Decide how to handle collection instancing or linked data #83889
Labels
No Label
Interest
Alembic
Interest
Animation & Rigging
Interest
Asset Browser
Interest
Asset Browser Project Overview
Interest
Audio
Interest
Automated Testing
Interest
Blender Asset Bundle
Interest
BlendFile
Interest
Collada
Interest
Compatibility
Interest
Compositing
Interest
Core
Interest
Cycles
Interest
Dependency Graph
Interest
Development Management
Interest
EEVEE
Interest
EEVEE & Viewport
Interest
Freestyle
Interest
Geometry Nodes
Interest
Grease Pencil
Interest
ID Management
Interest
Images & Movies
Interest
Import Export
Interest
Line Art
Interest
Masking
Interest
Metal
Interest
Modeling
Interest
Modifiers
Interest
Motion Tracking
Interest
Nodes & Physics
Interest
OpenGL
Interest
Overlay
Interest
Overrides
Interest
Performance
Interest
Physics
Interest
Pipeline, Assets & IO
Interest
Platforms, Builds & Tests
Interest
Python API
Interest
Render & Cycles
Interest
Render Pipeline
Interest
Sculpt, Paint & Texture
Interest
Text Editor
Interest
Translations
Interest
Triaging
Interest
Undo
Interest
USD
Interest
User Interface
Interest
UV Editing
Interest
VFX & Video
Interest
Video Sequencer
Interest
Virtual Reality
Interest
Vulkan
Interest
Wayland
Interest
Workbench
Interest: X11
Legacy
Blender 2.8 Project
Legacy
Milestone 1: Basic, Local Asset Browser
Legacy
OpenGL Error
Meta
Good First Issue
Meta
Papercut
Meta
Retrospective
Meta
Security
Module
Animation & Rigging
Module
Core
Module
Development Management
Module
EEVEE & Viewport
Module
Grease Pencil
Module
Modeling
Module
Nodes & Physics
Module
Pipeline, Assets & IO
Module
Platforms, Builds & Tests
Module
Python API
Module
Render & Cycles
Module
Sculpt, Paint & Texture
Module
Triaging
Module
User Interface
Module
VFX & Video
Platform
FreeBSD
Platform
Linux
Platform
macOS
Platform
Windows
Priority
High
Priority
Low
Priority
Normal
Priority
Unbreak Now!
Status
Archived
Status
Confirmed
Status
Duplicate
Status
Needs Info from Developers
Status
Needs Information from User
Status
Needs Triage
Status
Resolved
Type
Bug
Type
Design
Type
Known Issue
Type
Patch
Type
Report
Type
To Do
No Milestone
No project
No Assignees
10 Participants
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: blender/blender#83889
Loading…
Reference in New Issue
No description provided.
Delete Branch "%!s(<nil>)"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Currently, dropping collections always creates a collection instance. This should be optional somehow, there are valid usages for both cases. Should it just be an option of the drop operator? So users could change it after dropping. (May be annoying when dropping heavy collections from external libraries, which can take a moment.)
Similarly, some users seem to expect that object data is linked by default. This is already an option of the drop operator (disabled by default). However it seems like they actually expect that data is a new "instance", rather than a full copy. This could mean either de-duplicating already existing data (so that dropping the same material asset onto 10 objects doesn't create 10 local materials), or linking data rather than appending.
Added subscriber: @JulianEisel
Added subscriber: @lowpolysaac
I'm not a developer but from my user perspective, I expect assets to be linked data always by default (if I want to duplicate things I can later use library overrides, etc right?).
In my opinion I think users should decide beforehand how the asset is going to be placed in the scene (full copy, linked...) just like now there are 2 options to import data (link or append). I believe this is better than first dropping something and then making the decision of "what I want this to be?" (the drop menu could be useful for deciding other things). This type of workflow not only would prevent the waiting for heavy collections issue but in the end is educating new users about how Blender manages data internally (Shift + D is not Alt +D, etc). But again it's my opinion. So far the asset browser works great!
Added subscriber: @wilBr
@Severin..... I think like @lopolsaac.. linked copy as default with option to change this behaviour..
Another thing is, if possible, keep asset in browser if it was deleted from scene, asset keep existing in blend file but not in scene.
great work.. thks
Added subscriber: @Dimitar
I would also expect assets in the same file, when dragged into the scene to be linked, but even more so than the object data. For example if the asset object is a cube with subdiv to level 3, then the linked asset should also be a cube with subdiv level 3. if subdiv level has changed of the original asset, so should the linked assets. It sounds a bit like a linked collection instance. Yet, it would be great if instead of a linked collection instance, it remains a mesh object, so e.g. further modifiers could be applied to the linked assets.
