提供: wiki
移動先: 案内検索

Expansion of Real-Time Baking

It would be nice to have Cloth and Smoke sims Bake and display finished frames as they are calculated like the Fluid Sim does.

Blenderbench - A benchmark based on Blender

It would be great if there was a 3D-Benchmark similar to Cinebench based on Blender. Just like Cinebench, it should test Render/Raytracing-Performance (single- and multithreaded, both for Yafray and the builtin renderer) and realtime OpenGL drawing speed. Additionally I would strongly recommend to add other CPU-intensive tasks that users generally wait for, such as Particle Simulation/Baking, fur or cloth.

If there was a Blender-Benchmark (available for all operating systems that Blender runs on) it would be great free advertising for Blender, since all benchmarking sites and also magazines (such as the one I write for! ;-) would use it. Don't ask how much free advertising Maxon got thanks to Cinebench... It would not require that much development work (yes, I know what I'm talking about I have done my own scripted benchmark with programs coded by others! ;-), since all the code that the benchmark runs is already there, you'd only have to script it, add some example files, measure the time/fps and build a nice standalone executable that people can quickly download and run (such as f.ex. if you're on a product launch and have to do a quick test of some new CPU/GPU! ;-)

-- DeeKay

Sketching Interface for 3D Freeform Design


A feature that will make the modeling work flow quicker by using the traditional drawing techniques on a graphic tablet. This feature will allow the user to draw a 2d image that will be generated in a 3d object automatically. This allows the pencil to paper artist to create 3d models of their character or object by using traditional drawing techniques.

I have enclosed some research that has been undertaken by a university with a similar modeling system that produces low quality 3d models.


demostation video [2]

commercial products that have similar interactivity. [3] [4] [5]

-- nsterjov

Shockwave 3D Export

It would be great to have the ability to create 3D models in Blender, and to be able to export to Shockwave.

Some work has started on this here... Shockwave3d Shockwave 3D Export -- MalDuffin - 14 Jul 2005

Creation and replay of a changes file

A possiblity to save not only the current project, but all the steps to the current result into a .blend.replay file. This file can be played then with Blender like a movie. The User see all the steps (maye one in a sek.) starting by the first one and ending by the last step the other user have done. This would be a great feature for all these people who want to learn to work with Blender. And not to forget as a developer you can make with that very fast and easy tutorials which shows the others how to use a new feature.

-- Florian

Integrated Object Repository System

Imagine this: in blender, you click "add>Mesh>Repository>>animals>horses>horse1" and get a complete model of a horse that somebody half-way around the world made earlier that day. Right now, if we want to insert someone else's object, we have to search for objects online, download them, and append the objects from within a file. What if I could create a repository for objects on a web server? Have some standard way of doing it. Maybe we could have a file called "repository.xml" in the root directory of a web server, which contains a hierarchial database of files: filenames and the absolute URLs of models. Then in the "add" menu of blender, the user could add a repository by just clicking on an "add repository" menu item and entering the repository URL (or the absolute URL of the xml file). The xml file would be downloaded, checked against a schema, and a sub-menu of model names based on the xml file will be listed. You click on one of the model names and presto! you download an object and enter it in your sceene at the cursor.

-- socratesone

First run registration screen

It might be worthwhile to have a first run screen that gives an option to register for an announcement list and other possibly useful lists (ie voting for features), and possibly optional collection of demographic data (so we can get some idea of who is using blender, and where). The first run screen could also allow the setting of preferences such as default language.

-- Blenderwiki.TomMusgrove

Particle features

  • It would be extremly comfortable to have a particle-ratio setting for i.e. adjusting visually connected particlesystems (i.e. 3 items into a machine and 1 product out). You could set to have an average particle output of i.e. 4.5 particles per frame.

I could imaginge 3 values of implementation:

-The easiest way would be to implement a simple numberfield-output that shows you how much particles are generated per frame with the current START, END and AMOUNT settings. This is essential and helps to avoid using a calculator.

-In second instance it could be a numberfield-in-/output outranking the AMOUNT numberfield, which means modifying START/END will keep the value but variate the AMOUNT value.

-In third instance it could get an animatable value outranking and overriding also the START-/END-numberfields by IPO-animation. May be we will need a button to switch on IPO usage and START/END off. Imagine: You could reuse an emitter for example to

-let a gun shoot several independent times

-let an exhaust produce smoke regarding to the driver playing with the gas... -- M jack - Jannuary 2, 2006

  • I am missing the feature to have several particlesystems connected to one emitter like it was in older blenderversions. - (M jack - Jannuary 2nd 2007)
  • Damping fields/boxes: This field-type should affect a particle with a higher or lower damping value. 2 examples to explain: Bullets as particles shooting into water should get slower very fast. This could be achieved by creating a box and attaching that damping field. Otherwise smoke could be accelerated within a chimney.

