Blender Notes

The only problematic thing about Blender is that, like many of the best tools for professionals, it is horrendously difficult to learn. Fortunately there are some good resources to help.

Screencast-keys is a non-standard addon, found here, that will highlight what keys are pressed so that observers can follow along. I think it could be helpful for knowing what key you accidentally might have pressed. Note I haven’t tried this.

Why start modeling from scratch when you can use someone else’s model that they want you to use? I don’t yet know which of these sources suck and which ones are good, but I’m listing them here for completeness. I got the following list from here.

Since Blender exhaustively uses every possible key combination, you can’t have obscure ones that you never use lurking around in some miscellaneous interface features. My window manager is pretty good about never needing anything that doesn’t use the OS key, but I recently discovered that [C][S]-e seems to be bound to a new-fangled system feature of "ibus". (This is interesting because Gimp seems to know how to override it when using this combo for "Export".)

It looks like the way to cure this is to run ibus-setup and go to the "Emoji" tab. Click the "…" button; click the "Delete" button; click the "Ok" button.

It might be smart to get rid of the Unicode one too (which is [C][S]-u by default).

Maybe these could be rebound to the OS key if they’re needed some day. I’m not sure that’s even possible but it is what makes sense.

Here in the USA many things are measured in Freedom Units. Blender can play along. Go to the properties tabs and look for the "Scene Properties" tab which will be a little cone and sphere icon. The second item is "Units". Choose "Imperial" and "Inches" and things should be fine.

Sometimes I need to get CAD models imported into Blender and that isn’t always easy. One important trick is running the models through FreeCAD. Try exporting to STEP files from the CAD program and having FreeCAD import that and export STL files. Blender can then import those pretty well.

Full official details are here and actually helpful.

The old way to split areas into subdivided areas: drag the upper right corner down (horizontal split) or left (vertical split). To combine areas, make sure they are the same format (e.g. one big pane doesn’t go right into 2 horizontally split panes) and drag the upper right corner up or to the right.

In 2.8+ things seem much easier to arrange. Just go to the edge of the viewport you’re interested in fixing and look for the left-right or up-down arrow icons. Once those are visible, you can use RMB to bring up a menu that will allow you to "Vertical/Horizontal Split" or "Join Areas". Also look at the "View → Area" top button menu for another easy way to split. This allows you to customize easily without the drama of old Blender.

A huge tip for people with multiple monitors is that you can detach windows. In old Blender this is done with the same drag of the upper right corner, just hold down Shift first. Which way you drag doesn’t matter—the window is just cloned with its own window manager decoration.

In 2.8+ there is a similar technique with [S] and dragging the plus to the left or something like that. You can also use the "Window" button on the top row menu to create a "New Window" — you may need to do a lot of shuffling after that though. Another way is to use the "View" button on the top row menu and then "Areas" (which is what these Blender function regions are called) and then "Duplicate Area into New Window". This will usually duplicate the wrong window but at least it only gives you one area to reset the way you want it.

There are 1000s. Here I will try to enumerate the ones I’ve encountered. Also remember that you can go to Preferences and select the "Keymap" section and learn a lot about what the currently configured bindings (and possibilities) are. Probably best to not go too wild with changing those unless you really know what you’re doing. Note that the tooltips often have the keybinding for the operation and an extra interesting hint is that if you turn off tool tips in the Prferences → Interface you can get them at any time by holding Alt and hovering over the feature. F3 searching also shows the keybinding when available.

This video has a lot of tips about shortcuts for the node editor.

Since you have all those stupid useless keys sitting there that you never use, you might as well use them, right? Thought the Blender devs. Well I had a similar thought which was to get a less idiotic keyboard that didn’t have all that extraneous cruft. But Blender is really keen on being the one piece of software that justifies stupid keyboards. But there are tenuous workarounds.

The numpad numbers tend to change the view.

I’m still trying to sort out these features but am noting them here so they don’t get completely forgotten.

Note that the number keys are 10 number pad number keys; if the number keys want to change layers you may need to set "Emulate Numpad" in File → User Preferences → Input tab.

In Blender 2.8 some new useful interface features appear. Now you can go ahead and use the limited number keys (assuming you don’t have a stupid number pad) to select vertex, edge, or face editing. So how do you get quick access to viewports? Use the "`" backtick key to bring up a new style wheel menu from which you can use the mouse or numbers to select the view you want.

Often when modeling I would like something more helpful than everything showing up default gray. This can be done by going to the little down arrow box to the right of the "shading" menu which is located above the "n" menu on the upper right. Then you can choose "Random" and the objects will be colored in random different colors instead of default gray. That’s often enough for my simple needs.

A very common workflow technique is to freehand sculpt 3d assets on top of (or in front of, etc) a 2d reference image. These images don’t hang around for final rendering and are not part of the model per se. They are just available as helper guides to put things in roughly the right place so they look good with respect to reference material.

Sometimes you try to render something shiny (or anything actually) and it looks very computery because there’s nothing normal out in the wider world reflecting naturally off the object. This is especially problematic with isolated objects which look pretty unnatural just floating in space. To help cure this, you need to tell Blender’s "world" about what sort of ambient background is out there.

A decent place to start looking for environment textures is this site (aka this URL).

You could do worse than to start at Poly Haven and go to their HDRI section. Pick one that is somewhat like the mood of the scene you’re going for and download it.

Note that this file can be a .exr or .hdr file. Best to stick to OpenEXR which was created by ILM and seems wholesome. Blender created the OpenEXR Multilayer format which is an extension of that which is slightly less universally supported and probably not necessary if you don’t know you need it. All of these seem compressed but lossless.

To get Blender to start using this HDRI as a background: Go to the World Properties tab. Click Color and under Texture choose Environment Texture. This will then provide a file box where you can specify the path to your file.

If you want to adjust how your image is sitting on the background open up the Shader Node Editor. Choose the World pull down (next to the editor type icon). Then add two nodes: a texture coordinate node (from Input section) and a mapping node (from Vector section). Connect the Texture Coordinate’s Generated output into the vector input of the mapping Node. Connect the mapping node’s vector output to the hdr image node’s vector input. By playing with things in the mapping node, especially the Z rotation, you can get what you need. This process is explained in this video.

HDRI stands for High Dynamic Range Image and is often made of images taken with a camera at multiple exposure settings so it can provide more information about what the scene is doing in different conditions (no regions washed out or too dark). Here are some good examples.

This kind of image can be assembled with Gimp — here’s a decent beginner guide. And a more in depth guide.

It should be possible to make the dynamic range "high" with Gimp. To make stitched panoramas, you can look into Hugin. Compiling it is no fun but it seems to be available with apt install hugin.

In a pinch, there are some low res environment images that come with Blender that are used in the Material Preview mode. These can be found in your installation at. /${MYPATH}/blender-3.5.0/3.5/datafiles/studiolights/world

I find the distinction here can be tricky to get used to.

The origin is the tri-colored unit vectors with a white circle around it. The 3d-Cursor is a red/white circle crosshairs. Position the 3D-Cursor by LMB; note that it should stick to the things (e.g. faces) sensibly. Note that this is less obvious in wireframe mode. When in "Object Mode" in the tool shelf, there can sometimes be an "Edit" submenu; in that can be found a "Set Origin" pull down. This includes "Geometry to Origin" and "Origin to Geometry". Also "Origin to 3D Cursor" and "Origin to Center of Mass".

