[Bf-committers] Multiple soft body settings per object

[Luke Magill]

I don't know if anyone is interested, but I have a video of my current
progress:

https://youtu.be/-qHT1R8aXM4

Thanks!

Luke

On Tue, Jun 9, 2015 at 9:30 PM, Luke Magill <luke.magill at gmail.com> wrote:

> Hi there!
>
> My name is Luke and I've been an amateur Blender user for a few years and
> I love it! Recently, I had a cool idea where I wanted to try to emulate
> muscles of the human body to create a more realistic style of animation.
> This was mostly to satisfy personal curiosity and perhaps to create an
> interesting new style of art. However, at some point I hit a snag where I
> wasn't able to move forward.
>
> The problem can be summarized as follows: the built in soft body engine of
> Blender is great, but what I really needed was to have more than one soft
> body setting per object. This was to allow me to sort of pin different
> things together, like pinning skin to muscle. Skin and muscle have very
> different levels of softness but sometimes the skin and muscle for one
> person may need to share points (be connected or glued together). What I
> needed was to have one edge connected to a point with a certain push/pull
> value and a different edge connected to the same point with a different
> push/pull value (this is just an example, the idea is that any soft body
> setting for which you can select a vertex group would be able to have
> multiple settings).
>
> I couldn't figure out any way to do this or properly emulate this within
> blender. True, there are ways to copy data from one point to another but
> the soft body forces wouldn't transfer which would leave me with an
> unrealistic result.
>
> To solve this problem, I decided to break open the internal C code to try
> and add this capability myself. Now before I get into details I want to
> tell you a little about myself. I got my bachelor's in computer science 6
> years ago and since then I have been a professional web programmer. The
> vast majority of my experience with programming has been c#/Java and
> scripting languages like PHP/Python, etc. I also do front end stuff like
> html, css, and javascript. I don't have a lot of experience with very low
> level code like c or c++. However, I'm always eager to try new things and,
> to be honest, I was surprised at how approachable the blender code base
> was! In any case, please be patient with me if I have any trivial
> misconceptions about c coding.
>
> This email/document is meant to be a project proposal and specification. I
> will include as many details as I can and I invite everyone and anyone to
> provide feedback and criticism.
>
> Feature Description
>
> This section describes the feature without discussing any implementation
> details. Implementation will be covered in the next section.
>
> In general, blender soft body settings can be split into two different
> groups, general settings and vertex group specific settings. The vertex
> group specific settings allow you to select a vertex group (or no vertex
> group indicating all vertices) to apply the settings to. All vertices not
> part of the specified vertex group will not have the settings, and will not
> be part of this animation. This has a disadvantage if you are trying to
> simulate heterogeneous soft bodies for which part of the body may have
> different settings than another part.
>
> To alleviate this problem, I propose allowing MULTIPLE soft body settings
> for the vertex-group type of setting. For reference, here is a list of the
> different settings associated with vertex groups
>
> Vertex Group Specific Settings:
>     Soft Body: Friction, Speed, Mass
>     Soft Body Goal:
>         Goal Strength: Default, Minimum, Maximum
>         Goal Settings: Stiffness, Damping
>     Soft Body Edges:
>         Settings: Pull, Push, Damp, Plastic, Bending, Length
>         Stiff Quads: Stiff Quads, Shear
>         Aerodynamics: Simple|Lift Force, Factor
>         Collision: Edge, Face
>
> All of the remaining settings are general settings and will not be
> affected by this change.
>
> The question naturally arises what will the soft body settings be of a
> vertex that is a member of multiple groups? If we imagine trying to emulate
> jello that is very jiggly on one side but rigid on the other with a smooth
> transition, the most intuitive way to do this would be to have a "rigid"
> vertex group and a "soft" vertex group. Then using weight painting, you can
> smoothly transition between the two groups to create a nice effect.
>
> To enable this, we institute the following rule: a vertex's specific value
> for any setting listed above will be the weighted average of all of the