Options: Default (spherical falloff), Box (inside the meshbox), Mesh (inside the mesh-geometry -- M jack - Jannuary 2, 2006

  • Particles baking and intelligent caching: When working on complex particle-scenes and using UNDO it can take quite a while to process and happens although nothing concerning the particles is affected. Maybe this could be solved by baking of particle movements or even better by a intelligent cache-system that avoids the recalculation.

-- M jack - Jannuary 2, 2006

  • I would like to see more work done with the particle system in blender.

I have seen so many things that can be done with particles, in 3ds max, and maya. I hope someone can start a project here at about it. Or tell me where I can start the project, Thanks. Contact Me

-- Blenderwiki.biomechonline - 03 Dec 2004

- There's still a missing feature with particles. The possibility to get back in time. In earlier versions, we could control that with the IPO "time" factor. Maybe we colud control that with negative values in the Tweak option?

- More particle integration with materials. When dropping a particle, the material propperties like transparency or color aren't been taken in the particle's life individually. That option is still a miss feature.

CVS support

I would like to see not only automatic numbering of files when saving, but also direct CVS support from within blender. This includes allowing cooperative work on the same file (for example, when you have a texture artist and a modeler working separately). I blieve the plain XML format makes it easier, as it will allow diffing.

-- RodrigoDamazio - 05 Feb 2005

Well, for binary files, CVS doesn't work so well. You can't really merge changes. XML blender files would help, but I still don't think it would work very practically. (What do you do when you need to manually merge, as is commonly needed in CVS) Now if/when Verse is ready, then you'll probably be able to do live collaboration. For now though, the best solution is using linked materials and object. You can load materials, meshes, etc. as linked from another .blend. To do that use "Append..." from the File menu or SHIFT+F1, then in the file loader at the bottom switch from "Append" to "Linked". Loaded data will be linked to the original file. So if the orginial copy is updated, with wil be up to date when you reload your file.

-- TedSchundler - 19 Apr 2005

A manual merge could perhaps be facilitated by a graphical dif tool (assuming XML here) would get three views of the object in question and graphical displays of the differences. From which the user could select/modify each to produce a final version to check in.

The three views/panes would be the original, the version being checked in, and the version currently checked in, the user then could choose the correct version, or manually modify one of the views to be correct. Then commit the change.

Different objects could have different views depending on the merge conflicts. Example: A mesh object view for vertex changes, would allow the user to see which vertex's have been moved (color highlighting on vertex differences). It could also have other layers that would show changes to texture etc. There could also be a ipo view to see changes in ipos.

In any case, XML would make this a lot easier to even begin to attempt.

-- George Bell - November 10, 2005

What you need is diff, merge and merge3 for .blend files, or a save format for which existing diff/merge tools would do. XML is not necessarily the right answer. Also consider Subversion (SVN). M.e 16:38, 29 April 2007 (CEST)

When you announce a new release candidate or a new release, could you please also include the CVS tag, so that the exact same code can be rebuilt (eg for another platform). M.e 07:06, 13 May 2007 (CEST)

Blender Mentor or Blender record/playback

Maya has a Maya mentor which can walk a person through the sequence of steps. What stops Blender from having one like that as well? Having a mentor module would significanly lower the threshold of starting with Blendor. It will most probably need to be implemented as a Python module. Therefore Python API will need to have a greater access to the Blender's events and objects (which ties in well with Python module redesign above).

Alternatively, if the full blown mentor is not an option, being able to record and then playback blender action would make it for a nice way to do tutorials. It could replace (or at least supplant) Video and still snapshot modes of teaching.

-- AlexandreRafalovitch - 21 Feb 2005

yes a record of the mouse moves and button and key clicks in the editor will be great. as zbrush has so you dont have to download whole movies just the strings of data. Can be used to make tutorial movies or to make macros for repetetive tasks.

-- Blenderwiki.ysvry - 14 Mar 2005

A working VRML97 importer

I now get file type unknown message all the time.

-- Blenderwiki.ysvry - 20 Apr 2005

CAD and CAM features

CAD; To be able to make 2d drafts of 3d objects. Maybe a script that can make a DXF export of 2d views of front, top and side. (With dashed hidden linens) Quick and dirty could be opengl to .ps (gl2ps) and then .ps to dxf (pstoedit). This file could then be opened by a regular 2d cad application to make measurment arrows etc.
Resources: (OpenGL referens on hidden lines) (OpenGL to .ps) (.ps to dxf)

CAM; To be able to produce g-code like this 3ds max example:

-- GnuttaG

Enhancing DupliVerts

I'd love to see a feature somewhat like the following:

DupliVerted Object appears at vertex locations in parent object, but its size, position, orientation are varied via an IPO curve having vertex colors as input.