In 2.8+ to reposition the origin of your objects, select the object in object mode, click the "Object" button on the bottom menu, choose "Set Origin", and pick the thing you need such as "Origin to Geometry". Or in Object Context Menu (RMB in OM) look for "Set Origin" - also easy.

Most people align things by just eyeballing it; this is wrong. But Blender doesn’t always make it easy to do perfectly. And perfectly is the correct way!

The object alignment is pretty easy. There is a whole Align addon that ships with Blender. It can be useful and I think mostly obvious how to use it. But you can also just copy transform coordinates if you only have a couple.

Those little 2x5 grids in the menu bars are layer slots. To change the layer of an object, select it and press "m". This brings up a layer grid to select where you want it. To view multiple layers you can click on the boxes in the grid using shift for multiple layers.

The basic grid seems like the kind of thing that should be controlled in user preferences (like Inkscape) but it is not. Turn on the "n" menu and look for the "Display" section. There is a selection for which axis you want the "Grid Floor" to be (usually Z). Then adjust the scale and size.

Objects can be dragged around and placed in different hierarchical arrangements in the "Outliner". I’ve had this sometimes get stuck and it’s pretty strange, but reloading can cure it.

Having a good object hierarchy can make operations easier since it allows finer control of hiding or excluding from rendering.

To create a different object that is not part of an existing mesh, use [S]-d for "duplicate" and then hit the "p" key which will bring up the "separate" menu allowing the "Selected" object (or "All Loose Parts") to be made into their own objects.

The complimentary operation is to join objects. For example, if you input some reference lines on a boat hull and each is its own object, you can’t put faces on them. You must join them to the same object first. In object mode, select both of the objects and pres [C]-j and they will be part of the same mesh.

Parenting is a very important concept that allows objects to be organized. I used it less than I should have for a long time because the metaphor was confusing to me — children are not normally the first to appear in a family followed by the parent. But in Blender, selecting objects to make a parent child relationship, that is exactly how it is done.

For me a much better metaphor that conveniently starts with the letter p, makes the selection order and one to many aspects seem natural is to mentally replace the Blender verb "parent" with "piggyback". Or "piggyback on to". As an added bonus, "piggyback" implies a physical connection that is usually relevant in so-called parented objects. While it’s true that multiple passengers don’t usually ride the piggy at the same time, it’s totally plausible if the piggy is strong enough. (Certainly more plausible than a child having only one biological parent - ahem.) This dictionary definition of the verb piggyback pretty much exactly sums up what Blender parenting is: "to set up or cause to function in conjunction with something larger, more important, or already in existence or operation".

Each object can be thought of having one single slot for a parent. Just like if you move a file into a directory, you make that directory a parent of the file. There is an analogy to mv fileA fileB fileC parentdir since the last item in the arguments becomes the parent for all the other items. Of course the better analogy would involve a directory and subdirectories since all items can themselves be a parent. To get the terminology straight, imagine that files/subdirs are parented to the directories that contain them.

It is possible to assign a single parent object (the "carrier") to multiple child objects (the "riders"). As long as the parent object is selected last (called the "active object", glowing the brightest orange) the parent operation will parent each of the other selected objects to it.

One of the options in the [ctl+p] parenting menu is the "Object (Keep Transform)" option. This is subtle and often optional for most operations. The basic idea of it is that if an object is transformed because of its parent, when you try to parent it to a different object the question is if it will revert back to the basic object properties of the child first or if it will get a kind of "apply" to keep the previous parent’s influence and only then follow the new parent. An example would be if two characters were going to hand each other an object. The object would be parented to A’s hand and at the moment of hand off, you want it parented to B’s hand. What you probably don’t want is that when parenting to B, the object returns to where A’s hand was at the start of the scene where the object was originally modeled (and its transforms applied) and originally parented to A. That’s where it would go if it were no longer under A’s influence, but with Keep Transform, you can accept it’s new position/rotation/scale as the basis for this child object.

If you want to temporarily suspend parenting effects, with the parented object selected you can go to the "Active Tool and Workspace settings" tab (which looks like a screwdriver and a wrench) and choose "Options"; then one of the "Transform" options will be to "Affect Only… Parents". Check that and the parent can be adjusted with the children staying fixed.

For bones, the general strategy is that distal bones piggyback on the proximal bones by default assuming the bones were extruded distally. For example, the tib-fib (shin bone) piggybacks on or is parented to the femur (thigh bone). There the femur is the parent, selected last during the parenting operation. The idea of "bone parenting" as a specific parent mode is when you want some other object (e.g. a robot’s form/shape mesh or an animal’s skin) to be piggybacked on a specific bone as opposed to the whole armature.

A very common task I have when modeling is something like a racing bicycle’s handlebars or its brake cables or a curved railing or drain p-trap, etc. Any kind of free-form piping or tubing. Of course there is a way to do this in Blender and luckily it’s not too horrifically difficult.

The most common problem I have to solve with Blender is not sculpting a cube into a bunny but rather taking some engineered product from the real world and modeling it. Real world things usually aren’t so organically 3d and there are aspects of them that the designer simplified. For example, let’s say I’m trying to model an intersection for a vehicle simulation, I can pretty much assume that the designer of the road first thought about it in two dimensions before moving onto worry about camber and slopes. My workflow is very similar.

I often take a reference photo as described and what I want to do is just trace some geometry around important features to get started. But Blender, as far as I can tell, is terrible at this simple task. Well, it certainly does not emphasize it as important workflow, but, like all things in Blender, it is possible.

In object mode (tab to toggle) use the menu Add → Mesh → Plane and put a plane near where the geometry in question starts on the photo. Then "gZ" to move (in XY only) the plane square so that one of the points is where one of your model’s points should be. Go to edit mode (tab) and get rid of half the plane with "ab" (select all, bounding box) and grab two of the most incorrect points with a box. Then "x" to delete the vertices. Oh, another way is to have them all selected and do [A]-m and "Merge at Center" which will leave you with a single point at the center ready for action.

Now you have a line with one of the points correctly placed. Still in edit mode, select the incorrect point and "gZ" it to a correct location. With it still selected, do "eZ" to extend that point into a new line/point and put that on the next feature point of your model. Continue repeating this extending until you have the whole thing outlined.

Note that it’s just weird lines unless you can close them back up. For the final point that should be back where you started, you can put the point anywhere near. Then use "ab" and select both the beginning and ending point and do "[A]-m" to merge them. You may need to reset their location but that should be it.

Now with all the points of this object selected you can "ez" to extend in the Z axis and give that some thickness. You can use "f" to but a top and bottom on it. Now you have a complex shape that you can sculpt.

Often with the frenzy of extrude operations needed to get accurate geometry, you wind up with some silly organization. For example, I might extrude along a wall and then up a door edge, over the door’s width, back down the door height and then continue along the wall. This would be based on measurements of the room and door. But at the end of it all, I probably want an object representing the floor or walls and a separate object representing the door. How does one break out the door object?

My best guess for doing this which seems to work is to go into edit mode with the object selected that has the geometry you want to liberate. Select just that subset you want to make a new object out of. Then [S]-d seems to reasonably create in-mesh duplicates. This is also a good tip for doing things like repeating framing studs or some other repetitive geometry that might reasonably all be in the same mesh.

