I have no scripting skills at all and started looking into it today, to automate some boring stuff in blender and is blowing my mind how easy is to make simple addons.
One thing that i would want to do is to control the bevel modifier just like the bevel tool, where i could control width and segments dynamically, is this some heavy stuff or is somewhat simple?
@Udjani not a real useful answer perhaps, but if i understand you correctly, that's what some addons like HardOps do by default with the Bevel modifier operation. If you already own these, perhaps there's your answer?
I'm only saying this in case you were'nt aware of it, though you probably already are ^^'
@Udjani accessing modifier information is quite easy. You can test it using by selecting an object with a bevel modifier and change the width using the console.
@Justo Yep, i am aware of it, just thought that would be a good idea to invest some time learning this, so i don't need to download addons with dozens of functionalities just because i need one of them. Also a good thing for stability probably, i bought an addon some time ago and the download option disapeared from my gumroad page. xD
@Prime8 Yes, the way to change the values is the one that i was worried about, a better example would be with an array modifier. Hit a key to add the modifier then use the scroll wheel to control the ammout of duplicates. But this might be a littler more advanced, will see how deep i can go into it without learning how to code.
Just found this kickass little site with short text tutorials, which I personally prefer over going through videos: https://twominutetuts.com It's great! Let's share it with the people who're getting into Blender.
Yes, the way to change the values is the one that i was worried about, a better example would be with an array modifier. Hit a key to add the modifier then use the scroll wheel to control the ammout of duplicates. But this might be a littler more advanced, will see how deep i can go into it without learning how to code.
You can open the Text Editor and under Templates -> Python -> Operator Modal you get a simple template of a modal operator which is what you'd need to do something like this. In the invoke() method you can do some initialization such as adding an Array modifier to the object; in the modal() method you can react to input events such as changing a modifier parameter based on mouse delta movements. With this you can test the core of your scripted operator, but it won't be something you can share with others until you add a bl_info block in your script, as that is needed to make it installable and appear in the Add-ons tab of user preferences.
Well, you could share it without the bl_info, but then everyone would have to paste it into their Text Editor views and press the Run Script button in the editor header to 'register' the operator for their Blender session -- the next time they open Blender they'll have to do it again. If it's for something quick (a one-off script to fix something in the scene) then it could be delivered this way anyway...
Also, I don't know if you guys have been dealing with this same problem, but I'm having a extremely difficult time baking proper colors from Blender's materials. (The colors are always darker than the ones on Blender viewport when baked down in ID settings > colors from materials in Substance Painter.) Both on .obj and . Fbx
And its probably related to this: https://developer.blender.org/T53209 Looks like the dev didn't bother making the diffuse material export perfect... Which is a shame, using material colors as a base paint layer is a huge time saver instead of totally re-working on your object assigned colors within painter.
Would love to have some light shining on that issue
I love Bruno's concepts as well, especially what he did on Deus Ex. Your model looks great, though only some parts have roughness variance from weathering, is that on purpose?
Blender file using 8gb of ram, i have a scene with 100+ objects, they all have mirror and subdiv modifiers, but only 5m tri on viewport. I made a quick test on a new file with 50m tri and blender only uses 7gb. So what is going on?
I love Bruno's concepts as well, especially what he did on Deus Ex. Your model looks great, though only some parts have roughness variance from weathering, is that on purpose?
Yes Looking forward to improve on my upcoming works regarding time balance to narrow a proper materials workflow. As said, Blender <> Substance Painter isnt the best combination for me at the moment.
@jeperez Hi! I think your Gundam RX fanart is a clear improvement, using more complex shapes. But I think the lighting could be improved. Can you see it? With that high-key lighting setup it's lacking contrast or balance, it's kinda harsh. You can give it a lot more mood / atmosphere with the right lighting setup, and it helps with storytelling. For example, the mech is in a hangar being lit by spotlights and sinister rim light (this also gives you ideas of camera placement, like the viewpoint is from an engineer looking up at the mech).