Added subscriber: @sozap
Doesn't look easy to solve.
Link or Append should be easy to change and easy to discover.
Indeed seems annoying... maybe when an asset is appended once it's replaced in the asset browser by a local version.
Sounds that one may want to :
Added subscriber: @BukeBeyond
Linking should be the default behavior. If the user needs to embed the linked data, they can do so later (make Local), just like they could with embedded textures (as uncommon as that is used).
Currently, copy importing asset data is unusably slow and inefficient. Materials and textures get duplicated on import, using up gigabytes of GPU memory and taking extra time to redundantly upload the textures to the GPU, using far more resources than the entire original source library file, on just a few imported pieces from a kit library.
The new Asset Browser in Blender should be robust and able to handle large assets like film quality buildings and trees. Currently, this is unusable, requiring manual linking. The new Asset filter in the Link file browser is a nice feature, however.
It would be good to have some more context.
There is no doubt that the Asset Browser should support linking. But at the current state it is designed for specific use-cases, and I'd like to understand where exactly the demands for this are coming from before just adding it.
So what are the use cases you have in mind for linking in this first version of the Asset Browser?
I assume implementing linking (even optionally on the alt drag) will be easier than solving the material duplication problem later. Without linking, heavy assets are too slow to use and max out gpu memory with redundant textures. Imagine a film quality city library, with over 100 textures shared amongst 50 buildings. By the time you append the 4th building, you are using more ram, and have waited longer than loading the whole library file. Linking solves that.
Linking also brings lot of other modular options. You can convert to local append later, but not the other way around.
I'd say that this could be a common workflow for a lot of studios: you want to create several iterations of similar scenes but you want to control the overall look in this initial stage easily. You come and go making small adjustments for a while (colors, shapes...). And when you are happy with a base look, you can create instances of the linked objects/collections and an artist can start from there creating more detailed and unique libraries of barrels and tables that now look more "cyberpunk" and less standard. It's working from "big to small". So this point (1) is... fast prototyping/iteration for artists in an early stage of a project (?)
Sorry if this is not what you would like to know. My final and more important point is: instancing linked data is easy but the other way around is not (right?). I think unique copies of a mesh while I'm sculpting is useful though. Meaning: if I'm sculpting a soldier and I need a hand, I don't need a linked mesh at all...
I think, given the goals of the first asset browser milestone appending is the right behavior, the most important issue now is that appending several times the same data makes copies of it.
Linking is in fact useful when your projects are split into several files and from what I understand is out of scope of the first milestone.
But being able to link the data would make Asset browser useful also on bigger scale project as it is, even if it's part of a future milestone.
Working on commercial project as part of a team, we made an asset loader that handle character, props, sets . But something is missing for materials , node groups and other type of data.
Being able to link some compositing node groups, material groups , even if it's not perfect at this stage would be quite helpful right now.
What about something like this :
A toggle that allow you to switch between link or append, and also is informative about asset browser behaviour ?
All that said, Asset browser looks awesome, I really look forward to start using it , and I'm confident it will grow in the right direction from release to release.
I agree. I mean: I want a linking system for the reasons I exposed... but I totally understand that maybe is out of the scope for now. The Asset Browser looks great already, it's better not to rush things :)
The design of this isn't so much of an issue, it was planned to have early on:
But enabling things like that can have implications so I just want to understand context well.
Yes, that is great, the easiest path for you for now to enable linking, so we can start using the asset browser for our case of heavy modular assets.
In the future, the linking can be further refined, so it behaves more like append. For example, the holding object can be local copy but the mesh and materials are linked, so the user can start positioning it right away, and convert with make local to edit the mesh and materials when needed. I think the material and texture duplication problem with current append is a deal breaker until it is solved comprehensively later in Blender.
Added subscriber: @Harley
A very contrary opinion, but...
Ideally when dragging and dropping works well, you can start dragging a representation of a thing without first knowing the end action. This is because it is the drop zones that determine their own capabilities and sometimes have multiple options.
You might have multiple editors with very different settings and capabilities. So you might want to drag a thing to have one affect in one place, but another affect somewhere else. For Materials you might drag a thumbnail from Asset Browser to a model or to outliner or Properties or export by dragging somewhere else or you might drag to a different Asset Browser in the same Blender instance or another.