Once you have the geometry to create a new object from selected, press "p" which brings up the "Separate" menu. Since your geometry of interest is selected, choose "By Selection".

If you’re trying to hop some geometry from one object to the other, you can select the new object and the target object and use [C]-j to join the objects.

I never really use Blender text because it seems so weird. But I’m getting over that and it’s actually pretty well behaved. First of all a new font format has sneaked up on me: somefont.woff. This stands for Web Open Font Format. Sounds wholesome and Blender chose to use it, so it’s good enough for me. I’ve joked that you can spot a Blender work a mile away just by the distinctive default font. How do you get something different? It’s not too hard. I have created a blender directory in /usr/share/fonts where Linux (Debian anyway) stores its fonts. From there I link to my collection with my Blender stuff.

It may also be possible to have Blender not look for fonts in the system font location. This can be changed with Preferences → File Paths → Data → Fonts.

When Text objects are in Edit mode, the text can be changed. You can go to the Object menu and Convert them to meshes. Once you’ve done that, you can’t edit the text content, but you can do the normal 3d graphics stuff.

Loop cuts are very powerful and in dense meshes they can be quite intuitive. But in simple sparse geometry, they can get quite perplexing. The general idea is to go to edit mode and then edge select mode. From there use [C]-r to start the loop cut process; move the mouse around and little yellow guides should show you where your cut will happen.

You can use the page up and page down keys to adjust the number of cuts. (A mouse wheel, the normal clumsy way to do this, is not needed.)

When you’re happy with it, press enter. Now you should be in sliding mode where you can slide your loop cuts along the loops. If you did sliding by accident, you can use the RMB to put them back in the middle, without sliding. If you don’t want sliding at all, just press enter twice when accepting the cut.

Loop cuts provide a bizarre but serviceable trick for accurate technical modeling. A very common requirement in technical modeling is the need for a new feature a known distance away from some other feature. Imagine a face that represents a wall in a house - if you want to add a window, you need to make some cuts in that face where the window actually is. You can measure where the window is in the house, say, 40" from the corner of the room. To model this, do a loop cut. You get the orientation right, hit enter, and during the sliding phase, you slide this new window edge all the way into the corner of the room. Hit enter to accept this. That seems useless, but now you have the correct geometry to use "g" on the selected geometry, constrain it to the appropriate axis (with "x" or "y" probably) and type the exact value to put it in the right place ("40"). Now you have fresh geometry exactly where it should be based on numerical measurements.

Sometimes the loop cut extends farther than you’d like and you want to rejoin some cut faces. Use the face mode, select the faces to join and from the RMB menu use "Dissolve Faces".

This is a very strange workflow but doable once you get used to it.

Do you love triangles? I do. But Blender likes to hide the triangular truth of computer science from you. You can go into face mode and select non triangular faces and press [C]-t. This will triangulate them.

When I first was learning Blender I couldn’t quite understand exactly what the "modifiers" were doing or when or to what. For example, some modeling software keeps lists of boolean constructions instead of actual geometry. The resultant geometry is just calculated on the fly when needed. So are Blender modifiers like this? Not exactly.

Modifiers are a way to truly modify some mesh geometry. The confusion arises because they can be set up and sometimes previewed before the target geometry is actually modified. They can be stacked so that two modifiers are set to modify some geometry and if you like how it’s going then hit Apply and the actual modification will be done. Generally you’ll want to start at the top modifier and keep hitting "Apply" as the rest bubble up to the top to do multiple modifiers.

Modifiers show up under the object in the hierarchy while they’re being adjusted. They disappear when you "Apply" them.

There are a ton of these modifiers — so many that it’s hard to keep track of them. A very good resource is the Modifier Encyclopedia which contains descriptions and examples.

In computer science there are only triangles but Blender is strangely shy about letting that be known. If you like triangles you can divide meshes into triangles with the triangulate modifier. This modifier works well and has several fancy fine controls. Remember when dealing with modifiers, you may need to apply this in object mode, but the results won’t be interesting until you’re in edit mode.

Another way to do this without modifiers is Mesh → Faces → Triangluate Faces or [C]-T. Here’s a Python command line for the same thing.

Another way which is wasteful but aesthetically balanced is to "poke". This puts a vertex in the center of a polygon and triangulates radially from it. This is also in Mesh -> Faces -> Poke Faces or [A]-P.

Diffuse is when light scatters as on a rough surface. Specular is where the angle of incidence is equal to the angle of reflection.

I guess I wasn’t the only one marveling at the oversight of not being able to add a simple point. Or a line. Well, now you can! No more needing to collapse a cube — something I’m all too good at.

mesh_tinyCAD comes ready to go. Just turn it on in addon preferences.

Find mesh_tinyCAD features by right clicking into the context menu in edit mode; it will be at the top.

Here are things it claims to do, all of which are sensible.

Here’s an amazing resource that allows GIS data to be handled smoothly in Blender. If it’s working properly, it seems to be able to construct topo maps from nothing and populate them with simple building models. It looks so amazing that I am still interested even though it’s a struggle to install and get working.

Here’s a nice demo of it in action. Here’s how to install.

These modules are needed.

But wait! Remember that Blender has its own Python. So this apt install is nice, but doesn’t help.

Maybe restart Blender to be safe.

Go to Edit → Preferences → Addons and find BlenderGIS and enable it. Note the little triangle hinting that there are tons of settings you can play with to set up the behavior of the addon.

There is now a GIS panel in the 3d view properties menu. You should also have a GIS menu with "View Select Add Object" etc. Under that GIS menu, select "Web Geodata → Basemap"… I’m following these directions and… Well, dang.

Still, this looks very cool and worth keeping an eye on.

Here’s a very clever use for geometry nodes. When doing serious projects it’s easy to get very heavy assets that can bog down live editing. This very clever trick (from this video) shows how to have your geometry substituted out for a low poly place holder automatically.

This presentation expands upon the idea by suggesting a switch on a Delete Geometry node that selects random points to not show (using Utilities → Random Value node → Boolean). This allows you to develop with a very lean set of points, and then flip the switch for full detail final output.

Seems like this shouldn’t be too hard but there’s some node juggling that is required. Start with the Texture → Image Texture node. Put your transparent PNG in there; no unexpected settings. Next you need a Converter → ColorRamp node. Both of those will feed into a Color → MixRGB node — Color from the Image node goes to Mix’s Color2 and Color from the ColorRamp goes to Color1. If you get that backwards, you’ll get a reversed negative which may be a useful effect if that’s what you really need. The Color output of the Mix node goes into the Base Color of a normal Principled BSDF node. The BSDF node of that goes finally to the Surface of the Material Output node. Changing the Color bar of the ColorRamp changes what the background that’s not the image stuff looks like; note that each gradient position on the top color bar thingy has its own color.

That’s one way to do it. While working on a texture for tracer bullet fire I found another way. This guy’s succinct video seems to work and has a hilarious ending.

