Blender 2.8 ideas; Should the current Multires functionality just be removed?

General Forums Blender and CG Discussions
1

What the title says, as I think there are some strong arguments as to why Blender should just lose Multires functionality altogether for the time being.

  • Multires is all fun to work with until you need to do your first rebase, so most people will try it only to find it crushing their vision later.
  • The combination of Dyntopo sculpting and Cycles Microdisplacement allows one to get small details with no need for multires. The case for that could get stronger yet if Cycles started supporting Vector displacement (and if the viewport allowed one to paint such a map onto a model with the help of GPU tessellation). Yes, I’m talking about a possible future where multires in general becomes obsolete technology.
  • The whole thing needs to be ripped out and rewritten anyway according to the developers, what better time to do that (the ripping out part at least) than with 2.8.
  • The only people who seem to use the modifier are newer users who have not been around long enough to be aware of its issues (then they go to Zbrush or switch to Dyntopo and/or displacement).
  • It will take care of the endless train of questions and complaints about why spikes are forming or why the mesh seems corrupted, can’t complain anymore if the focus is on proposing Multires as a feature request rather than something breaking people’s work.

As of now, it almost seems like it’s ethically wrong for the Blender Foundation to advertise Multires as a feature (because 100 percent of the time, guaranteed, the data will eventually screw itself up). The time then might have come to just remove it with the idea that a developer might come and make a new implementation that works properly (unless of course newer technologies as mentioned in point number 2 actually removes any need to have it).

2

I guess its a New Year so we need to have our yearly rant about multires:RocknRoll:.

3

Well I don’t have any issues with Multires. I used it for the first time in my most recent work just fine. Dyntopo would have just made a mess of the topology meaning extra work to correct it.

4

You can never (and I mean never) add details in dynatopo on the detail level of Muti-Res. The most you can do in dynatopo is around 2M triangles. In muti-res you can easily go around 20 M (64 M is the max I worked with).

For people who have Zbrush, 3dCoat (or some other sculpt tool) it won’t matter so much if you temporarily remove multi-res. So if the plan to create a history based sculpting mode is agreed upon, then its fine to remove it. But if there is no such plan then maybe think about leaving it as is.

5

No, it shouldn’t be removed. It’s basically useless for actual multires editing (as in, being able to switch between different resolutions and have changes propagate between those) but its invaluable for sculpting at really high res. I don¡t know what black magic it uses, but it can never be replaced with regular subdiv …

6

I am blender user for 20 years.
I never had real issues with multires, just don’t change the topology.
It is really usefull in big scenes. I use it quite often.

Really, a very strange topic to discuss…
Don’t you really have anything normal on your task list?

7

With your suggestion people have to wait until render time to see the result of their work, unless tessellation is supported on the viewport.

Instead of asking to remove useful, and in this case VITAL features, why don’t you ask to improve the toolset?

At the moment to load a texture for your brush you have to preload it on another part of the interface, which is incredibly inconvenient.

A good thing to do is to implement the same technology Stéphane Ginier is using in his SculptGL Voxel Remeshing tool, so that you could quad-remesh your sculpt.

8

I would be all for keeping the Multires modifier in Blender if it could just work (as in, you could play with a sculpt all day, switching between different levels and rebasing, without receiving a mess by nightfall). Instead, you have to invent some ‘Blender’ way of doing things where only a small chunk of the total functionality actually sees use.

Psy-Fi has indicated before that the case can be made to see Multires as a broken feature as opposed to a highly usable one (because the way the tangents are stored is broken by design and will almost always lead to either spikes or fully corrupted geometry with NaN values. What’s even worse is that no one can report them as bugs because they are random). In that sense, it would not be removed simply because the BF hates sculptors or something like that, but it would be treated the same as other broken features in the past.

Either way, the only way to have working Multires in Blender is to rewrite it from the ground up (now I’m not a dev. by any means and this thread is just an idea, but I think an improved version could work by storing the tangents in the form of a vector-displacement texture via Ptex (so at its heart, you’re actually editing a texture while pushing geometry).

9

Multires shouldn’t be removed, it should be fixed; and in the process sculpting layers functionality should be added too, without sculpting layers there’s not much you can do on multires. It feels like a mockup for a possible feature.