If you make a thread over at the 3D Art Showcase & Critique area with all your images (like a "jeperez mechs" thread or something like that), I guarantee you will get better feedback than mine.
@jeperez Hi! I think your Gundam RX fanart is a clear improvement, using more complex shapes. But I think the lighting could be improved. Can you see it? With that high-key lighting setup it's lacking contrast or balance, it's kinda harsh. You can give it a lot more mood / atmosphere with the right lighting setup, and it helps with storytelling. For example, the mech is in a hangar being lit by spotlights and sinister rim light (this also gives you ideas of camera placement, like the viewpoint is from an engineer looking up at the mech).
If you make a thread over at the 3D Art Showcase & Critique area with all your images (like a "jeperez mechs" thread or something like that), I guarantee you will get better feedback than mine.
Thank you very much RN , i'll use this feedback to improve my scene, i'm already using hdri but i'll improve my lighting and will post a update as soon as possible,i'll take your advice and will create a thread on the next time, thank you very much! i have a problem once i start to model, i get frustrated soon and try to rush finish the scene and model as soon as possible, do you have some advice for that? i guess that is why all my models have a similar look and feel thank you in foward
Any one here using E-cycles? How does this compare to Blender with Optrix denoiser? Here is a Test using Optrix denoiser for Nvidia RTX cards in Blender 2.83: https://www.youtube.com/watch?v=L6MCRMike8o
Hi, I have a question regarding flipbook creation from simulations inside Blender. I am using 2.82 btw. So I want to create fire simulation coming from rocket engine. There are tons of tutorials about that and this part I know how to do. But in the end I want to put this simulation into flipbook or sprite sheet and use it inside Unreal Engine and Unity as particle effect. So is there any fairly automatic and easy way to do this? I am asking here because all I found is simulations inside blender but no way to convert it into game engine.
Loving 2.83 and going to stick with it for a long time. It's been great to watch blender develop into the amazing tool it is today. Throwback to my thread in 2016 "Do any good artists use Blender?" when I had low expectations lol.
I'd like to ask if anyone has good recommendations for evee or cycles tutorials I'd much appreciate them.
I hope they help you, I found them a great starting point. If any has any tips on where to go next I would love to hear them.
My
advice for anyone switching 3d packages would be not to adjust the shortcuts or switch the interface to be like
Maya/3DSmax or you will find none of the tutorials will make sense, try and get used to the
short cuts. While I found the expeience really painful and slow to start off as its very different to Maya, 3DSmax and Houdini it’s just like learning to
play an instrument and it will speed up your workflow for modelling loads in
the long run once you start to master the hotkeys keys (funky middle mouse translations, Tab to enter in and out off edit mode) and so on I promise you once you start to put it all to memory muscle it really does start to become a bit of a joy to work with.
While I’ve only been playing with
Blender for a few days its starting to make sense. I’ve got a long way to
go but I can see some real potential with it.
This video gives a great overview of the new features
I was wondering if I could get some help/advice for my rig. As you can see I'm rigging a horse body, and it's my first time rigging a quadraped so I'm a little stuck with this IK issue.
For the front leg, I have the IK bone in the same position as the hoof bone, with a custom shape. I've bent the joints in the direction I want them to move in, but the issue comes in when I set my chain length. Because of how a horse moves, I want to be able to move the lower shoulder bone, but selecting it in the chain makes the IK line go in front of the knee joint, thus making it bend backwards. Is there a way to remedy this, either by adding the shoulder movement in a separate method or changing the position of the IK constraint?
I have the same set up for the back leg, but due to its anatomy the IK line sits on the appropriate side of each joint, thus creating no issue.
I'm new to rigging in Blender, so I don't quite know all of the available constraints/what they do yet
That line is just a visualization of the range of bones in the chain. It does not indicate the direction that the bones will bend, which as you noticed is defined by the bend that they already have at bind pose. You can change this by rotating bones in Edit mode until they have a slight bend in the direction you want them to go with the IK solver.