The process that worked for me:

Update for Blender 4

I found that this task has been made much easier in more recent Blender versions. In theory the old ways should still work, but you don’t even need to do that. Just the obvious alpha from your image to the Alpha of the Principled BSDF and that’s it. Like it should be. But! That actually doesn’t quite work out of the box — and neither do the old methods. You have to go to the Material tab and then Settings and then set Blend Mode to Alpha Hashed. I had a project where I wanted my decal to cast shadows and was disappointed that it didn’t work. But then I saw Shadow Mode - turn on Alpha Hashed for that too and it will all work! Super easy!

Here’s a 24s video that shows this.

This is easy unless you forget the weird process. Which I have done.

The trick is to focus on the "Active Material Index" (tool tips will say that when you’re over it). This is at the very top of the material tab accessible with the pink checkered ball icon. This is not the "Browse Material To Be Linked" box with its own white checkered ball icon. No, you need the top one. You must add all new textures you plan to use on an object here. Click the + to do this. It is in the far upper right of the tab just under the pin.

When you have the multiple materials ready there, you can select faces you want to apply a texture to, then highlight the correct texture in the index, and then hit the Assign button.

There are many helpful modes in Blender to control how stuff looks but they can be hard to find.

Imagine you just did a photogrammetry extravaganza and you have a model with a zillion polys. Everything looks good but you’d like to economize that mesh a bit. But you don’t want to lose the detailed texture that the process managed to pick up. The general procedure is described in this video.

The first thing to do is duplicate the heavy model. Rename one "Thing_ORIGINAL" or something like that and the other "Thing". You’re eventually going to get rid of Thing_ORIGINAL but hide it for now. With Thing selected look for the Object Data Properties tab and in that look for Remesh. I used Voxel and a "Voxel Size" of 2.5mm. and an Adaptivity of 250mm. I had Fix Poles, Preserve Volume, Preserve Paint Mask checked. Click the Voxel Remesh and come back when it’s done. Inspect that and if it’s not quite satisfactory, try some different settings. The Quad remesh is also interesting and tries harder to have lined up neat quad faces but it’s less sensible about adhering to certain kinds of complex geometry. There is also a Remesh Modifier that can probably do this too. I almost feel like that’s faster since you can check what the settings will do quicker.

Great, the mesh is the way you like it, but the texture is lost. Basically you need to bake a texture rendered from your original to the new UV map for the new mesh. The first thing to do is make sure "Thing" has no texture. It may be pointing to an old image for the original and that will be comically wrong. Bring up the UV Editing workflow tab and create a new bitmap in the image viewer. Change the size to 2048x2048 or whatever you think is necessary. Then with Thing selected and in Edit mode, use u to bring up the UV Mapping menu and choose Smart UV Project. Hopefully you see the new mesh’s UV map somewhat sensibly laid out on the new blank image that Thing’s material refers to. Once you get there, it’s time to project.

With both of the Thing objects visible (though it probably doesn’t matter), you need to select both of them such that the new Thing is the bright orange and the old thing is the dark orange. I found that means select Thing and then Thing_ORIGINAL even though other selection processes in Blender seem to do it differently. Next go to the Render Properties tab and make sure you choose Cycles for the Render Engine. This is temporarily required to do this process and will present a Bake subsection. The Bake Type is Diffuse since you’re really trying to capture the actual light bouncing off this thing which includes its color. Under Contributions make sure only Color is checked. Now make sure Selected to Active is checked and since you were careful with the selection order, this should be ready. I left Cage unchecked and made an Extrusion length of 10mm. The Max Ray Distance was ok at 0. Now click the Bake button and take a break while that completes. If all goes well, you should have new pixels on your texture image that neatly paint the new mesh.

Delete the original and the original meshes if you like. Probably smart to save or embed the new texture bitmap.

Note that with Node Wrangler enabled, you can use [Ctl+Shift+t] to automatically setup maps. The addon will look for common names and this behavior can be refined looking at Preferences → Addons → Node: Node Wrangler → Edit tags for auto texture detection in Principled BSDF setup.

I have struggled with ordinary glass panes in windows before. This tip is quite good and does seem to do the job very well. Here is the summary of what that looks like.

Fancy people setup every part of their virtual universe but I’m often just trying to show off some single artifact I’ve been working on. The default dark gray background that shows up on renders is often at odds with dark items. To change the way the rendered background looks when it is empty, go to "Properties → World Properties → Surface → Surface" and set that to "Background". Now you can change the color and "strength" of the background. Note this does not do anything to make wireframe or solid view look any different.

Apparently to change the background of the 3d Viewport for wireframe and solid mode, you can go to Edit → Preferences → Themes → 3D Viewport

The color of the grid lines can be changed here. But keep scrolling down… farther… keep going… almost there… until you get to an entire section questionably called Gradient Colors. In there you will find Theme Background Color which has options. By default the eponymous "gradient" is turned off, but turning it on can be a festive thing to do. In this context, I get the feeling that "low" means the bottom of the screen, not a color property; high is the top.

One of the most frustrating things about trying to get high-quality realistic rendering is noise artifacts. This manifests as a sparse field of bright pixels where that is not accurate. Here are some places to start when trying to figure out how to minimize this problem.

I think this process has been streamlined with some necessary transformations automatically sorted out. Now just doing the export step works.

When imported into Unity, it opens and looks right in the scene. Here are how the axes are adjusted.

The location of the imported asset seems correct when placed exactly by setting the Transform Position attributes in Unity.

It may be wise to set the units in Blender to match what you’re trying to achieve in Unity.

Hmm. Yes, there still are problems. The two clues that things didn’t work quite right are that in the icon of the model asset, it is pointing up. When it’s placed in the scene it seems correct, but that icon is not consistent. And the position is just a little bit off by a weird amount.

This video is quite specific about what the problems are and what you can do to minimize Blender/Unity conversion problems.

I did the easy first method of clicking the "Experimental transforms" box during the FBX export. That seemed to cure problems.

Use [C]-b to select a box to render. This puts the powerful and resource hungry rendering engine to work only in this region of interest.

To clear this rendering ROI box [C]-[A]-b. This is important if the ROI box was done in the camera because that is how the final animation will be done.

In the Sampling section of the Render properties, the Samples box contains two fields, "Render:" and "Preview:". The Preview one is for what is rendered in the preview box set with [C]-b. This can be very helpful to determine the level necessary for a more complete render.

If you really have your act together, you can be very explicit of course, but I often need to move around looking at the scene as the camera will see it. To do this easily, use [backtick,1] to get to the camera view. Then go to View → Navigation → Walk Navigation. Now you can use WASD to pan around (also E for up and Q for down) and the mouse to aim the camera. Once you think it’s a decent shot, press F12 to render it.

Another one that sometimes comes up is what I’ve seen other programs call "twist". This is where the camera is rotated about its long axis. To change this — perhaps while looking through the camera view, or not — select the camera and [r] for rotate, and then [z,z] to align the rotation about the local Z axis which is the direction the camera points. If you don’t really want the camera to be changed but just need this effect in the viewport view, you can use View → Navigation → Roll Left/Right.

Although it is a somewhat exotic application, one thing Blender can do really well is show you roughly what some real cameras would have in frame by matching those cameras in the model and rendering. The render is exactly the same (or trying to be) as what a real digital camera of the same resolution is seeing. The trick then is how to match your Blender virtual cameras to the exact specifications of your real ones. This Q&A addresses the issue. The basic strategy is to go to the (SLR-looking camera) Render icon and go to Dimensions and set the pixels there. That sets the aspect ratio. Then go to the (Movie camera) Data icon and look for "Field of View". This field of view angle seems to apply to the axis with the largest number of pixels and the other axis is scaled appropriately. This means a 1000x2000 render with a 90 degree FoV will have that 90 degrees in the vertical orientation.

