Hey @WickedP, Thank you for reaching out to us. This is rather hard to answer as is. It is not that your question would be inadequate or the code example unclear, but this is also a subject I would have to try out myself to know for sure. The problem with pseudo code is that it is not executable, when you want a definitive answer, I must ask you to provide compileable code which demonstrates the issue. About your Issue Since you say that sending such image file to the Picture Viewer works, we can rule out that you made any mistakes in setting up the Filename in memory mode or properly encoding your image data. What is noteworthy is that Filename is a Cinema API type from the 'old' days which is today just a hollowed out shell that wraps UrlInterface. When using the memory feature of Filename you are actually using maxon::URLSCHEME_MEMORYFILESTRUCT (not to be confused with URLSCHEME_MEMORY). Internally we tend to use yet another file virtualization scheme implemented for URLs when embedding virtual files in scenes or renderings, maxon::URLSCHEME_RAMDISK. There could be a threading issue with your Filename (check if the issue also happens when you pass RENDERFLAGS::NODOCUMENTCLONE to your RenderDocument call). The most forward fix would be however to use a ram disk, as that is what is proven to work with render engines. The problem is that there are currently no public examples for creating URLSCHEME_RAMDISK urls in the public SDK (as it is sort of a semi-official thing). The only place where I used it was in the Asset API docs, but only to ensure ram disk assets are being cached, not to create new ones. But ram disk scheme urls have a convenience interface, RamDiskInterface which makes it quite straight forward to use. Find below some mock code for your use case (did not compile it due to the lack of a project for this topic). Cheers, Ferdinand using namespace maxon; Result<void> DoRamDiskStuff() { iferr_scope; // Error handler for this scope because we are using the Maxon API. When you want to merge // this with Cinema API code, i.e., use it in a function that is not of type Result<T>, // you would have to use here iferr_scope_handler to terminate errors into the return // type of your function. // Create a new mounting point for ram disk files. Use a unique ID in reversed domain name notation // to avoid conflicts with other mounting points or use the other constructor to create an automatic // hashed mounting point (which will not be persistent over reruns). RamDiskRef mountingPoint = RamDiskInterface::Create("net.mycompany.ramdisk.foo"_s) iferr_return; Url root = mountingPoint.GetRoot() iferr_return; // There are now basically two ways to create ram disk files: // (1) direct creation where we effectively treat the Url just like a file. // (2) lazy creation where we provide a callback that will be called when the file is first accessed. // Our payload. const Char payload[] = "Hello World!"; // Direct creation: // We create a file under our ram disk mounting point, it would have an URL such as // "ramdisk://.../file.txt". You can directly use this URL in all parameters and file // operations. We can also turn an Url back into a Filename using cinema::MaxonConvert. Url directFile = (root + "direct_file.txt"_s) iferr_return; // Now we write to the file, this is just standard Maxon API file I/O. InOutputStreamRef inout = directFile.OpenInOutputStream() iferr_return; inout.Write(payload) iferr_return; // And this would be the way to read back. Char buffer[SIZEOF(payload)]; inout.Seek(0) iferr_return; inout.Read(buffer) iferr_return; // Deferred/Lazy creation: // This way is in a certain sense a bit easier (as stream handlers can be a bit scary when new // to the Maxon API). The advantage is of course that we only pay the costs when something is // actually needed. The downside is that is this can happen at an inconvenient time. Which is // also why I showed in the Asset API examples for how to build caches in advance. But in your // case where you create the URLs yourself, this makes no sense. // This is just this one call which we invoke on our ram disk mounting point. We provide a callback // that will be called when the file is first accessed, and in this callback we create an IoMemoryRef // with our payload and return it. Url lazyFile = mountingPoint.CreateLazyFile(ToSingletonBlock("lazy_file.txt"_s), [payload] () -> Result<IoMemoryRef> { iferr_scope; DiagnosticOutput("Lazy content creation triggered."); IoMemoryRef mem = IoMemoryRef::Create() iferr_return; mem.WriteBytes(0, payload) iferr_return; return mem; }) iferr_return; // Reading back would work the same here. return OK; } In your specific case, I do not know if BaseBitmap::Save will work directly on a Filename that wraps a ram disk url (via Filename file = cinema::MaxonConvert(myRamDiskUrl, MAXONCONVERTMODE::WRITE);), but you can try and it should work. When push comes to shove you can always write data directly as you showed yourself.