Going into detail:


Parent Object has vertex colors, Duplicated Child C at vertex V is placed there according to Size/dLoc/Rot Ipos.

On the Ipo x axis there is one of the three RGB channes of Vertexcolors, in 0-1 range (so we can control position with Red and size with Green, or SizeX with Green and SizeY with Blue etc)

The value given by the Ipo curve for the colors of vertex V affect accordingly dupliCild C

This would already be great.


For each Channel there are three Ipos, not the regular one, graphically they should be drawn one in regular color, the other two fainter or thinner. The regular one acts as in BASIC mode described earlier, the other two are constrained to remain either always above (the first) or alway below (the second) the main Ipo.

These two additional curves acts as min max range. If they are used (turn them on/off with a toggle) The value of the Ipo is no deterministic but "stochastic", (via Blender or Perlin noise) within the given range.

With these additional feature it is possible to create much more interesting fur, grass, forests etc.

update [later on, 10 Aug] Timothy suggests to use weights instead that vertex colors and to have Ipos linked to vertex groups.

-- Blenderwiki.S68 - 10 Aug 2005

Colorbands on World

The current HorizontColor ZenithColor way of defining a standard environment is not good enough. A Colorband like option allowing for defining more than just two colors and how they do blend fron Horizon to Zenith would be very nice.

-- Blenderwiki.S68 - 12 Sep 2005

Support for Multiple Filetypes in Open Dialog

It would be great if the open file/append file dialogs would by default open and display any filetypes that there are import scripts for. Perhaps have it by default iterate through all the import scripts until one works. The dialog could also have a way to force the use of a certain script to open a file (in the case of 2 scripts for the same filetype) but by default the first one that works would be the one that is used. This would also enable the removal of the import menu item.

-- George Bell - 2 Nov 2005

New Spring Type Mesh and Curve

I would like to see a new "Spring" mesh and curve added with controls for the size, forward and reverse, distance between lines and number of vertices for the curve (or both). This would be very helpful when creating DupliVerted items or a spring object quickly that can have SoftBody properties applied to it. There is a tutorial here Spiral Staircase that can be used to create a spring curve or mesh but, it seems too time consuming. I have looked around for such a beast but, have failed in finding it. I may not have looked hard enough or missed it, if so, please erase this and put the link in my user discussion please :) --Chrysm 08:20, 2 November 2005 (CET)

Image Preview In File Browsers


This is quite trivial and I can live without it (there are bigger fish to fry for the developers) but, I thought I would throw it out anyway. I would like a preview of what the .blend file is in Windows Explorer. 3D Studio Max has this feature and it is very helpful when you have a folder full of .blend files. I am not a developer so, I don't know how much it would take to hook into the Windows API to make this preview. (Preview being the preview on the left hand side of Windows Folder View when you have a picture selected or a .3ds file if you have 3D Studio Max Installed). --Chrysm 08:29, 2 November 2005 (CET)


To have thumbnails in the most Linux file browsers like Nautilus(Gnome) or Konqueror(KDE) you need to implement the "Thumbnail Managing Standard":

--Tobias 21:59, 7 November 2006 (CET)

Create a metadata workflow

Inkscape[1] has some nice things, that Blender is missing: 1. defined metadata in .svg files (Inkscape uses RDF in the SVG Files)[2] -> we need it in .blend files

2. an metadata editor in Inkscape where you can enter your name, keywords etc. and select a license -> There we should use the Creative Commons[3]

3a. The Inkscape-Team has started the “Open Clip Art Library”[4] where they collect SVG-Images. The new version[5] uses ccHost[6] a very nice tool from Creative Commons[3] -> We “just” need to add support for .blend files to ccHost, and we have a very nice Blender model repository.

3b. Another nice thing with metadata in Blender files is, you can extend desktop search tools like Beagle[7] to find your files.

3c. And if we add the metadata into the rendered bitmaps too we could even find them.



Improve the Blender for CG movie project managerment

Database: the buildin batabase can deal with the bata and help to control the project progress.