So I would say do not have any setting in Asset Browser for this. Instead leave it up to the drop zone. If there are multiple actions or multiple options for that drop zone action then expose them then at time of dropping. In most cases the drop zone is simple and the action is obvious so just do that immediately. Otherwise open a popup menu and ask the user how they wish to proceed. Again, most cases it will just proceed. In some cases you might have to answer "Link" or "Append", while in other cases there might be multiple options and checkboxes, etc.
I think all those refinements are great, perhaps multiple link append preferences for diverse asset types.
But if the basic linking does not make it to 2.92, we are looking at another 3 months for it to gain traction. Once people run into the duplicate materials and textures problem with appended assets with no other easy option, they may completely give up on it, and never look back.
Ok ! sorry to try to rush things, it's great to know that it will be addressed at some point, as users we don't have a good understanding of what's easy to do or not ...
@Harley , I'm not sure if that the best fit for append or link, even if I'm sure this idea can be quite helpful in some cases .
To me you either want to link or append, but I don't see why one would switch from one to the other in the same blender "session".
To me having to click "link" or "append" every time I drop an asset gets a bit in the way.
How I would use the asset browser :
I'm modeling a props and want to kit-bash some parts, I set the asset browser to append . Then I drop a bunch of meshes, try some things, until I'm satisfied with the result.
Now I want to add some materials, I will probably use append , but that depend on the project, I won't be mixing linked and appended materials anyway.
I want to do some layout or set dressing on a shot, I activate Link, and during my blender session I would link every asset in the shot.
Given the task you're on it's really one or the other, you don't want to mix things, or hit one instead of the other by mistake. IMO that should be a global settings that you can set easily.
It's also something that could be set using scripts : If I work as a lead pipeline on a project I may do some automation to activate link or append depending on the context ( if the .blend is a shot or a props).
That way artists don't have to care most of the time about how it's set. It's always the right choice depending on the stage we are in the project. From experience on a project not every artist is technical and will care about this.
Link or Append look quite similar at first sight so it's easy to mess up and it's quite hard to fix once the issue is propagated to different assets or shots.
Sorry for the long post, I just want to give more user cases, thanks for taking time to read this !
No, but that is only thinking about one type of transaction. Think instead of the Asset Browser being the center of a number of transactions, some of them unknown to us now. So yes, you might not switch between linking and appending when dragging to the 3DView. But that drag could instead go somewhere else entirely. For instance you might one day be able to drag directly from Asset Brower to another Asset Browser. Or drag from there to your OS desktop.
Indeed, that would be great ! but I still think link and append should be solved differently especially if we don't know yet these extra features that would require a popup.
Something like : dragging a .blend from the desktop to the asset browser, show a popup that let you either open the file or use it as a temporary asset repository could be a nice additional feature and make much more sense to me.
Dragging an animation to a rig may let you have some options like replace, add to NLA etc... All these are great case for a popup to me, but again link and append should be handled differently IMO.
Linking assets is at the core of an animation production and ideally it would be great for studios to be able to customize the asset browser and don't expose link or append to the user, it should be set for the user depending on the context (asset modeling, set dressing ...) .
That way we can prevent many error made by mistake that may propagate to many shots until someone notice it. It's quite tricky to replace a local rig by a linked one once some constraints have been added and animated. But I understand that other uses cases have different needs.
Dragging in a blend library, to check out its assets and quality is a great idea. In fact, this shows that the asset browser, and the regular file browser for append/link/open should converge into one. Convergence is always a good thing. :-)
Of course the terminology can be made far simpler and more intuitive. We programmers often rush naming things by how the car engine works, and not how the driver drives the car.
Append means, to chain sequences in an order. What order is that on the user viewport? When copying files or directories into another, how confusing would it be to call that Append? So that could be renamed just Copy, Copy in, or even Import.
Link sounds just fine.
Make Local? Is the library asset Remote, or on another machine, a url? Name that Copy/Import like above too.
Luckly down the road, these renames and convergences would not break existing projects, just reduce the friction for new users to discover and to understand Blender's power more rapidly.
Added subscriber: @user1
When i provide an asset on blendswap, it is usually a collection of things, with parentings, drivers and f-curves. I guess the user of these assets expects to paste the asset from his repo and with a single click be able to change the f-curves, but have e.g. the root-empty still functional (example asset: https://blendswap.com/blend/27143).
Currently the best he can do is make the instances real and fix the hierarchy manually. All the drivers are broken at this point.
Added subscriber: @RobertS
I don't know if this is the correct place but:
Open file C and find the link now broken
(This feels unacceptable for a modern, robust system...)
Worse:
I am removing the
Needs Triagelabel. This is under the general rule that Design and TODO tasks should not have a status.If you believe this task is no longer relevant, feel free to close it.