Page MenuHome

Retopology in Blender
Confirmed, NormalPublicDESIGN

Assigned To
None
Authored By
William Reynish (billreynish)
Jul 31 2019, 2:07 PM
Tokens
"Love" token, awarded by hitrpr."Burninate" token, awarded by Rasta."Love" token, awarded by dlc17."Love" token, awarded by ILJI."Love" token, awarded by 14AUDDIN."Love" token, awarded by Krayzmond."Love" token, awarded by christianclavet."Burninate" token, awarded by Alumx."Love" token, awarded by R2792."Love" token, awarded by crazyWikka."Love" token, awarded by gwobbler."Love" token, awarded by tilapiatsu."Love" token, awarded by brilliant_ape."Love" token, awarded by dpdp."Burninate" token, awarded by MetinSeven."Like" token, awarded by Fracture128."Like" token, awarded by knightknight."Love" token, awarded by S_Jockey."Like" token, awarded by iWaraxe."Love" token, awarded by Brandon777."Love" token, awarded by mauriciomarinho."Love" token, awarded by sineo."Love" token, awarded by Shimoon."Love" token, awarded by bnzs."Love" token, awarded by fiendish55."Burninate" token, awarded by hjrabi."Love" token, awarded by runswithfork."Love" token, awarded by sakana3."Love" token, awarded by Draise."Love" token, awarded by helloidonthaveanyideaformyusername."Love" token, awarded by VertexPainter."Love" token, awarded by aliasguru."Love" token, awarded by Frozen_Death_Knight."Love" token, awarded by A.Lex_3D."Love" token, awarded by michaelknubben."Love" token, awarded by FaisselB."Love" token, awarded by johnsyed."Like" token, awarded by belich."Love" token, awarded by lcs_cavalheiro."Love" token, awarded by Tonatiuh."Love" token, awarded by koloved."Love" token, awarded by TheFlow."Like" token, awarded by Jaydead."Love" token, awarded by ixd."Love" token, awarded by erickblender."Love" token, awarded by EitanSomething."Like" token, awarded by mendio."Love" token, awarded by 0o00o0oo."Love" token, awarded by Jules."Like" token, awarded by amonpaike."Love" token, awarded by Zino."Love" token, awarded by Alrob."Love" token, awarded by xrg.

Description

In Blender’s overall workflow, which includes modeling, sculpting and animation, a major missing piece is retopology.


No Retopology?

As part of the overall focus of sculpting improvements, this is an area we should address.

After talking to @Pablo Dobarro (pablodp606), we agreed to set up this document to help define how we should approach this from a UI and workflow perspective.

Issues

Display

One of the major issues currently, is that we have no great built-in way to display your retopology mesh on top of a high res sculpt.

By default, your mesh will intersect with the high res sculpt:

You can enable In Front, but that makes the *whole* retopo mesh display on top of the reference, also the parts that should be occluded:

Not only is it confusing to set it up, but the result is sub-par.

What we really want is something more like this:


The occlusion is correct and the faces aren't intersecting. The retopology mesh is half transparent so you can see the underlying detail.

Snapping

There are a few issues with snapping:

  • You have to enable some very specific settings that are not obvious
  • There's no way to only snap to a certain object or Collection - you can easily accidentally snap to an item somewhere in the background
  • Snapping is only active while you are using any of the transform tools. If you for eg subdivide an edge, those edges aren't snapped to the source. Users then have to select those items and move them a bit to snap
  • Some users use the Shrink-wrap modifier, but this has issues with the mesh is too far away from the source - the projection becomes unpredictable
Tools

Compared to more dedicated solutions for retopology, Blender is lacking many basics. Things like drawing a 'strip' of quads, or defining a patch of quads. Or tools to automatically handle appendages just by drawing lines along it. Or even just a basic tool for 'drawing' topology onto the source mesh, which is then auto-filled with quads.