--Sumomofan 21 March 2007 (CET)

Outliner & OOPS Schematic show all links

Large image textures are being used inside Blender files, and when no longer needed, should be de-linked and removed. The OOPS and Outliner are very useful for finding things inside of a blend file, especially with the new search function (yea!). If an image is used in a normal texture, it can be found by searching for its filename. However, it cannot be found if it is used by an image input node. One of the holes, I think, in the OOPS / Outliner is that they do not show nodes, especially image input nodes, linked to a scene. I think it would be useful if they did, for then you could do a search (the new function) in the Outliner and it would find the image file name, even if it was used by a Node as input or output. The OOPS can now show images, but they are just floating there if used by an Image Input node. Since you cannot X delete datablocks in the OOPS (new feature?), and the OOPS does not show you where it is linked, you cannot get rid of it. This floating image says "delete me I am not linked" but when it isn't deleted, it is confusing.

Before you could only see images as part of a material's texture - which shows up in the Outliner correctly. But what if it is a UV map? Then it shows up as an island in the OOP schamatic and not at all in the Outliner, just like the Image Input or File Output node. How about images for the image project modifier? Or composition? Or VSE Strip? Same problem. If these objects could be included, it would really help tracking these images down and freeing up precious memory space. Thanks!--Roger 05:57, 27 March 2007 (CEST)

Incremental Curve Normals

In the following link, I will create a detailed review for a new curve normals system.

Description of "Curve Normals": Curve normals are a necessary component in all major 3D applications, which make certain operations--such as bevob's and path animation--possible. When placing an object--or object data, such as is the case with bevobs--it is necessary to have complete orientation data, of which a curve tangent alone does not constitute. It is the curve normal that dictates which direction is "up" at any point along a curve, and the combination of the curve normal and tangent vectors together give us a complete orientation.

Wavez's Incremental Curve Normals Proposal

Wavez 11:04, 27 April 2007 (CEST)

katt's crazy-long feature request list

CommandLine Parameter Input File

This request works with the issue#1 on

Rationale: As Blender becomes powerful and flexible more amd more features are added as an options in the command line. Some operating system have a limit on the amount of text you place in one line or sometimes placing it in one string of line is not that tidy to look at. This feature will allow placing options/parameters in commandline to a file. The contents of the file will override/extend the other options placed in calling blender.

Changes in Blender:

    1) Create additional commandline option for a parameter input file.


    1) User creates the text file and types options/parameters separated by line-break(s)


    1) User creates a file named renderOptions.txt and types in
         -b myBlendFile.blend
    2) User may then type at commandline
       blender -paramInput="pathto/renderOptions.txt"
    2) User may also place extra options, in addition to those found in the parameter file
       blender -paramInput="pathto/renderOptions.txt" -s 10  -e 20

--nbjayme 8:03p, 18-Jun-2007

Per-camera background image

I think it would be VERY useful if each render camera could be loaded with its own background image.

When trying to reconstruct game ship models from multiple, 2D in-game views, for example, it's a pain to have to change the background image each time one switches cameras. Other times, there are x,y,z orthogonal reference diagrams of a ship, and same thing applies.

Additionally, it would be great to have a button in the camera panel to make it adopt the size and aspect ratio of the background image loaded in it, as its rendering params; so that one doesn't have to go F9 and manually change the rendering parameters to avoid stretching of the background image.

Finally, if background image offset could be adjusted by dragging the image, it would be very nice.


Collaborative 3D editing

This may be something like what Jahshaka may have implemented or may have said they would have like to have implemented. However, this feature could include peer-to-peer and centralized hubs that give users the ability to hook clients together and edit scenes in real time together. Whenever users want, they could save versions to their computer and what not.

Centralized repositories with smooth connectivity

It may be possible to persuade the Wikicommons ( community to enable the uploading of .blend files. If this happens capabilities could build into the Blender client to automatically upload (with proper licensing) and download blend files from the commons. This could also streamline collaborative editing.

Customizable keyboard bindings with a GUI

Maybe this has been suggested else where...

Contraint / Modifier Settings

  • Save Constraint / Modifier settings
  • Load Constraint / Modifier settings
  • Copy Constraint / Modifier settings
  • Paste Constraint / Modifier settings

Bartrobinson 15:05, 6 February 2009 (UTC)

Invert zoom button (preferences)

It looks like this 2.4x option disappeared in 2.5

Wireframe Mesh for Fluid-Domain

With blender 2.5 there comes the new smoke-feature. I think it is quite handy. I most like the function, that if you add a domain for smoke the domain-block goes wireframe. That is very helpful if you want to add things into the domain. Unfornatly the fluid-system does not have this feature. If you set the domain-block, it stays solid. That's laborious. I hope that would be changed.

P.S.: I'm new here and I hope that I added my request in the right way and in the right section. If not please tell me how to change right.