I don't know what kind of behavior you're looking for, but to me it would be okay to leave the shoulder bone outside of the leg IK chain, so the animator can control the shoulder by itself, it's what I see in humanoid rigs at least:
Other things:
- The IK target bone should not be parented to any of the bones in the chain. It can be parented to the root bone for example.
- Only the location of the head (the pivot point) of the IK target bone is used. Its rotation makes no difference, so it's useful to keep it straight (aligned to one of the world axes) to make it easier to give it a custom shape.
- In some cases the per-bone Inverse Kinematics settings can be useful. They're in the Bone Properties tab of a bone in Pose mode. You can set rotation limits, lock an axis and set stiffness:
@RN Here's a gif of what happens with my leg. When the chain is set to 3 it works as intended, only without the shoulder movement. When I increase the chain to 4 to get the shoulder movement the knee no longer bends as desired. I included the line in my earlier image because I had thought that the placement of the joints on each side of the line affected what way they bend.
I would like the shoulder to be part of the IK for realistic movement without having to adjust the shoulder separately, as you can see in the gif above the shoulder moves with the leg.
The original IK was a separate bone, just in the same position as the hoof bone - it was a technique I followed from a tutorial but keeping it straight is probably for the best.
I've downloaded the latest stable version, 2.83, and nothing was done about the issues with subdivision modelling. We still have horrible perfomance, same modifiers, all worse than in 3ds Max (2.9 brings back a similar perfomance of 2.7x, but it's not enough). The multires modifier also offers a shitty perfomance (the new multires in 2.9 only works in sculpt mode, and also lacks perfomance, a lot). Sculpting can't be compared to Zbrush or Modo's one.
The app is still a low-class app in perfomance terms, It's impossible to work with it. And you wonder why Blender is still not used in too many studios?.
Too many minor improvements in this release, but not what is really important. The issue was adressed and closed as "solved". Resignation is what i have.
2.9 alpha is there, but they improved it just a little. I'm sad.
A shitty box like that, with the new subdivision modifier in 2.9, is very slow, similar to what we had with 2.78. I'm using 2 pcs, one with a GTX 1060 and other with a GTX 1070ti.
Finally we can invert vertex groups on most (all?) modifiers with 2.83. I was often wondering why this wasn't implemented before. @Blaizer Blender definitely has some performance issues, fortunately some improved over the last releases, others didn't, but you make it sound like Blender is completely useless. When does the subd calculation run on the GPU when modifying the cage?
@Prime8 Sorry if i sounded in that way, it wasn't my intention. I was just referring to subdivision modelling. Blender may not be the best 3d app, but it has some cool features, that's for sure. For subdivision modelling is very disappointing, and i would say useless for that aspect/work.
I really don't know what happens with the GPU performance in Blender 2.90. Modo 302 can edit that simple box with 8 iterations like nothing (Blender 2.90 can't move and edit that simple cube with less iterations, 4, something worrying for me).
It was said on the Blender Dev forum they were working on the subdivision modifier (the past year), and that the were making too many optimizations in order to take advantage of actual gpus, but it's like smoke for me. A Ryzen 3900x and a GTX 1070ti is not good enough?
I would like the shoulder to be part of the IK for realistic movement without having to adjust the shoulder separately, as you can see in the gif above the shoulder moves with the leg.
@atxr Thanks for the GIF. I can see now that what you want is contradictory: you want the shoulder to be included in the IK chain and for it to move, but you also want the knee to bend.
The IK algorithm solves the problem of "how to best rotate the bones in the chain so that the leaf bone (the last one) can follow the IK target". It tries to do that with the least effort possible, so it must've found that rotating the shoulder joint does most of the work and that the knee joint doesn't need to do as much. I don't think it's possible to change this.
I've never seen a humanoid / bipedal rig where the shoulder was inside the arm's IK chain. The shoulders were always separate from that, and controlled by FK (direct control by the animator). Then when you're animating you do it in two steps: posing the FK shoulder and then posing the IK/FK arm.