Setup

To get started with retopology in Blender, so many settings and display options have to be toggled and set by the user, that there exists entire tutorials on how to do this. The needed steps and settings are not only non-obvious, but also take time and focus away from the task at hand.


Solutions

Display

Add a Retopology overlay display option, replacing the current Hidden Wire Edit mode display

Snapping

Add a Retopology snapping option, which allows snapping to self while also projecting onto source, and snaps the current selection.

Tools

We can start by improving certain tools, such as Poly Build, and also add more edit mode tools. Even though they may be more constrained inside Edit Mode, we can try to use this as a test-case for more advanced gizmos and tools in general

Other examples of the kinds of tools we could add are things like:

A tool that lets users define a patch with handles, which is automatically filled with quads:

A tool that would let users draw a line, which then becomes a strip of evenly sized quads

A tool that would allow users to nudge around points with falloff, like a brush:

Addons

We probably don’t expect to provide a full and comprehensive set of retopology tools, at least initially. For this reason, addon developers could provide extra tools that integrate with the toolbar and tool system.

Event Timeline

There are a very large number of changes, so older changes are hidden. Show Older Changes

Something I'd like to add to the shading style:

  1. If possible the backface opacity of lines should be adjustable.
  2. If Faces are occluded by geometry they should probably also be rendered with slightly less opacity:

For as often as I quote Modo - the Topo display violates both those principles and makes editing the final stages of a complex model very painful.
In the picture you can see that while the wireframe can be distinguished clearly the opacity of the lines still makes editing very straining.
Also - much more important - the polygons in the center of the image are behind the high-poly mesh but are still rendered in front.

They also snap to the editing tools just as if there wasn't any geometry inbetween. Sometimes that's wanted - more often it is unwanted though.
(edit) Just realizing that the snapping behaviour needs to be there for modeling around ~90°+ angles. So that's probably a little more complex in practice.

Albert (wevon) added a subscriber: Albert (wevon).EditedAug 23 2019, 6:04 PM

I wanted to share a trick that I use to perform retopology. Subdivide the mesh but with the simple option. In this way, the contours do not round up so much, and the edition vertices remain in the same place as the originals. With the extra faces and the Shrinkwrap the new mesh is always seen above the reference mesh.

Edited:
I'm late.
Checking the thread, I see now that in the video of Jean Da Costa (jeacom256), he does exactly the same.

sineo (sineo) added a subscriber: sineo (sineo).

Hi folks, I was researching on good retopo method for a while now. And these are some of my settings. Hope it helps

https://youtu.be/47LF4o0hkXY
https://www.youtube.com/watch?v=Sb53GpN9YXg

Hi,

Blender has a bunch of tools integrated for retopology and some stuff is hidden, this is a problem for a good approach.

I think that Blender 2.81/2.82 have to merges all of this in an only one best tool for Retopology process.


My idea is:

Similar to Retopology Flow by CGCookies
Teaser

Poly-Build has had great improvements from Pablo Dobarro

Look at this Poly-Build 2.81

BSurfaces instead is an awesome tool for the lovers of the GPencil
Setup Blender 2.8 for Retopology

Other stuff is GStretch in LoopTools
GStretch + LoopTools or
MiraTools for Retopology

The Problem is that all of these tools and pieces of information are divided among panels and settings, that make this "Retopology Process" most difficult and time-consuming.


Now, if all of the stuff was merged together in Poly-Build (Called Retopology Master) would be very intuitive and would give a new process for Blender.

this Tool can be work in this manner

  1. Button for add a "Retopology Mesh in other colours" and set in front 001
  1. Switch to another Workspace called Retopology 002
  1. Work Similar to Quad Draw in Maya --- Maya Quad Draw Link on Youtube

Topology Brush in Zbrush Link on Youtube



Kind Regards,

Ros

I do a lot of retopology in Blender with just F2, shrinkwrap, and the default tools.