Thanks for your detailed answers @ferdinand ! At night, I suddenly had a flash of inspiration and found a simpler solution, use html tags: IDS_PLUGIN_HELP "description about......<br><b>Ctrl: </b> do something...<br><b>Alt: </b> do something... Cheers~ DunHou

The second ID is the sub ID. You can have a command (FOO = 100) which supports the sub-commands (BAR=1, and BAZ=2). So, you can then call CallCommand(100, 1) or CallCommand(100, 2) to invoke the two different things FOO can do. But as you said yourself, this is a rather unusual thing, and it is even more unusual that I manged to randomly click five things in a row that all have sub IDs What should also be said is that my little "extrude" log there does not really reproduce the modelling operations I carried out. It just runs the tools in the order they are invoked. And because when you run this right after your created the log, the extrude tool will for example still have set the same extrude depth from the last operation. But when you manually extrude something, and then run the script again, it will use these new current extrude tool settings, i.e., and not the ones from when your 'recorded' the log. The script log can be useful, but it is not the auto-script-recording feature users often whish it was.

Hello @mplec1, Welcome to the Maxon developers forum and its community, it is great to have you with us! Getting Started Before creating your next postings, we would recommend making yourself accustomed with our forum and support procedures. You did not do anything wrong, we point all new users to these rules. Forum Overview: Provides a broad overview of the fundamental structure and rules of this forum, such as the purpose of the different sub-forums or the fact that we will ban users who engage in hate speech or harassment. Support Procedures: Provides a more in detail overview of how we provide technical support for APIs here. This topic will tell you how to ask good questions and limits of our technical support. Forum Features: Provides an overview of the technical features of this forum, such as Markdown markup or file uploads. It is strongly recommended to read the first two topics carefully, especially the section Support Procedures: How to Ask Questions. About your First Question The bake tag and by extension cinema::BakeTexture (and its Python equivalent c4d.utils.BakeTexture) unfortunately do not support Redshift, even for the older Xpresso based materials. Redshift uses the tools found in Redshift/Tools/Texture Baking for baking. And while you can programmatically create a bake set and then programmatically click the 'Bake' button in it (which both would also work in a headless version of Cinema 4D, such as c4dpy), the following dialog which opens to set baking details and actually start the rendering is sealed, i.e., you cannot interact with it from the public API. And opening such dialog would also fail in a headless Cinema 4D instance. So, I am afraid there is currently no solution for your problem. You can technically export the whole scene to a format such FBX or USD, and use the builtin baking output (which would also work in a headless environment, as long as you do NOT pass SAVEDOCUMENTFLAGS_DIALOGSALLOWED to the save/export operation). But the output of that automated baking is often of poor quality compared to manually baking object(s) via bake sets. Cheers, Ferdinand

Hello @Dimitris_Derm. No, with islands I do not mean different objects. With islands are polygon (selection) islands meant. And that is just the term that is used for these things. The "Plane" object below has two polygon islands, the left and right rectangle shown in the viewport, each composed out of 4 * 5 polygons. They are islands because they are topological disjunct from each other - you cannot 'get' from one island to another without jumping over a gap. [image: 1777019034509-a88ac530-c105-430c-bd11-f922f5688881-image.png] The same can be applied to selections, as shown below. Now the left polygon island in the mesh has two polygonal selection islands. You cannot get from one selection island to another without jumping over a gap (they are topologically disjunct). [image: 1777019065284-bf41be6a-3993-4eb1-b603-438df9aa1bb3-image.png] This is a requirement because as my code example demonstrates, when bridging in island mode, you just specify a polygon and a point (and set the flags), and the tool then 'grows out' the to be bridged patch from the given polygon, based on the active polygon selection. And this growing will stop at topological boundaries. So, when you would bridge from the selection in the lower left corner of the left rectangle (to some unspecified target in another mesh), it would only grow into these four polygons. The selection in the top right corner would never be part of the operation. Cheers, Ferdinand

Hey, just as an FYI, I added this as an example to the 2026.2 rendering examples. You will be able to find it under \scripts\04_3d_concepts\rendering\render_document_tiles_2026_2.py. Since the script will use some 2026.2 features, it does not make much sense to post a preview here, as you will not be able to run it right now. The example also does the kernel border thing we discussed here. Cheers, Ferdinand

Ah ! They are called Proxy Tags and I can see the actual tag with the MSG_GETREALTAGDATA ! Thank you ️

Your approach is not necessarily worse, one could even argue that it is better. I personally would always avoid manually binding to an OS DLL via ctypes, but that is more a personal preference.