Freestyle rendering mode is a way to achieve interesting non-photorealistic effects in final renders.

One problem I had was where I had a very simple blocky shape with simple 90 degree angles and on a couple of the straight lines, there was a strange fluctuation of width. This made no sense, but I tracked it down to a particular setting. In the "Freestyle Line Style" section there is a property called "Chaining" which in theory can link geometry together. Turning this off cured the problem. I never figured out what it was trying to chain or how since the defective features were single edges with no complexity whatsoever.

Also note that you can change the level of detail of the rendering to speed up test renders. To find this go to the "Properties" bar; choose the "Render" button which looks like a camera; look for the "Sampling" section which is between "Freestyle" and "Geometry"; in this section look for the "Samples:" fields; specifically look for the "Render:" field. A value of perhaps 1024 is pretty decent and takes a while and a value of 32 or 64 can make things go quickly for previews. I don’t know exactly why sometimes this Sampling section is absent but this seems to be the important thing to adjust for grainy quality problems.

If you have access to a fancy GPU, rendering might be faster. But it might not. This is not a slam dunk in the Blender world like in other applications. If using a GPU consider adjusting the values in "Render→Performance→Tiles→Center→(X,Y)" to something larger than the normal "16" that seems pretty decent for CPU rendering. Note that if you do have a fancy GPU that is competitive or superior, you can run two instances of Blender, one rendering with the CPU and another with the GPU. I can’t easily think of a better way to punish a computer’s entire capability!

Sometimes you don’t need a video and you just want a still render of your scene. Go to the UV/Image Editor window type at the bottom of the rendered image. Got to Image and then Save As Image. Also F3 works.

Note that default Blender 2.8+ seems to want to open a completely new window when asked to render. This can be fixed in the preferences. Look for this.

Interface → Temporary Windows → Render In → Image Editor

It often makes a lot of sense not to render to a full video file. If you have hundreds of frames and Blender crashes at frame 300, if you’re rendering directly to a video file, you’ll need to solve the problem and start from the beginning. Rendering to images, therefore, usually makes a lot more sense. This also allows you to distribute the load of rendering between many different computers.

The important trick is how to assemble the still frames into the final video file that you want. You can use another tool like this.

This will take everything named something like 0003.png and create the video file. Also explore ffmpeg if that’s currently in vogue on your system (see my video notes for annoying details).

Here’s another trick to get an animated gif.

Or here are some fancy ImageMagick options.

Note especially -delay 2 which seems to be the minimum value. Anything less will make, strangely, add much more delay ref. Also note that if there are errors, the optimizer might need more room. See this.

You can also use the Blender video editor and "Add" "Image", select all the images and eventually export it when you’re happy. See video editing below.

I think it’s a weird omission that Blender doesn’t have a native backend (that I know about) for animated gifs. But that’s fine. In this rare instance of Blender being frugal with features, there are plenty of ways to do the job.

First decide if you want a transparent background. This isn’t about gifs per se, but it tends to be a popular idea with gifs. Go to Render Properties (the camera tab) and look for the Film section. There should be a simple toggle for Transparent. Now back to Output Properties (the printer tab) make sure Color has RGBA if you’re going for transparent. Or not if not.

Set your Resolution X and Y as normal and pick File Format as PNG. Make sure you have your directory path (top of Output section of Output Properties) set to something sensible; I like /tmp/R/ for "render". Now Render → Render Animation from the top bar menu ([C]-F12, and remember you must use a real Ctrl key, not a remapped one).

Now you have a collection of PNG frames in /tmp/R/. How do you convert those to an animated gif? The most scriptable technique is to use gifsicle. Unfortunately, this program only assembles gifs from gifs. So you’ll need a process like this.

Note that gifsicle is a weird program with weird options. The delay seems to be in 1/10 seconds (so 50ms shown here). The disposal method may not be needed if you do not need transparent background. The output file is only found on standard output, so redirection is necessary.

One obvious example for normal people is letting AWS do the heavy lifting for you. This way you can optimize the type of engine you use (GPU or CPU) for your project and get all of that heat out of your house. Step one is to set everything up exactly like it needs to be in your Blender project. Make sure that if you were using a GUI session you could just open the project and hit [F12] (Render Image) and everything would work perfectly. If that’s the case, log into your remote Linux system.

Rsync your project to the remote system. Run the render with something like this.

This will produce a file chessScene0001.png with no GUI fuss.

To do a complete animation use -a for the whole thing or pick specific frames with -f 1..20 or something like that.

The // syntax indicates relative paths from the blend file. You can also include explicit padded frame numbers with #, one per digit. For example test-##.png becomes test-000001.png. Sometimes you might want to just dump them to somewhere in temp. I don’t know if this is overriding the Render Properties but it might be. Seems you don’t need a -o if you’re happy with what’s specified in the .blend file.

Of course one problem with this strategy is that you can send up some geometry that is pretty lean and render it into lavish high frame rate bit maps; this can blow up the size considerably. So just think about that with respect to any transfer/storage fees that might exist. If you just happen to have access to an awesome private GPU machine, go for it!

Here’s a little script I wrote to help get a quick preview of full renders without doing all the frames. Lots of good information contained in here and will be a useful template for remote operation and custom project build scripts.

I called that script Command Interface Lights Out Rendering or the easy to remember ciloren (pronounced "Kylo Ren").

Note that running blender remotely from the command line may cause trouble due to a lack of an X windows system (answering system calls?). This thread talks a lot about the issue of Unable to open a display. I was finally able to get things going by prefacing my blender command with a DISPLAY setting like this.

OSL is used by the Cycles render engine. Examples of it can be found in /usr/share/blender/scripts/templates_osl.

I’m mostly interested in simple machine simulations. This should be easier than rigging a human ninja, but strangely it is not.

Here is an excellent simple tutorial rigging a simple robot arm. Here is a landing gear featuring pistons and hard pivot joints.

Here’s a good video showing the normal rigging of a normal biped figure.

This whole series, Humane Rigging is very well done. I watched it all just to marvel at the amazing complexity that high quality rigging seems to require. This video series was nice because it was not "easy" or for beginners — it told the mathematical truth to many of Blender’s weird decisions. Overall very helpful to understand things even if you’ll never rig and animate a battle between a robot squid and a Medusa.

Danpro has a nice vehicle rig series. I think if you watch these 10 times, the ideas will start to sink in a bit.

Here are some miscellaneous tips I’ve collected on how to rig mechanical systems. This could be anything from a robot arm to a canal lock. Although this is not the ultimate end goal of Blender’s rigging system — which tends to focus more on people and animals — it is actually quite tractable. The only difficulty is that this is not the primary work discussed in most tutorials.

When creating the bones, place them in the mechanical part that moves. For somethings this is simple; for example a backhoe might have a series of articulations in a line and the linked bones make sense. But in a robot arm, there may be a section where the end has a motor on it and the next section is at the end of that motor’s shaft shifted over some distance. Just create a new bone in edit mode or duplicate one and avoid extruding in this case. Then just parent the end bone to the one closer to the base using "Keep offset".