Thank you. F2 was designed for both retopology and complex organic modeling.
In fact, Stripe Modeling Workflow is unified workflow both for complex organic modeling and retopology.

All that models are made with power of stripe modeling technique. All this years F2 stripe workflow have been tested for any kind of model complexity production.

Here is free sample to check it out, it have a 500 000 downloads already.

My entire YouTube channel is dedicated to F2 Stripe Modeling workflow.

I agree that it should be part of Edit mode.

Well, it already is)

-Maybe make the F2 addon a default

Well, here are problems we faced during proper testing, like automerge incompatibility, so we have to left it as an addon.
But it's experience can be used for creating proper tools, as far as it was indeed hard tested for all possible cases.

Test process video sample:
https://youtu.be/Y74HmBnydSA?t=380
Finished model (free):
https://www.turbosquid.com/3d-models/free-max-mode-jumbo-regency-tables-reg-14/947889

Some specials:
https://www.blendswap.com/blend/14139

This is possible because Blender is a superior tool for retopology)
Enjoy!

I made an addon that lets you store presets for some of the view and snap settings in the 3d viewport. I also have a script that I use that switches all of the Viewport Display settings needed for object that is going to be used for retopology (In Front and Wireframe enabled). I think this would be a good approach for the problem of the long setup times when switching tasks in the 3d view. Blender could still allow the power of modifying all of the snap and display settings individually (instead of having separate retopo settings), but also have a preset system built in with a few predefined setups - retopo being one of them. This would make switching to retopology workflow as simple as changing the preset in the display settings and the snapping settings.

Beyond this, it would be nice to have an object selector in the overlays dropdown that lets you pick the retopology object, which overrides the viewport display properties of that object with the needed settings. This would keep the user from having to change settings in multiple places in the interface.

https://blenderartists.org/t/presetterprofessional/1164735

Since we are working with pretty much complex models that requires variable density retopology, so we made prototype of [QuadBridge] tool, that can bridge two variable density loops with quads using several topological schemes.
This tool is based on years of experience of quad modeling and retopology workflow.
However, it is still under development. It looks like that.

Can be interesting.
It is mathematically proven that any kind of organic surface can be made with quads, so, I guess, even quad-based dyntopo is possible to achieve at some point.

About "a tool that would let users draw a line, which then becomes a strip of evenly sized quads".
We made such prototype with distance support, but, unfortunately, it turned out to be not very useful on our models.
Anyway, it was nice experience. Here is GIF

@Paul Kotelevets (1D_Inc) Here's a more automatic way with notes but better interpretation of quads to follow the flow of the models, made time ago for sculpt (with some modifiers on it), as it is now it's not really useful for retopo but could be kinda useful with some adaptations and implementations like a new note tool that works (create quads) in real time + radius setting, or quick post-edit tools and some other options and widgets. But doesn't Retopoflow have something similar?

By the way, also made this 1-2 months ago, needs so much improvements but can be a start for something pretty nice, what do you think? https://twitter.com/jfranmatheu/status/1188037853055475712

@Paul Kotelevets (1D_Inc) Here's a more automatic way with notes but better interpretation of quads to follow the flow of the models, made time ago for sculpt (with some modifiers on it), as it is now it's not really useful for retopo but could be kinda useful with some adaptations and implementations like a new note tool that works (create quads) in real time + radius setting, or quick post-edit tools and some other options and widgets. But doesn't Retopoflow have something similar?

Is it an addon? Looking cool)
Retopoflow consists from a nice concepts, but it is pretty much hardware consuming, and it is not applicable to heavyweight models.
That's why we made a simple unassuming to hardware solution that directly supports edit mode and have some visual brush setup to perform adaptive stripe density, so it was the closest solution.

We rarely encounter regular density models, therefore, as a rule, we are forced to process everything manually, so making stripes is the least problem we have got to solve)
For sure, such tool will have its scope, for example, topology sketching, but not in our cases.