Hey @Dimitris_Derm. , no there is currently no dedicated plugin class for MoGraph effectors and fields in Python, only the scripting objects exist at the moment. Cheers, Ferdinand

Hey, well, "intended" would be the wrong word, but I am aware. The whole resource parsing situation is a bit of a mess at the moment, both in C++ and the Python API. Why is it like this? A description is defined by a res file (more or less its GUI declaration) and an h file which declares symbols for parameter IDs. The resource for the loop tool looks like this: CONTAINER ToolLoopSelection { NAME ToolLoopSelection; INCLUDE ToolBase; GROUP MDATA_MAINGROUP { BOOL MDATA_LOOP_SEL_STOP_AT_SELECTIONS { } BOOL MDATA_LOOP_SEL_STOP_AT_NON_QUADS { } BOOL MDATA_LOOP_SEL_STOP_AT_POLES { } } HIDE MDATA_COMMANDGROUP; } I.e., it indeed only defines three parameters. But the header file looks like this: #ifndef TOOLLOOPSELECTION_H__ #define TOOLLOOPSELECTION_H__ enum { MDATA_LOOP_SEL_STOP_AT_SELECTIONS = 1100, // BOOL MDATA_LOOP_SEL_STOP_AT_NON_QUADS = 1110, // BOOL MDATA_LOOP_SEL_STOP_AT_POLES = 1120, // BOOL MDATA_LOOP_FIRST_VERTEX = 1130, // LONG MDATA_LOOP_SECOND_VERTEX = 1131, // LONG MDATA_LOOP_POLYGON_INDEX = 1132, // LONG MDATA_LOOP_BOTH_SIDES = 1133, // BOOL MDATA_LOOP_SWAP_SIDES = 1134, // BOOL MDATA_LOOP_SELECTION_TYPE = 1140, // LONG (must be SELECTION_NEW, SELECTION_ADD or SELECTION_SUB) MDATA_LOOP_SEL_POLYGON_OBJECT = 1150, // LINK }; #endif // TOOLLOOPSELECTION_H__ I.e., it not only defines these three parameters, but also all the others. Because there are all these "hidden" parameters which are written into the data container of the tool, but never show up in the GUI. What collides here is (a) the a bit irregular (but not illegal) behavior of a resource to define more parameters in its header file than are used in the resource and (b) the questionable decision of our resource parser to ignore hidden parameters. Our resource parsing is automated, i.e., I cannot just go into a file and just add these parameters for a docs build. I could touch the resource parsers for Python and C++, but I do not have the time for that right now as they has been written in a very opaque manner. My advice is simply what the docs suggest: Search in the header files directly. Go to your Cinema folder, make sure that the resource.zip is either unpacked to the local folder (the folder existing does not mean necessarily that it has been fully unpacked) or an external folder of your choice. And then simply search in that folder with an editor of your choice. [image: 1775557617243-c122abc5-8cfc-40da-90be-534532600b4a-image-resized.png] At some point I will replace the resource and symbols parsing for Python and C++, because we made there quite a mess in the past with questionable parsers and manually curated lists. But for now this cannot be changed and using the files directly is the way to go when you want to look at descriptions. Cheers, Ferdinand PS: The C++ docs are NOT inherently better in that regard. The parser shares there the same flaws. The reason why you find some symbols there is because developers have duplicated them from the resources into the frameworks.

Hello @vaishhg, Thank you for reaching out to us and your comprehensive yet sparse problem description, this really makes things easier for us. The problem is a bit that you somewhat forgot a crucial part: The scene data :). You provided this: Create a Cinema 4D scene with a Main take (frame range 0-6) Create child takes (V, E, R, A) and override the render data's frame range on each (e.g., V: 0-9, E: 0-23, R: 0-13, A: 0-17) But I cannot recreate a scene with these instructions, unless you meant the second step in a rather metaphorical way (switching the render data itself via takes). I also talked with some of our support specialists, and the consensus seems to be that what you claim to be the case - a scene with take overwritten FROM/TO values - is not possible. What you can do, is create multiple render settings and then activate them via takes. But Takes are a tricky subject, and I would not rule out that you somehow found a way to create take overrides for parameters of a render data instance. If so, please share the scene. Find also a scene attached which demonstrates what I discussed above: A scene that switches the active render data based on takes. It will print this with your script: Take 'Main': FRAMEFROM=0, FRAMETO=90 Take '0_45': FRAMEFROM=0, FRAMETO=45 Take '0_90': FRAMEFROM=0, FRAMETO=90 take_renderdata.c4d So, to formally answer your main question: When my assumptions above are true, your scene is printing the same values because each take does have the same values, as these two parameters are not directly overridable. Cheers, Ferdinand

For now I am good. If I need further help, I will send the endpoint and function by email.