Many mechanical systems have very limited motion. If a robot arm joint only moves the elbow in one axis you can restrict the motion from pose mode by clicking the little lock icons next to the axes that do not rotate. It may be smart to lock off the scale too — no reason for that to change usually.

Y is the local axis that does a twist meaning the bone doesn’t move but the local axes of X and Z roll around.

Naming things well is always helpful. Here is a good convention I have seen:

Another good idea is to name the whole Armature something sensible (instead of the default Armature) — something like RIG-landing_gear.

For control only bones you can turn off the deform property in Bone Properties.

It is smart to have everything zeroed out before starting. Have your mesh sitting at 0,0,0 and all transforms cleared (e.g. [Alt+r] to clear rotation, [Alt+s] to clear scale, or use [Ctl+a], etc. ). Note that while you’re posing things in Pose Mode you can reset rotations with [Alt+r] back to the rest pose.

I’m often fine with using [Shift+z] to switch to the wire mesh display, but many people like to select the armature and then go to Object Data Properties (the little running stick figure) and then go to Viewport Display and click In Front (goes well with [Alt+z] partial transparency). Just under that is the setting for Axes which can also be helpful when troubleshooting, especially alignment/roll/twist problems.

To hook up the mesh to the armature you do not use the same kind of process you’d use with an organic model. With animals and people, you tend to parent the skin mesh(es) to the skeleton armature with [Ctl+p] and then With Automatic Weights. But with a rigid mechanical assembly, this is wrong. You want to use With Empty Groups. What this does is make groups for all the mesh components (the robot’s upper arm mesh object, lower arm mesh object, etc.) but leaves them unassigned. Automatic weights tries to smoothly distribute, for example, the upper and lower arm around the elbow. But for a mechanical model, you’ll want to assign an all or nothing approach yourself. Now after parenting with Empty Groups if you go to the Object Data Properties tab for each of the parented robot component meshes, you’ll see under Vertex Groups a list of all the bones. If you go to Edit Mode you’ll have the chance to select vertices — all is popular with mechanical rigs — and click the Assign button that appears in the Vertex Groups area. If some part of your mesh is moving with the wrong bone, you can try to Remove those vertices from that bone in the same way.

Note that when you assign 100% weights of a vertex group for an object to some bone, you are creating an Armature Modifier. Check for it in the modifier stack. This is ignorable if you have no other modifiers. If you do, however, it’s good to have this modifier calculate before any subsequent expensive ones are calculated. In other words, this will help put the simple mesh in the right place and then make it complex rather than having to do the latter and reposition a complex mesh. To move order of the modifiers, use the little down v button and select Move to First.

When making an Inverse Kinematics system, you are basically making a second armature that uses an IK constraint and there is a slider value between that armature’s influence and the normal forward kinematics one. Since the end will control things closer to the base it is not quite right that it is parented to it. For example, if your FK system allows positioning an upper arm, then a forearm, then a hand, that’s how it is parented. But if you want to position the arm by specifying where the hand is then the hand in the IK rig can’t be parented to that chain. You need to start by clearing the parent relationship of the IK armature’s "hand" (or whatever is controlling things in an IK way). To do that, with the hand bone selected, use [Alt+p] and Clear Parent. That hand bone in the IK chain actually must be re-parented to the base so that it behaves sensibly if the whole mechanism is moved; do the parenting in the normal way keeping the offset.

This series is really excellent for carefully developing a custom rig for a robot arm complete with FK and IK controls. The one I linked to is part 5 of an excellent 6 part series and shows the main tricky bits with the IK drivers.

Often I create something like a robot and rig it and I use the rig in the normal way. But sometimes I just need my robot unrigged in a cool pose, perhaps as a static background prop. How do I get my robot all lined up neatly to take on the pose set with the rig and then forget about the rig?

The answer is to apply the Armature modifier for each mesh. Now that mesh is as if you put it there with normal transforms. Delete the rig if you want.

A 4 bar linkage is a good test of Blender’s ability to do sensible things with complex assemblies. Here’s a description of a simple process for setting this up.

Rigify is a 1st class addon present in stock Blender that seems to pre-emptively do a lot of typical rigging jobs so you can skip a lot of typical work. I found it interesting that you can add ([S]-a,a) a rig from bird, cat, horse, shark, wolf. I don’t know why those animals, but they could be helpful. There’s also a fancy human, basic human and basic quadruped to get you started.

CGDive has a superb video series that covers rigging thoroughly. Without a resource like this, it really is hard to make any sense out of how Rigify really works and how to get the most out of it. I found the videos on custom Rigify concepts to be exceptionally valuable — those are part 1 and part 2. (These links are updated for 4.0 Blender but he has an older 2.8 series that still has some value too.)

When you add a meta rig but it is hidden inside the model, you can select the "metarig" object (that’s Rigify’s automatic name), go to the Object Data Properties (the little running stick figure icon), then open the Viewport Display section, and make sure "In Front" is selected. Since I like wire frame mode more than normal users this bothers me less than it seems to bother every single Rigify tutorial creator; yet even I can’t understand why this is not the default.

It seems like it is important to make sure that you clear any lingering transforms on the model you’re trying to rig and also the metarig which you may have just tweaked to match your model. It seems that it may be sufficient but not necessary to make sure both are cleared (with [C]-a) but it does seem at least necessary to make sure they are not different.

The donkey work here is positioning the auto-generated metarig components (in Edit mode) inside your model’s exterior mesh. Here are some handy tips.

If you use the meta rigs in a predictable way and keep them largely intact, instead of parenting the mesh to them as you do by hand, you can go to the armature properties and scroll down and find a "Generate Rig" button (must be in Object mode). This will let the Rigify addon really do a fancy job of rigging your model. Once the fancy rig is generated, you can parent the mesh to this fancy rig (use "with automatic weights"). In theory you should be able to (save a copy first and) delete the meta rig since its job is done; I found this helpful since otherwise it just sits there in the wrong position. Actually it may also be present in the layer #31 slot armature object properties → Skeleton → Layers grid. So safe to delete unless you want to retry completely (so Save Copy!).

To do a simple rigging of a simple single object with a posable bone I’ll use the example of a rowing oar.

I have had problems being able to reset my rig back to the default pose. Sometimes when I want to start posing a rigged model I get "Cannot change pose when Rest Position is enabled." On the object data properties for the armature there is a button that says "Pose Position" and another that says "Rest Position". Clicking the first can just scramble the whole thing. So how to start with the rest pose as the first pose? In Pose Mode hit [Ctl+a] to bring up the apply menu which will look pretty different. Select "Apply Pose as Rest Pose". This has worked for me so that I could switch to "Pose Position" an have it not scramble.

Sometimes however this is not enough. Maybe the rest pose is fine and when you go to Pose Position button in the armature properties all hell breaks loose. How can you get the pose position to just chill out and be like the rest pose? The trick I’ve found is to first go to pose mode and use [a] to select all bones. Then use [alt+g], [alt+r], and [alt+s] (all three!), to basically do what the Pose → Clear Transform → All menu item does. With the pose transforms cleared, you’re left with the rest position when posed. Start again and enjoy!