UPD

By the way, also made this 1-2 months ago, needs so much improvements but can be a start for something pretty nice, what do you think? https://twitter.com/jfranmatheu/status/1188037853055475712

Looking interesting, but it relies on line drawing that requires a lot of motorics actions, so such workflow will have stamina/endurance limits on massive objects retopology.
We've got the same issue during making Gstretch tool from LoopTools that operates with Gpencil strokes.

For example, we solved that issue by making F2 addon, so we operate with vertices, like in regular modeling.
F2-based Stripe modeling/retopology workflow bypasses this limit.

But it is nice that you are able to make such kind of tools, you've got an impressive programming skills!

When developing the tool, take note of the principles from the Retopo MT addon:
https://gumroad.com/products/cNGNb

When developing the tool, take note of the principles from the Retopo MT addon:
https://gumroad.com/products/cNGNb

Retopo MT follows Retopoflow way. It provides some parametric way of stripe editing, but it's workflow is a bit clunky, since it has not been tested on complex models.
Seems, author also has gone to F2-based workflow type later due to its flexibility.

When developing the tool, take note of the principles from the Retopo MT addon:
https://gumroad.com/products/cNGNb

Retopo MT follows Retopoflow way. It provides some parametric way of stripe editing, but it's workflow is a bit clunky, since it has not been tested on complex models.
Seems, author also has gone to F2-based workflow type later due to its flexibility.

@Paul Kotelevets (1D_Inc) Just in case you didn't notice @Vladimir Spivak (cwolf3d) had already updated to 2.8, you can support him there.

@Paul Kotelevets (1D_Inc) Just in case you didn't notice @Vladimir Spivak (cwolf3d) had already updated to 2.8, you can support him there.

For sure, a lot of addons are updated to 2.8. But update doesnot mean fixing workflow issues.

About diplay issues.


There are also special needs regarding historical restoration retopology - to see a texture of object (so it is needed to be supported) and surface coverage (so overlay display should be optional)
This allows the final surface to be reliable.

Actual mockup

R1 = shading + Texture + Xray
Allows to see final shading

R2 = shading + Texture + Xray + transparency
Allows to see details

R3 = shading + Texture
Allows to see the accuracy of the coverage area

R4 = shading + Texture + transparency
Allows to see the accuracy of the coverage area and details for the allocation

The archaeological reconstruction retopology process is one of the most complex in terms of both shading requirements and attitude to the original and the resulting geometry.
As far as all details are stored in resulting geometry for reconstruction needs, such kind of retopology refers to variable density retopology type.
This type of retopology cannot be automated (yet, with a generic algorithms) due to the high level of semantic significance of the details.
Thus, such retopology process requires the ability to see the original model and overlay in several modes:

  • Textured
  • Xray
  • Shaded
  • Transparency

with ability to quickly switch between their combinations for different purposes.

@William Reynish (billreynish)
Some more notes:

Display
I tried out the retopology overlay (T70267) and it's already a great improvement but I feel like there's something missing for the workflow.
While in Edit Mode with this new overlay it's effortless to see and work on the topology but a common thing I to do in the retopo workflow is to compare the topology with a subdivided, smooth shaded result, possibly with additional modifiers to test the behaviour while retopologizing.
This is commonly done by switching in & out of Edit Mode and having all modifiers visible only in Object Mode.
But since the objects would intersect in object mode it's currently necessary to set the retopo object to be "In Front", which we don't want anymore.

I suggest to add another overlay or shading toggle for Object Mode that will show the retopology object in front of other objects, with the same depth rendering that T70267 does (just without the face color & wireframes to see the final result properly).
This could be context based on selected objects? @William Reynish (billreynish) What do you think?

Tools
The brush-like Tool is one of the most important additions IMO since it removes Sculpt Mode from the Retopo workflow.
Another addition to the design could be a second brush-like tool that functions like the new "Topology Tool" in Sculpt Mode (D6059).
This Tool has quickly become an essential part in my modeling/retopo workflows.

