Mirror editing is very broken

[laughsuggestion]

The mirror editing tools are broken. If you model a body, and then two separate leg objects (one mirroring the other) you will notice that they do not create the same geometry on both sides of the body. The body itself is ostensibly already centered, and the legs are equidistant. The issue really is that when we want "mirroring" the need is not for "place this object on the other side of the pivot and then do merge operations". The need is "I am doing this on the left, recreate it identically on the right." The effect we have here is more like a fun-house mirror; it is not a true representation of what's on the other side. Some of this might be small errors in how things are positioned, but diagnosing such errors is basically impossible with the tools we have today. To get the mirroring to be at all useful, you have to export to Blender, cut the entire object in half, and then use mirroring in Blender.

[laughsuggestion]

Expected output:

[huxingyi]

Hi @laughsuggestion thanks for the feedback. The issue is cut face is too little to build the mesh, you can change the cut face, or subdivide the cut face.

Mirror is working on the part level, and part is auto combined with the other parts. The mirrored part is auto combined with the other parts as well, since it's working on the part level, it's not guaranteed the whole model which the part is mirror to will have left and right with the same mesh.

[laughsuggestion]

The mirrored part is auto combined with the other parts as well

As a user, this detail isn't meaningful as a distinction. If I click "mirror" I would expect it to be mirrored. If I create a symmetric object in the middle (the rat body) and then attach a leg, and mirror it, I would expect the "auto combine" to combine in a way that produces identical geo on both sides. That this doesn't function as expected (from a user point of view) means I just have to continue box-modeling in Blender. It's faster to box model in a tool that supports proper symmetry tools than it is to model in Dust3D and then have to fix-up later in Blender.

I get that the mirror is "working" within the parameters you have defined for the problem. What I am telling you is that it does not function in a useful way that allows me to get models to a game-ready state. If this problem is intractable for the way this tool works, that's fine, I have other options. I just wanted to bring it to your attention that for Dust3D to be useful as a game-dev modeling tool, it needs to do some things better than how it is doing them now -- things that other standard tools do well already.

[huxingyi]

Can you provide the .ds3 file for debugging, I will see if the mesh generation can be improved to maintain the exact mirror.

[IndEE777]

Can you provide the .ds3 file for debugging, I will see if the mesh generation can be improved to maintain the exact mirror.

I have this problem too and it's a pretty big problem, for example when modeling a person. What I noticed is that if I put a mirrored body part on the right side, then when connecting it on the mesh drawing, it doesn't reflect the diagonals, but rather connects them from the bottom left corner to the top right corner on the mirror image, which is NOT A MIRROR IMAGE, since the opposite doesn't happen. This way the 2 sides won't be symmetrical, so you can't create a symmetrical body... The body, like the rat's body here, is completely in the middle, with algin, I placed it on the mirror axis, but since it doesn't reflect the mesh drawing, it only creates the object itself and the parts of the fusion without reflecting them, so the model won't be symmetrical either, sometimes we get a convex result on one side and a concave result on the other side at the same time, which is really not good!!! It's a shame, because I really like this program and I wanted to love it, but unfortunately it's not that useful. I also don't understand why options like the hollow and scuplt brush aren't in my latest version when they were in a video from years ago!!!

[huxingyi]

Thanks @IndEE777 for reporting, the asymmetry mostly generated during the removing seam and recombine the meshes after seam remover, will look into it to see if can make the algorithm generate more symmetry triangles.
The features like sculpt which was in early rc version are not in the final version, is because these features are experimental, and didn't got mature enough. What you see now in the released version will not got removed.

[IndEE777]

Thanks @IndEE777 for reporting, the asymmetry mostly generated during the removing seam and recombine the meshes after seam remover, will look into it to see if can make the algorithm generate more symmetry triangles. The features like sculpt which was in early rc version are not in the final version, is because these features are experimental, and didn't got mature enough. What you see now in the released version will not got removed.

This would really be important for the program to be usable! I really like the idea and its simplicity. But symmetry plays an important role in game development! I hope you manage to improve it and have the capacity for it, even though the program is free! It's a shame that more resources aren't allocated to it, mind you, because it's a very promising modeler. I see you've been back to it recently? There was a break in development for a few years? I want to make characters for a relatively lowpoly game, I started with a polygon modeler, but Dust3d would be a much faster and better solution - if that crappy mirroring worked. The problem isn't with the alignment, as you wrote above, that subdivision needs to be turned on, etc., because it can be put there so that it fits well, just like I solved it with plain squares in my picture and it turned out nice, except that the diagonal wasn't mirrored. It's certainly not easy to program such things, but I hope there will be demand and capacity for this to be fixed, because it would greatly affect usability!

[huxingyi]

Hi @laughsuggestion and @IndEE777, can you help to have a try, I have made some improvement, it's not guaranteed to be 100% mirrored, but it should be mirroed in most of the case, as long as the center mesh is symmetry.