Still having problems with this? Make sure you check all your bone constraints. I had some set on a bone that locked the rig into a weird pose position even when I tried to clear all the transforms to reset it to the rest position. At least turn off visibility on those modifiers to check if that’s causing problems.

Sometime everything seems pretty good with the rigging process but when you animate your model, some part isn’t quite right. Maybe the crease in a joint looks weirdly folded or you modeled a hand too close to the body and moving the arms picks up the hips' mesh. To fix this, you need to weight paint the significance of each bone’s contribution to mesh deformation.

Annotate documentation is sensible.

Annotations are very useful but a bit weird. They’re essentially a way to put stuff on your stuff. This makes one wonder why not simply use some more stuff? Why introduce a different kind of stuff for putting on stuff? The annotation system is a way to let you just free-form doodle pretty much anything anywhere in Blender at any time. This is fantastically useful for communicating with other people (and your future self!) subtle details about the Blender project itself.

The important forgettable thing to know is that holding down the [d] key makes LMB into a free-form doodle at the cursor.

Here are some more very useful keybindings.

But to get a little fancier with it you probably want to open the [t] menu and click on the little pencil with blue line icon. With this tool permanently activated you can then see its properties in the [n] menu. Those properties allow changing colors and organizing doodles into different annotation "objects" (they’re not really normal Blender objects because they don’t show up in the outliner). To get straight lines, polygons, or the eraser mode, hold the LMB over the annotate tool icon in the [t] menu. Also [Shift+space] has an Annotate option that you can hold down for all the modes.

The Placement mode is very useful. Surface Placement will allow you to scribble on the face of some geometry, flat or complicated. View Placement will stick the doodle to the 2d screen regardless of how you orient your model in 3d. 3d Cursor Palcement just lines up your doodle to the current view and puts the depth at the 3d cursor; what "depth" exactly means is a bit complicated since the annotation is drawn after (on top of) the model by default; this can be changed clicking on the In Front icon in the [n] menu after clicking the annotation pull down (next to color).

I really find the default dull civil war blue to be a terrible color for highlighting things. This is easy to fix in Preferences → Editing → Annotations → Default Color. Problem solved.

Note that annotations use some of the same machinery as grease pencil features. Annotations can actually have onion skin properties like grease pencil — they probably use the same system.

To get started, you must add a grease pencil object in a normal 3d editing situation. Once you put that in your scene, you can change the mode to "Draw" and start playing around.

Note that you will not see the proper line thickness or color if you are in wireframe view mode — try solid.

Here are some interesting key bindings.

You can mark edges of a model and then in Object mode, go to the Object GUI top button and look for "Convert To" and "Grease Pencil". Go to the lower left menu ([F9] if you missed it) and look for "Only seam edges". You can adjust some other things too.

Enabling the addon Add Curve Extra Objects will give you a lot of things to make quick grease pencil items from. Stuff like arrows and stars.

With several grease pencil objects it can be confusing which one you’re working on. Check the outliner for a dot or an icon to the left of the grease pencil object name. That is where you can change which object is being worked on.

This information was so useful, I’m going to reproduce it here.

Collections can contain grease pencil objects (fat squiggly snake icon). The grease pencil objects can contain a grease pencil data block (similar icon with endpoint vertex squares). The grease pencil data block generally contains layers, the basic two are by default called "Lines" and "Fills". Layers contain frames which are selected by the playhead placement. Frames are composed of the strokes and fills themselves which are composed of points (perhaps vertices). There is also a concept of channels which is a bit more mysterious to me — the manual provides this hint, "With Auto keyframe activated, every time you create a stroke in Grease Pencil object Draw Mode a new keyframe is added at the current frame on the active channel." This manual page also has hints about how it all works.

Here’s one workflow to animate a typical thing. The manual talks about redrawing new frames each time, but I find it helpful to just morph the previous position. This doesn’t have to be the previous frame. With the dopesheet grease pencil mode visible, make sure your "layer" is visible. The layer can be cross referenced over in the object data properties under Layers. If you use [i] to insert a new keyframe all of your previous artwork will be cleared; perhaps ok for starting at the first frame. Draw your lines in draw mode. Then select the keyframe on the layers channel (you must have the mouse focus on the dope sheet) and [Sh+d,x,5,Enter] to create a duplicate keyframe of this art 5 frames over. Then scrub over 5 frames (or whatever you need) to the next important position. Edit or sculpt your art into the new position. This will be mutating the geometry on this new copy of the keyframe. An advantage of this method is that you preserve perfect correspondence of all the parts. E.g. the leg in the last keyframe is the leg, drawn the same direction, in this one. This means that later, you can go back and interpolate and expect things to go pretty well.

The grease pencil data block tab — near the Materials thingy when a GP Object is active — allows you to add and remove layers if the ordinary "Lines" and "Fills" is not enough. I find that adding a GP Object with "Add → Grease Pencil → Stroke" starts you off with a quintessential stroke that is inevitably not what you really need; it seems quite ok to use "Blank".

If you’re not seeing the grease pencil strokes or they are faded out while you’re trying to create them there are many possible explanations.

Note that to see what’s going on along the animation timeline, you need to have a dope sheet pane visible and you need to set its submode to "Grease Pencil"; only then will you see the keyframes that govern what grease pencil art is visible at any given time.

An important tip is to make sure you isolate the subject’s animation from its motion. This sounds like they are the same or similar or related but it’s best to decouple them. For example, if you have a character walking, it’s tempting to assume that in frame 2 the character will have moved forward because of the walking. But you must resist that temptation and animate as if the character was in a space suit and miming the action during a space walk with a black featureless background. Once the characters internal motion relative to itself is animated, only then should you provide the actual motion in context. This is relatively easy because you can just keyframe the object in object mode and it will all look sensible.

The animation system in Blender has an easy mode that is deceptively easy. You can set auto keyframes and move a thing around and hit play and it’s good. But behind the scenes all kinds of crazy stuff is going on and any animation that gets even slightly complex will start to be very confusing very quickly.

For example, in the timeline, new keyframes often do not appear when you might expect them. The crazy thing is that you need to use MMB to pan the display down because the keyframes are there, just panned up above the title bar. Crazy! That is a broken system! But be aware.

This video is an excellent introduction to the true nature of Blender’s animation system beyond dumbed down clutching at keyframes.

There are many different editor panels which are used in controlling animation.

Try this video too.

In Blender there is an important but somewhat advanced concept (not necessary to know about for simple cases) called "actions". An action is where keyframes are really stored. Just like objects contain mesh data, actions can be thought of objects that contain keyframe data. Actions also contain curves which specify how the keyframes make the transition. This makes actions a kind of container that contains all the data needed to animate something.

Actions can be linked to objects and the properties that the action changes over time with its keyframes and transition curves will animate that object. You can apply multiple objects to one action. This is the reference count number that shows up in places. If the reference count drops to zero because you’ve unlinked all objects from an action, it will disappear when the animation system recomputes. To prevent this and save an arbitrary action disassociated from any particular object, you can make it have a "fake" object. This is like "phony" in Make and just keeps it around and valid though it doesn’t really have a real object target (yet).

To create this reference to a fake object you can "Save" an action by clicking the shield icon in the Dope Sheet editor’s Action Editor mode. Similarly, the X there is to delete or, technically, unlink actions. This is also where you can create a new action to work with.