A horse is a ungulate (image taken from this page), the joints are still the same as in a human leg. So if I had to rig that horse I would do it like a human.
The original IK was a separate bone, just in the same position as the hoof bone - it was a technique I followed from a tutorial but keeping it straight is probably for the best.
You can do both, keeping the position the same as the last bone, but then use any rotation at all --since it doesn't matter for the IK target-- like keeping it straight to make it easier to model a custom shape for it.
@Blaizer If I try to edit the same mesh you showed in your screenshot with the same modifier on 2.83 (I did not try 2.90 yet), I do get FPS drops as well. But as soon as I turn down the view port iterations to 3, I hit 60 FPS no problem and there is no visual difference apart from maybe a few pixels in the contours. I am not saying this couldn't be better, but do you really NEED that many iterations for seeing what you are editing?
@Blaizer I didn't follow the subdiv development closely, I have to admit, if I remember correctly they switched to opensubdiv in 2.8. I cannot find any settings in 2.83. In 2.79 the old modifier was faster when editing, I can understand your frustration if that is part of your workflow.
Quick question. Lets say I have a single object in a certain location, and created a circular array with it (using an empty object for object offset). After that array in the modifier stack, I have created a lattice modifier, and distorted the array using a lattice.
When I now want to edit the original single object and switch to edit mode, the array snaps back into a position in 3d space using the position of that original mesh (lattice deformation is ignored). Only when I switch back to object mode, the array takes the lattice deformation into consideration. Adjusting this requires a certain amount of 'blind flight' as you can not see the actual final result.
Is there some way to preview that lattice deformation? Meaning, can I somehow display the object/ array with lattice deformation applied (e.g. an additional, ghosted view of the geometry) when I switch to Edit mode?
@Blaizer I'm a little confused. I just tested this on my build—also GTX 1060, but paired with a i5 6600. It should perform worse than yours.
Here's what I see on 2.83. You said this version was particularly bad at this but it's not an alpha, so should be preferred for tests.
While I got a slight performance drop imo it's not unusable to the point of name-calling—this I reserve for when working with particle systems! It should also be noted I'm already using roughly 40% of my pc resources for another tasks.
I suspect something's up with your preferences. Double check everything is correct: Using all cores or auto, GPU selected... Also check if your drivers are up to date and try a .zip install without porting preferences or add-ons instead of updating a past version to avoid carrying any usage-induced glitch over. If the performance issue goes away you can try to port preferences, then enable add-ons one by one, because it's likely one of them causing the issue.
And in case you don't already do this, follow @Prime8 advice. Most people do subsurf modelling on Blender, and they use the modifier instead of subdividing the model to take advantage of the configurable viewport levels and quality in beta versions.
There is still no way to make a shift-extrude, like in 3D max, without programming? I really miss this one.
Yeah, there doesn't seem to be a design task for it on the Blender maniphest. You could upvote this suggestion on the right-click select site if you haven't already. I don't miss that functionality though since I don't use gizmos in Blender. :S Using a shortcut for extrude faces along normals is very fast and offers fine control over the ctrl+RMB extrude.
@birb : this issue is that these topics are highly subjective relative, and also not quite debug-able like a regular bug would.
For instance your GIF seems to show that there is no issue. But is it actually the case in a practical scenario ? The fact of the matter is ... what could be perceived as harsh "name calling" on @Blaizer 's part is actually completely justified, as the performance on something as simple as a simple bust with lvl 2 subdiv on it is indeed unacceptable. I'd go as far as saying that it is actually noticeable even on your GIF, as things are lagging behind the cursor and you are stating yourself that you are getting some performance drop. On a simple subdivided cube !
Here's something that shows the issue clearly : viewport navigation is butter smooth, yet merely moving verts/quads slows things down to a crawl making it litterally unworkable for anyone used to working lighting fast. All that on a very simple bust model.