--Jansende 18:59, 25 November 2009 (UTC)

I notice 2.49b has "screenshot all". It would be useful in making tutorials for Blender if it had its own screen recorder. The fact that Blender will know when change happens(thus only taking a snap shot when needed) and where it happens (only record the change from previous frame)should mean it could capture frames with minimal CPU drain. I tried making a tutorial with Camstudio and it was too slow. This maybe a non-issue and I just have a slow old computer. :)

Multithreading Support for Soft Body & Cloth Collision Baking

--Gregory Kramdia (Algomorph) 8:31, 8 December 2009 (NY time)

If at all possible, we should start thinking about extending multithreading support for the baking features of Soft Body & Cloth collision (F7 - Physics - Collision Tabs under Soft Body & Cloth).

To site some data, with dual-core Athlon X2 QL-64 (2.1 GHz, 256 kb L1 / 1024 kb L2) baking of 100 frames with emitter animation of 1,000 hair strands takes roughly 2 minutes. Extrapolating, we get approximately one hour for 100 frames of 30,000 hair strands, provided a linear relationship between bake time and strand number, or 10 hours for a 40 second shot within a scene.

30,000 is a good ballpark figure to plan for relatively realistic human hair animation (real figure approx. 120,000 for human females).

If both cores (multiple threads) could be used for the simulation baking, the processing speed would increase by approx. 50% on a dual core machine (set 25% aside for system, just in case), approximately 250% (!) on a quad-core machine, and approximately 650% (!) on an oct-core machine.

Obviously, since the baking results depend on the end of each consecutive frame (i.e. frame #1 depends on frame #0), it is not possible to efficiently multithread based on frames. However, for the hair case - the soft body can be split up as long as self-collision isn't enabled, then the parts can be handled by separate threads (i.e. 1000 hair by each thread.) I'm sure there is a similar method to go about non-hair Soft Bodies and Cloth(i.e. split up by vertex groups).

Make saving save edited images

Blender ought to have an option to save *everything* when I save the .blend file.

For example, I just spent a good amount of time texturing color maps, spec maps, and so forth, saving often, but then it crashed. And it's not the first time I've lost texture work, at least 50+ other times. And lots of times I just save it and quit -- only to find all texturing I did wasn't saved.

I'm not sure if it's just a bad habit I have to not check every image I might've tweaked, but it does seem cumbersome.

Procedural builder

I'm not sure what the best name for this would be 'Procedural builder' or 'Procedural fractal builder' or something similar, however i see it as a very useful and quick method for modeling.

The idea comes from the method of creating fractals, not through mathematical formula, but procedurally. I have thought this through a little and i think it would implement better with its own window type.

The best way to explain the method is through example...


The image above shows two shapes a triangle on the left and a shape made with triangles on the right. The triangle on the left would be a base object which here as one socket node and one grow node, represented by the star and circle respectively. Each nodes has a normal. The base object has its own co-ordinate system and properties and grow nodes have property modifier values.

In the example above the growth node modifies the fractional size of the object and the procedure runs as follows:

  • start with base object 1 as model
  • Loop three times {add 'base object 1' to model with 'socket node 1' at 'grow node 1(age 0)'}

I have added 'age 0' at the grow node because the more the model grows the more grow nodes it will have.

This way of building models has many possibilities such as adding random factors in the property modifiers of the grow nodes. Limits to the growth of models may also be used. One of the properties of grow nodes would be the probability of an object being added there when an attempt to add an object is made.

There are many models that could be built in this way, trees, leaves, springs, buildings, wheels, etc. etc. and once the base objects and procedures have been made they can generate models as many times as is desired.

Hope it all makes sense. Please add to this and edit it to make it clearer and better ;-)

Jump 17:09, 8 July 2010 (UTC)

I have started writing an add on called Grow Modeller to do this. The terminology is not the same as above, but the main bits are more or less done. The wiki page is here.

--Jump 18:11, 12 April 2012 (CEST)

Multimedia enhancements

  • tracing bitmaps in the 3d view to convert them to paths
  • scan and capture from screen/[web]camera images, then import them
  • capture from screen/camera videos, then import them
  • export runtime generated executables shouldn't display the python console (do that in Blender only)
  • ASCII-Art rendering
  • text to speech for dialogue placeholding in games/movies
  • speech recognition for automatic subtitling
  • mimick sketchup's "followme" tool
  • measure distance between objects (ruler)
  • GUI for Audaspace
  • use Ming or ScriptSWF or something else to output native Flash files for both Blender and its game engine

Bye, IvanoA 12 March 2012

Use LIBELAS / LIBVISO2 to trace stereo photos / videos into meshes

They are both GPL libs, check these links for description:

Int_ua 2012-11-19