And yes, in general, the whole sculpting workflow in Blender should be revised. Maybe the workflow and templates project could help with that.

10

I’m currently making dinosaur skulls. Each skull has two finished products, a detailed 3D print version for Shapewaysand a lo-res realtime version with baked AO and normal textures which ends up on Sketchfab.

Both are produced from the same source – a single multires model.

Because multires is subdivision technology I can keep my UV maps for the realtime model without having to retopologise. I like Dyntopo for freeform sculpting but it doesn’t cut it for my specific working method here.

The whole thing needs to be ripped out and rewritten anyway according to the developers, what better time to do that (the ripping out part at least) than with 2.8.

Why remove before rewrite, when some of us are getting use from it?

The only people who seem to use the modifier are newer users who have not been around long enough to be aware of its issues (then they go to Zbrush or switch to Dyntopo and/or displacement).

I’m not a new user and I’m aware some people have issues. I don’t want to buy Zbrush and Dyntopo/displacement is not practical for my preferred working method.

It will take care of the endless train of questions and complaints about why spikes are forming or why the mesh seems corrupted, can’t complain anymore if the focus is on proposing Multires as a feature request rather than something breaking people’s work.

Maybe it’s the way I use it but I don’t get spikes. I also move up and down the levels tweaking and refining details. My multires subdivs go to level 6 or 7. Typically several million polys.

Multires is the ideal solution for the way I work and for me it’s 100% reliable. Don’t remove it unless an improved version becomes available.

11

Only when I read the developer’s take on the code (along with the opinions of reputable members like Doris who honestly tried to make the feature work), I’m not sure how it can just be ‘fixed’ when considering the issues are impossible to reproduce and deeply embedded into the design.

It might be possible to do to the Multires code what Tristen and co. did to the mess that powers the BGE (but you will need an experienced developer who can actually rework it without breaking every model using the feature, and you usually can’t count on someone to ‘fall out of the sky’ with a patch like the BF put it for Mai and the work on microdisplacement).

12

I got to ask do you even use multires or are you just swinging for the fences based on other people’s say so. You can’t really sculpt details that well using just dynamic topology, dynamic topology->to multires is a workflow I use and guess what I haven’t had a spike problem since 2.5X days.

There is nobody currently working on things like sculpting or particles, absolutely nobody. As broken as these features are people still use them to do work. So when some of you guys go on about this or that feature has to be removed from Blender because it’s ‘broken’ do you want those of use that use said feature to twiddle our thumbs in the hopes that a developer will magically spring up and start working on these features.

13

I did use Multires for a full creature model in a couple of recent projects. For one of them it only lasted until a re-base was needed (which then proceeded to screw up the tangents). I also tend to find it limiting that the editable part of the multires functionality isn’t exactly what it’s advertised to be (it works for some tools and completely destroys the mesh when used with others).

So in a way, I can now add some personal experience here (I used to think any problems could be resolved by making sure there was no bad geometry, but the broken rebase function for instance does not need bad geometry to mess up).

14

As a fairly proficient ZBrush user, I can tell you that as a general practice both for software unpredictability as well as workflow efficiency, you really should try not to mess with the topology with multiple sub d levels anyway. That goes for ZB and Multires.

On the software side: Yes, in ZBrush, you can go to the base level and change the topology, but it will often cause unpredictable results all the same. You would generally rely on other tools entirely for this sort of thing, such as creating a new mesh with the topology changes, and then projecting the sculpted detail onto it.

On the workflow side: If you’re getting into detailed sculpting work before you have proper clean topology, you’re not really working as efficiently as you should be. Yes, the detail work is fun, which is why one of the biggest mistakes people make in sculpting is to rush into that part. In practice, however, you should be nearly done with a project before even thinking about adding additional sub d levels. :slight_smile:

15

Ace

What is the problem with MultiRes? Just curious. I used it few times but not much as I am not a lot into sculpting.

16

the only problem that I have personally with multires is that u have some lag if u use it in high poly models, wich is something that I don’t recommend, and the spikes that sometimes appears in low-medium levels. to me dyntopo>retopo>multiress works fine as long as u don’t try to sculpt in low leves or delete higher

17

The real problem with multires is not the infamous spikes. It only happens if you try to edit the topology of the mesh after sculpting, which is not a recommended process anyway, not in Blender, not in Zbrush, not in Mudbox or any other app that uses regular subdivision for this; as @cgCody mentioned:

…you really should try not to mess with the topology with multiple sub d levels anyway. That goes for ZB and Multires.