Hi @ferdinand, and thank you! Your proof-of-concept and pointing me toward mxutils.GetSceneGraphString() was exactly what I needed to solve this. By using the scene graph dumper on a native UI-generated Motion Source from a rigged character, I realized it's just a standard Ojoint hierarchy with normal CTrack objects. I used GetClone(c4d.COPYFLAGS_NO_HIERARCHY) to perfectly replicate the Ojoint skeleton and injected the time variables into the container, and it maps and plays back perfectly. Thanks again.

Here's my prototype, if you are interested in my goal Right now, everything works as expected, but only with these lines of code. profile = GetCloneSpline(profile_orig) path = GetCloneSpline(path_orig) I'm happy with the current result. So, I'd love to get some additional advice on correctness and optimization. I think your previous answer was comprehensive enough, so I'll try to integrate some of it. But I'd also be very grateful if you could take a look at my plugin and perhaps give me some more specific optimization tips. If that's not too much trouble, of course! @ferdinand @ThomasB Anyway, thanks for your replies and advices!

Thank you. MSG_DESCRIPTION_USERINTERACTION_END is just what was needed.

Hey @SteveHill3D, Storing a tool state in a tag is possible, but a bit unusual. But generally you are on the right track. Tags are also internally often used as hidden data containers. When you write a modelling tool, you probably already found out that points and polygons are are actually stored in hidden tags on objects. I am not sure though if using tags in tools is a good advice, as injecting all that data all the time into the scene graph is probably not the fastest. It depends a bit on what you are doing, but in general the solution is a bit unusual. Cheers, Ferdinand

@ferdinand Hey @ferdinand, Thank you again for the clarification — checking the Video Post list was indeed the right direction. Just to share what worked on my side (Standard/Physical renderer), I check the presence of the GI Video Post and use its BIT_VPDISABLED state to determine whether it’s enabled. Here’s the small helper function I’m using: def is_gi_active(doc): rd = doc.GetActiveRenderData() if not rd: return False # GI Video Post IDs (verified on C4D 2024) GI_IDS = [ 1021096, # Global Illumination Video Post (C4D 2024) 300001038, # VPglobalillumination (older versions) ] vp = rd.GetFirstVideoPost() while vp: if vp.GetType() in GI_IDS: return not vp.GetBit(c4d.BIT_VPDISABLED) vp = vp.GetNext() return False Thanks again for your help. Best regards, Mustapha

Hello @atg, Welcome to the Maxon developers forum and its community, it is great to have you with us! Getting Started Before creating your next postings, we would recommend making yourself accustomed with our forum and support procedures. You did not do anything wrong, we point all new users to these rules. Forum Overview: Provides a broad overview of the fundamental structure and rules of this forum, such as the purpose of the different sub-forums or the fact that we will ban users who engage in hate speech or harassment. Support Procedures: Provides a more in detail overview of how we provide technical support for APIs here. This topic will tell you how to ask good questions and limits of our technical support. Forum Features: Provides an overview of the technical features of this forum, such as Markdown markup or file uploads. It is strongly recommended to read the first two topics carefully, especially the section Support Procedures: How to Ask Questions. About your First Question I can see why would think that, but you mixed things there a bit up. The CMake SDK build system was introduced with 2025.2.0 and became the standard with 2026.0.0. Between 2025.2 and 2025.3 we supported both build systems as a grace period for developers to get accustomed. Chances are very good, that you can just copy the CMake setup, i.e., the cmake folder and files such as CMakeLists.txt, CMakePresets.json, and sdk_modules.txt to a 2025.0 folder and it will generate a correct build system for you. But the supported range is only 2025.2+. For older projects you would have to use the old project tool based build system. Since I know what you are trying to do, I would recommend trying copying before you get into the old build system of ours. 2025.2 Release Notes 2025.2 Build System Docs (which covered both the old Project Tool and CMake) Cheers, Ferdinand edit: You will only find the old project tool tooling in old extended SDKs which supported it, such as 2025.2 or 2025.0.1

hey @Gene, So, you were just trying to add information for future readers? That is of course very welcome and explains my inability to extract a question from your posting. I would not have been that generous with the word bug either. Since this has never been fixed, chances are high that we came to the same conclusion internally. In general, you should not expect Command calling patterns to make perfect sense in every case. There are multiple cases where Command is triggered more often than one would think at first glance. I would also not try to interpret some deeper sense into all that. Cheers, Ferdinand

Thank you so much Ferdinand! I have a way forward now for my plugin using the EffectorData ModifyPoints() method. Kind regards Fredrik