Anyone else still need a SUR generator?
-
Well, I agree that LS sur builder is not the perfect tool (and I think he is aware of it). The modders from our mod are also pushing me to finish the tool Gisteron remembered. This would really be a breakthrough. I plan to make time for it as soon as possible, but there is also much work for university this semester. Also the testing of the tree which has to be constructed will take some time. I personally expect it to take at least a month before I can get a version without errors. And there still is the issue that Freelancer could use a modified version of the tree algorithm, but I will cross that bridge when it comes to it…
-
I’d use it in a heartbeat. I already use LS’s builder now for complicated models
-
i looked at the code of LS sur builder its as good as it will get
i experimented about using diffrent algos
but it was always the same results
i did have a thought about some thing
but it prob take a majoir reright
and i didnt test was to take the outermost point and work in the way basicly revers the process
and alway keep the outer points and link them
basicly shrink rap itforgot to add to show you what i mean take a box devide its surface into triangles you see the results but if it came in the way i think it will catch the outer points
-
BBalazs wrote:
Hi ST!
…Hi BBalazs, nice to see you again.
For your SUR problems, see my reply on the new thread “Problems with making SURs? Ask Here!” so we can keep this thread clear for the SUR generator tool discussion.
BBalazs wrote:
However, remember LS’s SUR Builder? The very last version works great. The only problem was that it forgets some meshes - here is the solution: create a copy of your model, use milkshape’s Subdivide 4 on the problematic part’s meshes, export as CMP, and build a SUR for that. As group names are the same, it will work for your original model. And no errors, no nothing. Perfect collission detection.
Exactly. The only problem I have with sur builder is that it does not tell us that it missed a cmp group. It would speed things up dramatically as we would not need to manually check every part on a large model to see that it has a sur part.
Of course once we know which part is missed we can fix it by making more vertices.
And a shield bubble option would make it absolutely great.
BBalazs wrote:
The SUR Builder, nevertheless, has an option for this - in theory. It could remap an existing single-part mesh into a multi-part mesh based on your Groups in Milkshape. By building a new CMP, this feature could counter the CMP exporter’s limitations. Unfortunately, this doesn’t seem to work all the time, particularly if you have a very lot of groups in your single-mesh CMP.
There is a misunderstanding here. SUR Builder cannot separate a model that has been exported as a single group into multiple groups. Single-group / single mesh means the modeller exported the CMP model as “Number of Groups = 1”, instead of “Number of Groups = n” and then setting the “Group Quantities:” to the correct number of parts in each of the “n” groups.
So the same limitation is 18 groups max to export from Milkshape.
Then SUR Builder can see the separate groups.
If the model has more than 1 group, then SUR Builder can either treat the whole model as 1 group and make only 1 sur part to cover all of the model, or generate a sur part for each group separately.
This is what is meant by single-part and multi-part in SUR Builder.
BBalazs wrote:
So, if this unlucky fellow that could never get a working sur made was pulled out of the trouble by LS’s SUR Builder, I’d say it is -the ultimate- SUR tool to all your needs.
Nevertheless, for high resolution SURs (tha need far over 18 convex meshes, like planetary terrain), either a fix to the SUR Builder’s CMP remapping method, or a massive expansion to the CMP Exporter’s part limits would fill in that last gap.
Yes, we’re just back to if SUR Builder at least told us which cmp groups were not generated, then we could easily fix those groups by dividing and convexing them out a little. But it can take a long time to spot which ones have no sur part manually.
TheDVDMan and anyone else for help with surs, please use the new thread “Problems with making SURs? Ask Here!” so we don’t clog this one.
This thread is better to continue discussion on whether anyone feels he can make a new SUR generation tool.
-
Try this…
Click Here to download SB 0.004
Additions:
Selectable script folder, allows the ability to mix dif builds, as in copy good file over bad file, reselect complete sur folder and build sur
Missing hulls (as in no faces) appear in the selected script folder as MissingHulls.log, use this to know which files need replacing -
I get a failed initialisation error. I also tried compying it into the folder of 0.003, to see if that helps, but it didn’t. 0.003 works fine, however. I am probably missing some dlls or libraries that are specific to 0.004 but are not needed by 0.003 - that’s just a guess, though.
-
You need the new DLL pack I made a couple of weeks ago…
-
LancerSolurus wrote:
You need the new DLL pack I made a couple of weeks ago…A DLL pack? Does it contain some plugins or fixes something?
-
It’s the Visual C++ runtime package?..
-
MS updates the DLLs for Visual Studio on a regular basis. I made up a new DLL pack just like the old one. The old one I had made was 3 years old. You olny have to run it one time though, it will istall them in the proper folders.
-
I’ve had some fun with the sur builder. Now, it seems that the multi-part option builds surs based on milkshape meshes (“groups”).
I’ve attached my favourite testing cmp, the 127 milkshape groups terrain and the sur file generated by the sur builder.
When I load the cmp into sur builder, I see it generate the sur meshes for all the groups one-by-one. In the script folder I see the details of the sur mesh - all neatly there.My problem is that if I task sur builder to make the sur, it will succeed in it, however, when I load the model into HardCMP, only one sur part will show - the very first one (or perhaps the last one?) ever created. This is probably because the cmp is actually exported as a single-component cmp, whereas the sur is actually multi-component, with all the components sharing the same name.
Sidenote, that in-game, I can’t see the solar mesh, even if I increase the sur file’s radius.It is just so nice to see those so many sur groups being built. I really feel that there is only some minor thing going wrong. Perhaps if the CMP could be remapped to those 127 parts? Or is it just something in joining up the various parts of the sur?
Nevertheless, nice job LS! Congratulations - your utility is the only way for me to produce 100% working surs, and I actually trust these files for being clean - much better than any splicing method.
-
With the terrain disappearing - did you try opening the cmp in FLModel Tool and saving it immediately with no other changes (this adjusts the size of the bounding box).
-
Nothing to do with LOD (Level of Detail).
But because the bounding box is often incorrectly sized, a new large ship model sometimes disappears from view when you are not looking towards its centre, even though it is next to you.
This fixes that problem sometimes.
-
Well, after resizing (the radius didn’t change that significantly, some 25% different), the terrain was still not there.
I’d assume that flying close to the only part that shows in HardCMP would render it visible, though. Still, when talking about such a large model, this is not as easy as it seems, so I couldn’t verify this.Still, the Sur Builder -can- and -will- build proper surs, even for me
It’s just getting something with more than 18 sur parts to work, that’s hard. -
I want to say the sur builder take in charge vertexes, not edges. LS, I purpose that you can add an edge option to make more realistic hitboxes.
-
The hitboxes you get are absolutely realistic. I’d say they are 99% perfect convex decompositions of your model’s parts.
If you feel the hitbox is not realistic, try braking up the CMP into more parts, or if that’s not the case, I find that a smaller duplicate radius (like 0.1, but try 0 as well, which disables the feature) often helps in improving precision.
Edit: The CMP Remapper (which seems to be disabled in 0.004) worked perfectly for me in the case of most common models. In the case of those that are built of many groups, the remapper also seems to do everything correctly, its only mistake is that it fails to add data04.bin - which makes sense, because as I’ve checked, data04 does not exist (is not created). Do you know what data04 contains, or is there a way to add it manually?
-
Can you explain your edit BBlazs because I don’t under stand you. Wher can I get or find data 04.bin please?