Plz don't make it too complicated with too many Tools, just a few that work well with the combination of modifier keys, so the workflow is easy and streamlined.
if you add too many options it will break the flow when doing retopolog and slows the process alot.

Plz don't make it too complicated with too many Tools, just a few that work well with the combination of modifier keys, so the workflow is easy and streamlined.
if you add too many options it will break the flow when doing retopolog and slows the process alot.

Seems, you are asking for proper workflow design.
Not sure if it is possible.

Proper workflow design means, that there should be a hard test example to evaluate performance of a process (its simplicity, speed, reliability, endurance) with final result.
Well, speaking of manual retopology and organic modeling, conditions are basically known.

Modeling - 100% quads based surface, variable mesh density, single reference, simplest toolset - less than 3 hours for such kind of a model (raw footage).
Image reference - Pearwlorks CE-137a
Workflow test result:
https://youtu.be/Tsaa_D6CcSo

Retopology - 100% quads based surface, variable mesh density, single reference, simplest toolset, maximum details preservation - less than 8 hours.
Free base mesh for test purposes can be downloaded here or here.
Workflow test result:
https://youtu.be/Y74HmBnydSA

The problem is that to make retopology fast, user need to know topology rules.
That means, he should know how to apply basic topology schemes, so it is more about skills than tools.


In addition, there are many influential factors. Even clunky LMB selection can sufficiently reduce retopology speed more, than any kind of tool.
Most tools have never been tested under extreme conditions, that's why it is impossible to find hard tests of concepts like retopoflow, polyquilt, poly buld, maya quad draw related tools, and other handy looking concepts.
Maya quad draw and Topogun tests also shows poor perfomance results (like 4 hours for uniform density croc mesh)

It is too hard to make such tests, it takes just years to figure out the best way, and the bar is too high, so I am not sure that someone can poissibly handle that kind of workflow design task.
So I believe that the common steps, like edit mode topology relaxing brush/methods, reprojection of the retopo surface onto the base mesh methods and enhancing shading modes will bring the most effective predictable result.

Aryan (AryanBro) rescinded a token.
Aryan (AryanBro) rescinded a token.
xan2622 (xan2622) added a comment.EditedMay 9 2020, 1:40 AM

On one hand, it would have made sense to add a workspace dedicated to Retopology, because since version 2.80, Blender has been having many workspace tabs (at the top), one for each step of the 3D creation (Modeling, Sculpting, UV editing, Texture Painting, Shading..) and Retopology is part of it. That being said, consistently switching between a "Retopology" tab to the "Modeling" tab would be time consuming and counter-productive (I am glad that it has been decided to integrate the retopology tools into the Edit mode).


I have tested RetopoFlow 3 beta 2. (Github, Twitter, Installation instructions)
It has many great features (and devs are working hard to make it better) but the way its UI is obfuscating the rest of the Blender UI is very annoying.
You can read my other "grievances" about its interface and why I think it restricts the "freedom of edition": https://github.com/CGCookie/retopoflow/issues/810


There's another great add-on that worths it being tried: Poly Quilt. (Github, Twitter, BlenderArtists forum)
Once it is installed, go to the Add-on Preferences, Extra Settings, Select the "Future" branch, Check for updates, Update, Save Blender preferences and Restart Blender)
It's a very interesting add-on because it allows to perform many actions with just one tool. I like the feature that allows to create a sequence of quads by dragging an edge.


Here are some interesting retopology features (taken from different tools or softwares) that could be an inspiration for Blender's future retopology tools:

That being said, consistently switching between a "Retopology" tab to the "Modeling" tab would be time consuming and counter-productive (I am glad that it has been decided to integrate the retopology tools into the Edit mode).