Sometimes there is an unlabeled orange bar near the bottom - just below the marker labels if there are any. This is showing "the cache". This seems related to Scene Properties → Rigid Body World → Cache. It seems that this line indicates which frames the simulation system starts and stops which is defined globally only once in the Scene Properties section just mentioned. The simulation start and end frame setting seem quite important; if the system isn’t even trying to simulate physics in the frames you expect, then things won’t work.

If your system came from elsewhere but is not behaving how you expect when you modify it, look specifically there at the "Bake" and "Free Bake" (un-bake) setting.

There can be confusion in Physics Properties → Rigid Body → Settings with the "Dynamic" and "Animated" checkboxes. The tricky thing is that you may need to "animate" the "Animated" setting — this means that you’d want the rigid body physics to be active during some part of the shot but not the whole thing. The confusing part is that "Animated" can kind of mean "not animated" because it means you’re animating "by hand" and not letting the physics simulation take over. It’s confusing enough to put keyframes on these checkboxes but keep in mind that "the Animation System" as shown in the tool tip for "Animated" means the normal one where you do the animating with keyframes, not the physics simulation.

When the value fields in the properties menu show up yellow that means they are showing the value of a selected keyframe. When they show up green, that is one of the frames which is being interpolated. (So it seems to me.)

The red record button on the tool bar is to do automatic keyframes. With this engaged, LocRotScale changes are automatically assigned to the current time location as a keyframe.

To move all of the animations such that the effect is inserting more frames in the animation, you can go to the dope sheet and press a to select everything. Then press "g20" to move everything 20 frames forward.

The rigid body animation can position an object based on physics, stuff like falling and bouncing off things. The biggest point of confusion I had was on the "Animated" property. This property itself can be keyframed (animating the animating) and doing this is how you activate the computed physics action. What I found confusing was when you want the object to do animated stuff on its own, you must turn off "Animated". Then the RB physics can take over. N.b. that "Animated" means: you, the Blender user, will be animating this by hand - not the rigid body physics system.

Here’s the tool tip on the Animated property: "Allow rigid body to be controlled by the animation system." Since this is found on the Physics Properties → Rigid Body → Settings panel, you think it means the rigid body animation system. No - the opposite. It means the normal manually specified key frame animation system. So that’s quite confusing IMO.

When doing hand-drawn 2d animation it is very helpful (and not at all cheating!) to have some live action video of such action to use as a reference. In the outliner, select the camera and then in the properties panel (where materials and modifiers live) look for the camera settings. There you will find a section for "Background Images" which works as expected. You can even bring in video clips.

I found it helpful to use youtube-dl to bring in the video and then use something like this to remove the audio and crop it down.

The "-ss" is the start time and the "-t" is the duration of the clip, allowing you to isolate just a small set of frames. And the crop parameter allows you to just take a portion of each frame. The crop format is "crop=width:height:upperleftX:upperleftY".

If you want stills you can do that too by just having the output file look something like this: jumpaction%03d.jpg.

Pinning

When doing stuff with Python, it’s good to have an "Info" pane open. By default this will show important messages that have registered to be shown. But if you want to see all possible messages you need to activate that explicitly with this.

When developing, this is extremely useful.

Of course fancy addons will want fancy modules. And since this is not system Python, it will appear that they are missing even when your system can use those modules just fine. You must first install pip in your Blender python.

Now it looks like GDAL is the real PITA. Sure Debian can get one no problem with apt install gdal. But things are tougher in the world of Blender. You have to go to this repackager’s repo and pick the right one. I guessed until I found one that worked. No idea what the compatibility parameters are.

This also downgraded the Numpy. Why not? Go for it! After that, I could go to the Python console in Blender and import gdal came back with the goods.

The one Python function that did not work which was a bit of an impediment for my purposes was the input("Prompt:") command (formerly known in Python2 as raw_input). I understand the limitation but what if you need to send user input to your Python activities?

This helpful tutorial has some answers. It says…

From the Python console this will dump vector (object) coordinates for selected vertices.

Use i.index to get the vertex number instead of the coordinates.

bpy.data - Project’s complete data. bpy.context - Current view’s data. bpy.ops - Tools usually for bpy.context. bpy.types - bpy.utils -

The from_pydata() function is a way to take Python data and make Blender data. Unfortunately, the documentation is perhaps hard to find and probably non-existent. However this is helpful. And this. And this.

Basically you need something like this.

In the from_pydata() function call, you need a list of vertices which are 3 member sets of floats. The second field is edges and the third is faces — you can only have one of the two. Leave the one you don’t use an empty list. The edge or face list is a list of sets connecting vertex indices. So an edge connecting the first vertex to the second would be [(0,1)]. There can also be quad faces with 4 values per set.

It’s common to see some structured names in class definitions, for example in the examples found in /usr/local/src/blender-3.2.2-linux-x64/3.2/scripts/templates_py. These have the format of ADDONNAME_XX_class_name. The first part is whatever your addon is to kind of isolate the namespace you’ll be using. (Not sure what to think if you’re not exactly making an "addon" per se.) Next comes the type code and they are as follows.

Then comes the name of your class.

More details about this are described here though I could not find it in official docs.

The bl_info dictionary is defined outside of any classes and contains helpful information about the addon.

The category field can be 3D View, Object, Modifiers, Mesh, Import-Export, maybe others.

Here is some code that creates a simple renaming addon. It creates a text box in the Object Properties menu to rename the currently selected object.

It can be very frustrating to have everything ready to assemble, but no sound is playing. Here are some ideas about that. But what I had experienced was far more ridiculous.

I had to go to Edit → Preferences → System → Sound and change Audio Device from PulseAudio to None. Ah, but then comes the tricky bit. Then you have to go to Edit → Preferences → System → Sound and change Audio Device from None to PulseAudio. It’s obvious really, right?

The first time you run Blender, there are probably some things you will want to adjust.

After you make all these initial changes, it is wise to not repeat the process every time you use Blender. Go to the main "Info" section’s "File" menu and choose, "Save Startup File". After doing that, you’ll be loading up Blender with your presets ready to go.

Often you just want to do a simple thing like cut off a bunch of stuff at the beginning and end that you don’t care about. The basic process is as follows.

Reasons for doing this might include the following.

The way to do this is to create an image separately.

Similar to overlays, this technique can help present multiple simultaneous video streams. The perfect case example is trying to visualize a side by side comparison of two graphics cards. Assume that I take a video with card A called A.mp4 and a video with card B called B.mp4. I want to show the left half of A on the left of the screen and the right half of B on the right side of the screen.

That’s it for the transform strip. Make sure the transform strip is on top. The B strip needs to be visible, but you can turn off A’s visibility and just let the transform render what is needed from it.

Here are some settings hints that did work.

Click Animation to begin render of the video.

It seems that a completely reasonable way to study video editing in Blender is to watch some videos on the topic edited therewith. The videos I found helpful were by "Mikeycal Meyers". The problem with the videos was that they were so comprehensive and patient that there are hours of material. That is a worthwhile exercise to initially learn Blender video editing, but after the first viewing, I found I needed a simple reference to the stuff he talked about. Besides providing a quick reference for cryptic key bindings, if I still have trouble, this list of what the videos contain can direct me to it. I commissioned my son to make the original list this is based on.