Hello,
We’re having issues with flame importing clips in the wrong colourspace when multiple flame ops are working on the same project in different workspaces. The steps to reproduce are;
- User A launches flame in project A with workspace A
- User B launches flame in project A with workspace B
- User A opens MediaHub tab, sets colour management for an import e.g. Mode=View Transform, Invert=False, Display=ACEScg,View Transform=From Rules, Display=Rec.709 video. Crucially, they’ve set the colour management to a new setting, i.e. made selections in the UI to change the settings
- User A clicks away from the MediaHub to a different tab ← this specifically seems to be what triggers something across users.
- User B runs our import tools, which are essentially a wrapper around;
`flame.mediahub.files.options.set_tagged_colour_space(colourspace)
flame.import_clips(‘/path/to/files’, reel)`
While you would expect User B to have clips imported in their given colourspace, in fact, the clips get imported with User A’s settings. In the example given User B’s clips should be in ACEScg, but are in Rec709.
Since all subsequent imports will also be using User A’s settings, the only thing I can been able to find that clears this is for the user to manually tab to MediaHub, set the colour mode to something else, then set it back - then it will start importing in the correct settings.
I note in the Flame documentation mediahub settings do indeed state options are saved automatically and at a project-level, rather than per workspace.
For reference here is a brief outline of our systems setup:
Every artist has their own AWS EC2 instance, each running its own local Flame install.
Projects themselves live on a central project server, and that project data gets exported over NFS4 and auto-mounted on every workstation. The mount options we’re using (acdirmin=0, acdirmax=0, actimeo=0, basically) are straight out of Autodesk’s own Flame project collaboration docs for multi-workstation setup, disables NFS attribute caching so you don’t get cache desync when Flame hard-links/renames files internally.
Finally, I tested this in both Flame 2027 and 2025, which both have the same issue. However I do notice PyMediaHubFilesTabOptions does have different attributes/methods availble, which allows me to do this;
flame.mediahub.files.options.colour_mgmt_mode = "Auto Convert"
flame.mediahub.files.options.colour_mgmt_mode = "Tag Only"
which seems to be enough to replicate the ‘on and off again’ manual workaround in 2025.
Though, side note: these are not documented. @fredwarren could we get those added to the Flame 2027 docs? The missing methods/attributes are;
colour_mgmt_display
colour_mgmt_invert
colour_mgmt_mode
colour_mgmt_view
colour_mgmt_working_space
side note to the side note, this throws an Error: basic_string: construction from null is not valid unless colour management mode is set to Auto Convert, which is probably not desirable.
import_transform
tagged_colour_space
Questions;
- Are there any changes we can make to our systems setup to avoid this issue? Or is this expected/unavoidable behaviour because of how Mediahub works?
- In Flame 2025 is there any way to set colour management via python api to anything other than ‘Tag Only’?
- Could we get this ‘shared mediahub settings’ behaviour re-considered? It can be quite disruptive to production on multi-workspace projects because users may not notice the files have been imported incorrectly and end up sending clients shots with incorrect grades. Even though there is a workaround on Flame 2027, it is still susceptible to race conditions.