Well, yes. Retopology is just dedicated type of creating surface process - of modeling,
Expanding modeling to retopology tools expands retopology to modeling tools as well, that allow to perform complex type of modeling like partial retopology, when when details are modeled on top of the basemesh.

AFX, RetopoFlow, Bsurfaces, Polyquilt.

All those tools provide solutions for regular density meshes topology (like head retopology), but retopology is not limited to that type.
The hardest part of retopology is topology with variable density meshes.
Also, those tools have a lot of issues.

Autoretopology.

Proper Autoretopology engine is pretty much tough and separate task, it will be hard to develop something above Instant meshes or Quad Remesher

xan2622 (xan2622) added a comment.EditedJun 4 2020, 9:03 PM

These two videos show how easily it's possible to extrude edges to create faces. Notice how convenient the snapping makes it easy to stick the face to the other quads:
Twitch video clip 1

The second video also shows the creation of faces after extruding edges but also shows how to create faces after adding a vertex:
Twitch video clip 2

The ziRail Maya plugin is a bit similar to TopoGun 3 (it allows to create patch polygons along strokes, strokes that are drawn directly on the mesh):
https://youtu.be/_ehmPPQRtm0 (watch at 0.25 speed 😉 )
https://youtu.be/toCyGQ8wZK8

These two videos show how easily it's possible to extrude edges to create faces.

The problem with this approach is that it is as awkward as shown in the video.

The ziRail Maya plugin

It was done in bsurfaces as creating functionality, then we enhanced it to editing functionality in Looptools Gstretch, then it was broken since 2.8, and it was reinvented by Oscurart
https://twitter.com/Oscurart/status/1264903106581725184
But the problem with this approach is that it does not give a strong advantage over regular extrude and scale.

The problem with this approach is that it is as awkward as shown in the video.

I don't think it's awkward. At all.
IMO, to create faces one-by-one, extruding edges to create faces this way (by snapping the closest vertex to the other geometry) is efficient and simple.

Proper Autoretopology engine is pretty much tough and separate task, it will be hard to develop something above Instant meshes or Quad Remesher

It's going to be a real challenge to beat Exoside Quad Remesher algorithm, but there are some open-source GPL projects that go to the right direction:

https://github.com/huxingyi/autoremesher
https://blogs.dust3d.org/2020/06/19/auto-remesher/
https://blenderartists.org/t/pablo-dobarros-master-plan-for-sculpting-and-painting-development-news/1150731/5829

I don't think it's awkward. At all.

Yes, I know. A lot of people think that such approaches are not awkward, because it is unobvious until you start making retopology at massive scale.
Same issue with, for example, Polybuild and Polyquit tools.

That's the common problem of retopology workflows and approaches.
You have to test it hard before to say that approach is good.

The problem of the edge-oriented workflow is that edge consists from two vertices, and you have to tweak them both to get result, that leads to more tweaking.
Also, edge-oriented workflow relies on extruding, and extruding was never a problem.
So this approach solves unexisting problem.

Probably the best free addon as of now for retopology is PolyQuilt mentioned above: https://github.com/sakana3/PolyQuilt. Its functionalities are simple and effective. The only issue with it is that it does not render properly with the shrinkwrap modifier applied, but you can always use auto snapping as a workaround but still, it would have been great if it did respect shrinkwrap. But its relax-tool is the thing of beauty and something that is very useful during the retopology process and other Polyquilt tool functionalities are basically essential. I'd be glad to see something like the Polyquilt tool implemented in Blender's edit mode out of the box. It would have been perfect, to be honest.

xan2622 (xan2622) added a comment.EditedAug 19 2020, 9:26 PM

QuadWrap looks interesting too:

Click to preview GIF image:


https://blenderartists.org/t/add-on-quadwrap-retopology-tool/1241087/

QuadWrap looks interesting too:

I bought it, it is pretty much early beta.

Anyway, core development is way more about making API that allow to write such functionality, rather than making such functionality, because of wide range of possible realizations of such functionality.