Of course there is always going to be people claiming that "it's just fine" - but that's usually people who don't have much expertise on the topic, or, people with high level of tolerance for slow workflows (like some Zbrush artists or Substance users who are totally fine with extremely poor viewport response because that's probably all they know so they're used to it). But to anyone who not only enjoys smooth response time but also actually *relies* on it to get work done it is extremely frustrating.
(Now that said I am admittedly running a somewhat unconventional setup : recent video card, but 10y old motherboard fitted with the most powerful CPU available for it. But that's somewhat the point - other applications do not show that sort of issues to begin with, so just throwing more hardware at the problem is not the right way to go. Something somewhere is just poorly optimized/not accelerated, especially in contrast with the progress made on the sculpting side of things.)
The problem is that merely voicing such frustrations doesn't help one bit, so that's why I would personally highly encourage people like Blaizer to not just repeatedly mention the bad performance, but also document it *comparatively* by recording the same operation in Blender and then in other apps with better performance (Modo, Silo, Max, and so on). Otherwise there is no point of comparison and it just goes in circles.
- - - - -
The bottom line is that whoever is in charge of this aspect of the program probably doesn't have the perspective/experience of a real user - as opposed to how the sculpt mode is being developed by an actual user of it. Unless that changes I wouldn't expect any performance improvement on that side of things any time soon, sadly.
Yup, that's been a problem since 2.80 and I also think Blaizer's complaint is justified. It definitely gets worse the more dense the mesh is and the more modifiers are stacked such as mirror and array. It should be a lot more responsive in 2.79. I've gotten used to modeling with the subsurf modifier disabled though and intermittently check the result. This is probably cuz I model the high and low poly at the same time for hard surface models. It might be a bad habit though (inaccurate initial placement of support loops) and I'm curious whether you guys prefer modeling with the subdivided result visible?
@pior When I said I was confused I wasn't being sarcastic or anything—I am genuinely confused whether the performance I showed was what he considered unacceptable. That's why went ahead and I recorded it. Because if it's not then there's something else contributing to the performance issues he described, making it solvable (or at least less noticeable!) without depending on BF.
Usually I just avoid engaging in this kind of discussion, but his computer spec is relatively similar to mine and the test case was simple to reproduce. It wouldn't hurt to test it and potentially help someone. Excluding install issues modifiers stack also impact performance, including disabled modifiers, and that's another thing you can change to minimize the performance drop. But to change it you need to be aware of it in first place.
And personally I agree about documenting the issue and contrasting with other examples. BF is an open-source project. Their growing userbase and limited resources make them very picky about what they'll prioritize, and I'd say they can be frustratingly prone to closing down even well documented issue reports. So if you want results your best bet is to be kind to the devs by being super through to minimize their workload. Like... don't even show videos of someone modelling something random on whatever software you're comparing to Blender and tell them to watch the X minutes of it, create a simple test case and record exactly what you're describing. Create a test file for blender with the test case, record a comparison and share the file. Be as objective as possible, this type of thing.
@birb , my thoughts exactly ! Here's good old Silo that I reinstalled just for that. Note that the recording is capped at 60fps but the in app performance is probably higher.
(how the hell to you embed youtube videos here ??)
Of course "workable" vs "unworkable" is subjective. But personally I wouldn't wish this kind of editing performance on anyone ... Personally that doesn't affect me much as I barely model these days, and if I do I'd be mostly relying on marked hard edges / marked creases / marked bevels. But TBH the poor Blender performance with subd is probably what pushed me towards these techniques in the first place.
@f1r3w4rr10r@birb i have installed both versions, 2.83, and 2.90. I downloaded the 2.90 alpha version to check out the new modifiers because i was said to do it. I complained about the poor performance of 2.83, and like i said, 2.90 has a similar performance as Blender 2.78 (opensubdiv was used in that version).
Just subdivide the box one more time, and you'll see how terrible is all. All the responsiveness is lost. In the other example, Blender is at the "limit".
If with a simple box we have issues, imagine a character model of 200k+ polygons and 2 iterations. I've loaded several models, and i simply can't work with them. It's like playing a game with 2-3 fps and micro freezings, when modo is having more than 200 fps.
