for a second and then disappear before you can choose an option.


If you have multiple renderable render setups, rendering a sequence will render them all, but they’re all output with the “masterLayer” name, so they all get overwritten with the last rendered render setup. Rendering a single frame in the normal way uses the correct filename, with the <RenderLayer> placeholder being replaced with the active render setup. Not sure if there’s a workaround, other than rendering each render setup individually, changing the filename string manually each time.


Here’s the release notes, I had some trouble finding it:

If Arnold is loaded, it adds a lot of extra attributes to things, and since they all start with “Ai” they all sort to the top of the CB, which is annoying. It also adds an “Arnold” menu to every menu set, which isn’t very good integration (why aren’t Arnold lights in the Create > Lights menu?), and sort of defeats the point of Maya’s menu sets.

The viewport with smoothed meshes seems much better. In 2016, smoothed meshes would cause tumbling the viewport to be very hitchy with tumble on object. In 2017, I don’t see any hitching at all. I think it’s still doing hit detection agaainst the smoothed mesh (except in adaptive, which doesn’t), but whatever they’ve done it’s much smoother now. This is nice, since you can leave smoothing on for a lot more meshes, instead of having to turn it on selectively.

OpenSubDiv adaptive is much slower than it was in 2016.5. With a mesh that’s just 15k polys, tumbling the viewport drops to around 50 FPS for me. In 2016.5 it was around 400 FPS. Regular OpenSubDiv CC with division levels set to 3, giving 1.1M faces, is still around 400 FPS. The GPU acceleration is probably more useful when it allows more deformers to be GPU-accelerated. (This is on a Geforce GTX 750, for reference.) I’ll stick with OSDCC/Maya CC.


Hello gfk,

In the past year we have been working on better OpenSubdiv integration into viewport for both OSDCC and OSDAdaptive modes. These are great feedback - thanks for the quick update!

Regarding the performance issue of tumbling OSD adaptive object, we are curious to understand what caused that problem. Could you please share your hardware / OS spec, and also which viewport device was used?

Please note in 2016.5, OSD adaptive mode was not supported in DX11 viewport yet - under the hood it was still using OSD uniform mode. This has changed in 2017.

Wenle from Maya team


Continuing to look at the time editor:

  • I was confused that creating poses in the time editor didn’t seem to save. They actually do, but it was putting the pose in the middle of a clip, and poses inside clips seem to have no effect. The pose data is still there, and shows up if I drag the pose out into a separate track.
  • I was also confused about ghosting. Turning on ghosting doesn’t ghost within an animation clip, it ghosts the beginning/end of the surrounding animation clips. If you want to ghost the animation itself, add ghosting from the Visualize menu normally.
  • To interpolate between poses, select two poses and Create Transition. That’s why it was stepping from one to another–I hadn’t added a transition.
  • Poses are just clips, but you can’t extend the length of a pose out the time editor since it’ll just drag the pose around. If you go into the AE, set “clip loop after mode” to “hold” and then increase “clip hold after” it’ll extend the clip, and once you do that you can extend it normally in the editor, overlap it with other clips, etc.
  • Exporting clips bakes them to individual keyframes, instead of saving the keys as-is, which isn’t great. I haven’t yet found any way to take a clip, export it back to regular keyframes as it was when you imported it (not baked to a keyframe every frame) so you can make changes, and then replace the clip in the composition with it. “Bake to Scene” does the same thing. (I was trying to find a way to extract a clip into the scene so I could edit it, then load it back into the time editor, without having to edit each clip in a separate scene and then re-import it.)
  • The time editor doesn’t support character sets, and you have to delete the character set and save a copy every time to import clips? That’s a rough one…

Please note in 2016.5, OSD adaptive mode was not supported in DX11 viewport yet - under the hood it was still using OSD uniform mode. This has changed in 2017.

That’s the difference, then. I use DX11 in 2016.5, due to some performance regression with OpenGL in 2016.5 (don’t recall what off-hand, but it was reported). I don’t use typically use adaptive and was only looking at it to see how it had changed in 2017. I also saw the change notes for improved viewport support for OSDCC and I’ll be giving that another try–I’ve been using Maya CC since I always had trouble with materials with OSDCC, but right now all of my materials are broken anyway without mental ray so I can’t test it yet.


Hi gfk,

OSD adaptive mode is suitable when preview character animation/adjust your character’s pose under subdivision mode.
Because the adaptive’s subdivision is accelerated by the GPU, the playback or modifying speed is much faster than OSDCC/MayaCC.
But for static meshes – using OSDCC/MayaCC will be a better choice :slight_smile:



Maybe I am doing something wrong, but I can’t seem to write decimals in the inputs of the multiplyDivide node. Works fine in 2016 E2.

To the Maya devs reading this forum: Looks like another great release, though. Keep up the great work.