On the software side: Yes, in ZBrush, you can go to the base level and change the topology, but it will often cause unpredictable results all the same. You would generally rely on other tools entirely for this sort of thing, such as creating a new mesh with the topology changes, and then projecting the sculpted detail onto it.

On the workflow side: If you’re getting into detailed sculpting work before you have proper clean topology, you’re not really working as efficiently as you should be. Yes, the detail work is fun, which is why one of the biggest mistakes people make in sculpting is to rush into that part. In practice, however, you should be nearly done with a project before even thinking about adding additional sub d levels. :slight_smile:

Let’s see the 2 most common scenarios:
1. Sculpting and detailing for animation/games: You will either do a highpoly sculpt and retopo after is done, or do the sculpt and detailing on a model with proper topology. Then the high poly detail is baked onto disp and normal textures. In this case, for Blender you can delete the multires modifier after the maps are baked (I’ve seen some complaints about Blender not being able to handle a character with a multires AND an armature modifier, which is nuts).
Doing this process properly wouldn’t need you to go back to the mesh and edit it while sculpting, so the spikes shouldn’t be a problem right? But, if there’s the need to change the topology, then the sculpt info should be projected into a new mesh, instead of modifying the original.
There’s real problem number 1: There’s no easy way in Blender to project Multires sculpt info from one object to another. :no:

2. Sculpting for concept art and/or stills: most of the times a clean topology won’t be needed for this since a lot of paintover and post-processing is done after render, so probably a combination of Dyntopo/Remesh/Multires can be used. Again, if the Multires comes at the end, and only for final detailing, the spike issue won’t be a problem.
Also, with a powerful enough PC, you could even do the whole thing in Dyntopo without ever touching Multires. :smiley:
Now, if the concept/sketch gets approved and the model needs to be optimized for production, then retopo must be done to have a clean mesh ready for detailing.
There’s real problem number 2: Despite having several addons for retopology in Blender, there’s still a hole in that department. We need proper retopology tools for Blender, or a better Remesh modifier, on steroids.

Problem number 3: No sculpt layers.

Problem number 4: No way to bake or use vector displacement maps.

Problem number 5: Vertex paint is still wonky (I hope the GSOC project for this gets added to trunk soon).

Problem number 6: In general, the whole sculpt workflow is complicated and unintuitive, so in terms of UI something should be done too. Hopefully this will be taken into account for the templates of 2.8 :slight_smile:

Now, if the multires modifier gets to be rewritten, great! All of this could be fixed and properly implemented.

BUT! If there’s no plans for a rewritte or fix then it should be left alone. Incomplete as it is right now, there’s still a LOT of things you can do with Multires.

1 Like
18

Maybe like experimental mode add a deferred mode.
So people can still use it, but its not advised, and its still there for backwards compatibility reasons.
For newbies unaware of the multiress problems its no good
I think its below the overall quality we have in blender these days, and its better to be avoided with a great warning…

Though cycles micro displacement is (i think ) a face level thing, to add a real Z displacement as in contrast to bump maps, and thus sculpting… is not a direct match with expanding upon this technology… i think, but i might wrong here.

19

Not officially perhaps, but I can confirm Secrop’s instructions worked to both create a VDM, and use it in a displacement modifier.

The results looked correct to me.

Edit: Produces World space VDMs. Is only correct along the same plane here. Try on each face of a cube, or rotate the displaced geometry in Edit mode to demo potential errors.

However the same map didn’t work as a sculpt brush texture, but perhaps there’s a trick I’ve yet to figure out?
Was trying to achieve this workflow (from 2 minutes on):

20

it’s pretty useful for games, you create your low poly mesh, subdivide it a few times then sculpt in hipoly details all while retaining the topology and ability to edit your base mesh. I only work on the highest subdivision and don’t really care about going back and forth between divisions, but not having layers or the ability to undo is killer. If there were a kickstarter to rewrite/improve this aspect I’d be all over it.