Working with "raw polygons" is a not an option for me. I've been using subdivsision modelling with Silo, Modo, Hexagon and other apps since year 2004 without these problems (with a shitty PC of that year), and i'm not going to change the way i work because a software is not well coded, nor optimized for a simple and very old task. We are in year 2020, not 2001.
Just re-watch the video Pior linked, and imagine a full body. If Silo is a very good app, imagine Modo. In Gacha terms, Modo would be like class SSR, Silo A, and Blender F. I was expecting to have a Silo Performance in Blender, but it's still a dream with version 2.90.
This issue was already described, documented, re-described, reported, and re-reported, and .blend files were attached to reproduce the problem in several ways. And not 1 time, but a hundred. I'm bored of it already, and i'm fed up of this. All case supports were mergued into one, and filed as "problem solved" because they are like sceptical with all, they think performance is more than enough (like you, birb). They don't want to see the problem because it's very complex, and at the end, the most important issues are not adressed.
I can't do much more, except making noise about an issue, here, and there. Don't expect me to act as a Blender fanboy, i'm pretty old for that .
Replies
One thing that i would want to do is to control the bevel modifier just like the bevel tool, where i could control width and segments dynamically, is this some heavy stuff or is somewhat simple?
I'm only saying this in case you were'nt aware of it, though you probably already are ^^'
bpy.context.selected_objects[0].modifiers['Bevel'].width = 0.04
Writing the code needed to change the values in the way you want it is probably more the issue.
@Prime8 Yes, the way to change the values is the one that i was worried about, a better example would be with an array modifier. Hit a key to add the modifier then use the scroll wheel to control the ammout of duplicates. But this might be a littler more advanced, will see how deep i can go into it without learning how to code.
It's great! Let's share it with the people who're getting into Blender.
In the invoke() method you can do some initialization such as adding an Array modifier to the object; in the modal() method you can react to input events such as changing a modifier parameter based on mouse delta movements.
With this you can test the core of your scripted operator, but it won't be something you can share with others until you add a bl_info block in your script, as that is needed to make it installable and appear in the Add-ons tab of user preferences.
If it's for something quick (a one-off script to fix something in the scene) then it could be delivered this way anyway...
model
Check out Bruno's concepts in here. Very fun to train hard surface designs, totally recommend giving one of those Mechs a shot.
Full Cycles Renders on Artstation
And its probably related to this:
https://developer.blender.org/T53209
Looks like the dev didn't bother making the diffuse material export perfect...
Which is a shame, using material colors as a base paint layer is a huge time saver instead of totally re-working on your object assigned colors within painter.
Would love to have some light shining on that issue
Your model looks great, though only some parts have roughness variance from weathering, is that on purpose?
Ivan tantsiura:
https://www.artstation.com/artwork/6am3ZN
some of my own inspiration:
https://www.artstation.com/artwork/lVJJea
some from warhammer tau empire:
https://www.artstation.com/artwork/lVgrzY
Looking forward to improve on my upcoming works regarding time balance to narrow a proper materials workflow. As said, Blender <> Substance Painter isnt the best combination for me at the moment.
Alternatively you can try a setup using the "color burn" node, has a similar effect. Example with procedural textures.
from gundam RX series thank you @RN
You can give it a lot more mood / atmosphere with the right lighting setup, and it helps with storytelling. For example, the mech is in a hangar being lit by spotlights and sinister rim light (this also gives you ideas of camera placement, like the viewpoint is from an engineer looking up at the mech).
Here is a Test using Optrix denoiser for Nvidia RTX cards in Blender 2.83:
I'd like to ask if anyone has good recommendations for evee or cycles tutorials I'd much appreciate them.
Hello
I've just started learning Blender this week and I am very impressed.
If your about to start on the Blender journey I would recommend the following.
I hope they help you, I found them a great starting point.