> settings for all of the vertex groups that apply to it, including the "all
> group" if there is a setting for that.
>
> Now let's talk about how the UI will enable the user to control this
> feature. There are already multiple examples in the blender UI of having a
> list of settings that the user is allowed to modify on a per-item basis.
> Some examples are render layers, textures and materials. The soft body
> settings will work the same way.
>
> First, we will move all of the General soft body settings to the top so
> that the user does not get confused thinking that it applies only to a
> vertex group. Then, we will have a list selector. For an example list
> selector, see the render layers list selector. In the list selector panel
> there will be a vertex group selector so that each item in the panel can
> have a vertex group. The only change in the UI to the vertex group specific
> settings panels will be that they will no longer contain individual vertex
> group selectors.
>
> Required Code Changes
>
> This section describes what I will be changing to enable this feature. The
> DNA, RNA, and Python UI features will be modeled after existing list type
> code that exists within blender, particularly the RenderLayer object.
>
> DNA changes: The settings associated with the aforementioned Vertex Group
> Specific Settings will be removed from SoftBody (DNA_object_force.h) and
> added to a new struct called SoftBodyVertexSettings. The
> SoftBodyVertexSettings will be a linked list, in which the first element
> has a *prev and a *next link (based on ListBase linked list). Methods will
> be added to softbody.c called SoftBodyVertexSettings_add() and
> SoftBodyVertexSettings_remove() to add and remove items to this list. These
> _add and _remove functions will be similar to BKE_scene_add_render_layer.
>
> RNA changes: These new DNA changes will be exposed through RNA. In
> renderlayer, these are exposed through SCENE_OT_render_layer_add and
> rna_RenderLayer_new. I will create two similar functions to expose my new
> DNA. All of the RNA which currently modifies the "Vertex Group Specific
> Settings" will have to be changed to modify the RNA for the currently
> selected setting, rather than trying to modify the SoftBody itself.
>
> Python UI changes: Again relying heavily on RenderLayers as an example,
> RENDERLAYER_PT_layers class is responsible for showing the list selector
> panel in the UI and RENDERLAYER_UL_renderlayers is resonsible for setting
> up the individual items in the selector table. I will have to add similar
> classes for soft body and also a vertex group selector to this panel.
>
> Softbody.c changes: The most extensive changes will have to be made to
> this file, because this is where the soft body settings are actually used.
> In general, the soft body system works by converting a soft body setting
> and a mesh (or lattice or curve) into a set of BodyX objects: either
> BodyPoint, BodySpring, or BodyFace. These objects contain the information
> required to compute each step of the SB animation.
>
> To make soft body have multiple settings, each individual item of these
> will have to be expanded to contain more information. To pick an example,
> inspring (pull in the UI), is currently only stored in the SB object
> itself, like this:
>     iks  = 1.0f/(1.0f-sb->inspring)-1.0f ;/* inner spring constants
> function */
>
> In the new system, BodySpring would have to have it's own inspring, so
> this line would look like this
>     iks  = 1.0f/(1.0f-bs->inspring)-1.0f ;/* inner spring constants
> function */
> (It's somewhat confusing, but sb in this function is SoftBody and bs is
> the BodySpring)
>
> In addition to utilising these new fields in computation, I would have to
> actually add methods for converting the soft body settings into a set of
> BodySprings. This will be done using an averaging algorithm described above.
>
> Filesystem changes: I haven't researched this yet but since I will be
> removing items from the DNA I will need to make changes to the file
> read/write systems both so that the new setting system can be saved/loaded
> but also so that legacy soft body systems can be dealt with in a reasonable
> way. I am hoping to write the core parts described above first, as a proof
> of concept, before delving into details like how these settings are to be
> saved.
>
> Conclusion
>
> Thank you for reading my proposal. At this point I would appreciate
> criticism but I'm also kind of fishing for interest in this feature. If I
> were to implement this would it be possible to have the change integrated
> back to main? I would like to contribute this new functionality if possible.
>
> Thanks!
>
> Luke
>