Here is the download link:
Linux: https://dust3d.org/download/dust3d-latest-preview.AppImage
macOS: https://dust3d.org/download/dust3d-latest-preview.dmg
windows: https://dust3d.org/download/dust3d-latest-preview.zip

Let me know how you go, really appreciate the feedback.

[IndEE777]

Hi @laughsuggestion and @IndEE777, can you help to have a try, I have made some improvement, it's not guaranteed to be 100% mirrored, but it should be mirroed in most of the case, as long as the center mesh is symmetry.

Here is the download link: Linux: https://dust3d.org/download/dust3d-latest-preview.AppImage macOS: https://dust3d.org/download/dust3d-latest-preview.dmg windows: https://dust3d.org/download/dust3d-latest-preview.zip

Let me know how you go, really appreciate the feedback.

Hi! I tried it, at least for a few seconds here at work, and yes, you can now achieve a real mirror image, although not in all cases, there are cases where they insert an extra part, but where it is symmetrical, the diagonals are really good and in opposite directions! That is a good sign that a mirror image is available, and it only needs to be adjusted a little. If you have any more ideas and can improve it, that is good, but it is already better than it was! Also, I think a snap to grid would be useful. Which would be in the 2D editor image to which the nodes could be connected, it can be optionally switched on of course.
I may be going against your ideas a little, I don't know what the specific plans are! I saw that there is a pistol modeled, and a rifle. For such models, which are also symmetrical and are the results of engineering work in reality, a precision tool like snap to grid would not hurt, so that it is truly geometrically accurate. I thought of something similar to this last time, but now suddenly snap to grid seems logical. I know there's Algin and stuff like that, but it's not quite the same. If I drag a node 1 pixel away, it can twist the whole body and then this reflection can go wrong, so Snap to grid is a faster solution to Algin and we can drag the points freely, we don't have to always center everything. But this is just a suggestion! I'm curious to see what other plans you have. Stitching line/loop is a really good solution! I really like that too!

Also, I tried this now, I copied the mirrored "hand", narrowed it down, then inverted it, it didn't repeat this on the mirror image, even though the inversion object is also mirrored. So it only made one side hollow, the other side remained untouched.

[huxingyi]

Thanks @IndEE777, glad it worked in certain level. The snap to grid is good idea, and it's intuitive to the canvas workflow. Will think about that.
For the not-work case on hollow, if you can move around a little bit to see if it's an bug or nodes placement issue?

[IndEE777]

Thanks @IndEE777, glad it worked in certain level. The snap to grid is good idea, and it's intuitive to the canvas workflow. Will think about that. For the not-work case on hollow, if you can move around a little bit to see if it's an bug or nodes placement issue?

I tried moving the inversion object but I didn't see any change. It had a hole on the right side, it seemed to work, but on the other side it didn't affect it at all. It's really good that there are actually few real functions in the program, but all of them are very good and useful! I'm with Artista in a 2D vector drawing program, I also use it to draw textures, because while you have to fiddle with layers elsewhere, Artista can do what I want to do very well. I also had this observation that although I would do the UV in KHED, the loaded object in khed is so small that I have to zoom in 50x, but the UV is concentrated in one point in the lower right corner of the UV map there. So I can't even pull the points apart, because in KHED the UV scale is on the sides and corners of the bounding box, but since everything is in one point, it's impossible to grasp it. Although if all the ropes break, I can still make a UV map there. Also, KhED is obviously not a standardized modeler, so maybe the problem is there, not here! I'm not that familiar with things.

[huxingyi]

You are not wrong on the UV, since if when there is no texture image specified, pure color of the part are using a small place of the final texture image to hold the color. This will definitely be fixed. Will keep you updated.

[IndEE777]

You are not wrong on the UV, since if when there is no texture image specified, pure color of the part are using a small place of the final texture image to hold the color. This will definitely be fixed. Will keep you updated.

Sorry for the question, but how did you restart the development now? I'm glad to see that many people liked the software. They write good things about it, and the idea is also good, why was it created. Or did the development not stop, there was just a break in uploading videos? I also wrote on YouTube that, for example, the tire on the car wheel was only roughly drawn in a circle and it's not a regular circle, that there might be some tool or preset that could be used to make certain regular shapes, such as a cylinder, or to specify how many nodes it should consist of, or I don't know. I tried it with one node and drilled out the middle with another, but this way you only get a circle if you subdivide the hexagon, but it also rounds the edges a lot, so it's not the best solution, although snap to grid can be useful here, because then you could draw a circle more precisely if you have something to adjust the nodes to.

[huxingyi]

I remember the comment you left. The development in the last couple of years, was paused for a while since relocating and work. However, now it will be continued, recently, there is no new videos been uploaded, but the development is not stopped, will keep uploading videos. Thanks for your feedback, it's very helpful!