My advice for anyone switching 3d packages would be not to adjust the shortcuts or switch the interface to be like Maya/3DSmax or you will find none of the tutorials will make sense, try and get used to the short cuts. While I found the expeience really painful and slow to start off as its very different to Maya, 3DSmax and Houdini it’s just like learning to play an instrument and it will speed up your workflow for modelling loads in the long run once you start to master the hotkeys keys (funky middle mouse translations, Tab to enter in and out off edit mode) and so on I promise you once you start to put it all to memory muscle it really does start to become a bit of a joy to work with.
While I’ve only been playing with Blender for a few days its starting to make sense. I’ve got a long way to go but I can see some real potential with it.
This video gives a great overview of the new features
https://blender-addons.org/
Blender 2.90 Alpha
The app is still a low-class app in perfomance terms, It's impossible to work with it. And you wonder why Blender is still not used in too many studios?.
Too many minor improvements in this release, but not what is really important. The issue was adressed and closed as "solved". Resignation is what i have.
2.9 alpha is there, but they improved it just a little. I'm sad.
A shitty box like that, with the new subdivision modifier in 2.9, is very slow, similar to what we had with 2.78. I'm using 2 pcs, one with a GTX 1060 and other with a GTX 1070ti.
@Blaizer Blender definitely has some performance issues, fortunately some improved over the last releases, others didn't, but you make it sound like Blender is completely useless.
When does the subd calculation run on the GPU when modifying the cage?
I really don't know what happens with the GPU performance in Blender 2.90. Modo 302 can edit that simple box with 8 iterations like nothing (Blender 2.90 can't move and edit that simple cube with less iterations, 4, something worrying for me).
It was said on the Blender Dev forum they were working on the subdivision modifier (the past year), and that the were making too many optimizations in order to take advantage of actual gpus, but it's like smoke for me. A Ryzen 3900x and a GTX 1070ti is not good enough?
When I now want to edit the original single object and switch to edit mode, the array snaps back into a position in 3d space using the position of that original mesh (lattice deformation is ignored). Only when I switch back to object mode, the array takes the lattice deformation into consideration. Adjusting this requires a certain amount of 'blind flight' as you can not see the actual final result.
Is there some way to preview that lattice deformation? Meaning, can I somehow display the object/ array with lattice deformation applied (e.g. an additional, ghosted view of the geometry) when I switch to Edit mode?
Here's what I see on 2.83. You said this version was particularly bad at this but it's not an alpha, so should be preferred for tests.
While I got a slight performance drop imo it's not unusable to the point of name-calling—this I reserve for when working with particle systems!
I suspect something's up with your preferences. Double check everything is correct: Using all cores or auto, GPU selected... Also check if your drivers are up to date and try a .zip install without porting preferences or add-ons instead of updating a past version to avoid carrying any usage-induced glitch over. If the performance issue goes away you can try to port preferences, then enable add-ons one by one, because it's likely one of them causing the issue.
And in case you don't already do this, follow @Prime8 advice. Most people do subsurf modelling on Blender, and they use the modifier instead of subdividing the model to take advantage of the configurable viewport levels and quality in beta versions.
I really miss this one.
For instance your GIF seems to show that there is no issue. But is it actually the case in a practical scenario ? The fact of the matter is ... what could be perceived as harsh "name calling" on @Blaizer 's part is actually completely justified, as the performance on something as simple as a simple bust with lvl 2 subdiv on it is indeed unacceptable. I'd go as far as saying that it is actually noticeable even on your GIF, as things are lagging behind the cursor and you are stating yourself that you are getting some performance drop. On a simple subdivided cube !
Here's something that shows the issue clearly : viewport navigation is butter smooth, yet merely moving verts/quads slows things down to a crawl making it litterally unworkable for anyone used to working lighting fast. All that on a very simple bust model.
https://www.youtube.com/watch?v=T0iz2JWOPkU
Of course there is always going to be people claiming that "it's just fine" - but that's usually people who don't have much expertise on the topic, or, people with high level of tolerance for slow workflows (like some Zbrush artists or Substance users who are totally fine with extremely poor viewport response because that's probably all they know so they're used to it). But to anyone who not only enjoys smooth response time but also actually *relies* on it to get work done it is extremely frustrating.
(Now that said I am admittedly running a somewhat unconventional setup : recent video card, but 10y old motherboard fitted with the most powerful CPU available for it. But that's somewhat the point - other applications do not show that sort of issues to begin with, so just throwing more hardware at the problem is not the right way to go. Something somewhere is just poorly optimized/not accelerated, especially in contrast with the progress made on the sculpting side of things.)
The problem is that merely voicing such frustrations doesn't help one bit, so that's why I would personally highly encourage people like Blaizer to not just repeatedly mention the bad performance, but also document it *comparatively* by recording the same operation in Blender and then in other apps with better performance (Modo, Silo, Max, and so on). Otherwise there is no point of comparison and it just goes in circles.
- - - - -
The bottom line is that whoever is in charge of this aspect of the program probably doesn't have the perspective/experience of a real user - as opposed to how the sculpt mode is being developed by an actual user of it. Unless that changes I wouldn't expect any performance improvement on that side of things any time soon, sadly.
Usually I just avoid engaging in this kind of discussion, but his computer spec is relatively similar to mine and the test case was simple to reproduce. It wouldn't hurt to test it and potentially help someone. Excluding install issues modifiers stack also impact performance, including disabled modifiers, and that's another thing you can change to minimize the performance drop. But to change it you need to be aware of it in first place.
And personally I agree about documenting the issue and contrasting with other examples. BF is an open-source project. Their growing userbase and limited resources make them very picky about what they'll prioritize, and I'd say they can be frustratingly prone to closing down even well documented issue reports. So if you want results your best bet is to be kind to the devs by being super through to minimize their workload. Like... don't even show videos of someone modelling something random on whatever software you're comparing to Blender and tell them to watch the X minutes of it, create a simple test case and record exactly what you're describing. Create a test file for blender with the test case, record a comparison and share the file. Be as objective as possible, this type of thing.
Here's good old Silo that I reinstalled just for that. Note that the recording is capped at 60fps but the in app performance is probably higher.
https://www.youtube.com/watch?v=vBAq7UYmWdY
(how the hell to you embed youtube videos here ??)
Of course "workable" vs "unworkable" is subjective. But personally I wouldn't wish this kind of editing performance on anyone ...
Personally that doesn't affect me much as I barely model these days, and if I do I'd be mostly relying on marked hard edges / marked creases / marked bevels. But TBH the poor Blender performance with subd is probably what pushed me towards these techniques in the first place.
Just subdivide the box one more time, and you'll see how terrible is all. All the responsiveness is lost. In the other example, Blender is at the "limit".
If with a simple box we have issues, imagine a character model of 200k+ polygons and 2 iterations. I've loaded several models, and i simply can't work with them. It's like playing a game with 2-3 fps and micro freezings, when modo is having more than 200 fps.
Working with "raw polygons" is a not an option for me. I've been using subdivsision modelling with Silo, Modo, Hexagon and other apps since year 2004 without these problems (with a shitty PC of that year), and i'm not going to change the way i work because a software is not well coded, nor optimized for a simple and very old task. We are in year 2020, not 2001.
Just re-watch the video Pior linked, and imagine a full body. If Silo is a very good app, imagine Modo. In Gacha terms, Modo would be like class SSR, Silo A, and Blender F. I was expecting to have a Silo Performance in Blender, but it's still a dream with version 2.90.
This issue was already described, documented, re-described, reported, and re-reported, and .blend files were attached to reproduce the problem in several ways. And not 1 time, but a hundred. I'm bored of it already, and i'm fed up of this. All case supports were mergued into one, and filed as "problem solved" because they are like sceptical with all, they think performance is more than enough (like you, birb). They don't want to see the problem because it's very complex, and at the end, the most important issues are not adressed.
I can't do much more, except making noise about an issue, here, and there. Don't expect me to act as a Blender fanboy, i'm pretty old for that