It brings text-to-character motion generation directly inside Houdini, using Kimodo for motion generation and KineFX for importing, retargeting, and continuing the animation workflow inside Houdini.

Demo: 
https://youtu.be/dqLJe_3Qlic?si=TEZFXFNnDwSD1ndD

The integration is currently available in the Houdini AI Assistant build on Gumroad.
Houdini AI Assistant:
https://rart.gumroad.com/l/HoudiniAIAssistant

Current workflow:
Prompt → Generate Motion → Adjust Character Position → Regenerate

The current build supports:
- Text-to-character motion generation
- Multi-prompt timeline
- 2D root waypoint controls
- Full-body keyframe guides
- Hand and foot guide tracks
- KineFX retargeting workflow

The idea is not just to generate a motion clip externally, but to make the result editable and usable inside Houdini. The generated motion can be imported as BVH, retargeted through KineFX, adjusted, and reused as part of a Houdini character workflow.
At the moment this is still an early Windows beta and focused on humanoid motion, since Kimodo is trained around human skeleton animation.
APEX support is something I’m also looking into.

I’m currently testing different workflows around:
Timeline-based prompt segments
Motion transitions between prompts
2D path control
Full-body constraints
Hand / foot end-effector guides
BVH import and retargeting reliability

I’d be very interested to hear thoughts from Houdini users working with KineFX, APEX, retargeting, or character animation workflows.
Houdini AI Assistant:
https://rart.gumroad.com/l/HoudiniAIAssistant



Would this kind of prompt-based motion generation be useful in your Houdini pipeline?
And what would be the most important control features for production use?

Thanks,
Radu

I'm trying to rig a simple hexapod like robot, with APEX.

I've watch quite a few tutorials but never found info on how to fine tune an IK (locking some axis and limiting the range of others).

It's pretty much a mechanical arm rig.

If anyone can help/point me in the right direction... thanks !

6ril_IK_APEX.hipnc

I'm having most trouble understanding the procedurals, with the above link he uses a HDA which has been very helpful but then mentions reimporting it back into sops animating and then importing back to LOPs, I dont quite understand this process and any insight would be super helpful!
If anyone could help I would be much appreciated

I would like to ask for some advice. I’m not entirely sure what the correct term for this effect is, but I really like this animation of a tennis ball spinning on top of another ball.

Could anyone please guide me on how to achieve this effect? Any tips, recommended techniques, or tutorials would be greatly appreciated.

Thank you in advance for your help.

I’m having trouble achieving the expected muscle definition using OTIS in Houdini.

For some reason, I’m not able to get the same level of visible muscle detail that I can achieve with Vellum (please see the attached reference images). In terms of realistic muscle deformation, the results from OTIS appear much flatter and resemble something that could easily be achieved with a simple blendshape, rather than exhibiting natural muscle bulging and volume shifts.

Additionally, I’ve noticed that OTIS does not seem to produce realistic muscle behavior similar to Ziva, where muscles naturally bulge and shift upward during contraction. Instead, the deformation feels limited and lacks dynamic volume change.

I’ve attached my Houdini file (Version 21.0.631) in case anyone would be willing to take a look and help me understand what I might be missing or setting up incorrectly that’s preventing proper muscle definition with OTIS.

Any guidance would be greatly appreciated. Thanks in advance!

Ottis Leg1.rar

Any pointers are appreciated, and I'm happy to go through an hour long KineFX class if the answer is five minutes in the middle--I'm still learning KineFX and am happy to be taught how to fish.

Thanks!

ADDENDUM: I had another thought that might be possible, but I'm not sure how to implement it--since joint movement is all rotation, I can calculate angular velocity using the point velocity node and then multiply that in a wrangle, but I'm not sure how I would then apply that back to the joints. A solver, perhaps?

I think it's  a dictionary, but I can' t get the syntax right.

What is the syntax to access the string  or the float in : 

boneCapture regn[0]

or

boneCapture w[0,0]

How do you even parse: "boneCapture regn[0]"

I have a question…  after much troubleshooting importing blendshapes, I figured out what the problem is, but I don't know the solution.  This character has high heels on, which is affecting the blend shapes I have on this character.  If I import this character with no heels on, then the blend shape is fine and imports with no issues.  How would I correct the blend shape’s transform with this character? to account for having on shoes/heels/etc?

Thanks,

Mirta

I have a bunch of points that all have a vector attribute as in the ss below. I then turned these points into joints and want to align the generated joints' Y axis to their respective vector attributes. I used a rig attribute vop that runs on points and used a foreach transform vop node to loop over each joint. But the problem is that although I could bind the per point vector attribute into the loop, I just couldn't get the foreach transform to work on each point's respective vector attribute, it only works on the first point's vector attribute. This is so frustrating

How on earth can a such a simple task be done in a rig attrib vop?

Thankx in advance;

AJ

ALIGN_JOINTS_TO_VEC.hiplc

I am trying to create a APEX graph, which loads animated lines, applies a sweep, and displaces the columns of each blade (marked in red).

Inside APEX:

It is possible to output the column attribute inside the sop::sweep node. After the sweep node, I loop over every point. Inside the loop, I multiply the column Attribute

with the transform I would like to apply. (In this case the transformation vector should get multiplied by values of 0,1 and 2 for each column)

However, the geometry will be transformed as a whole. I know it is a logic/skill issue on my side, but how can I define points or groups and locally transform individual points (or other operations) in APEX?

I am fairly new to APEX.

Thank you very much!

Cheers

Jules

APEX_Groups.hiplc

test_anim.hiplc

I want to create a scene where a ball flies in and collides with the tree, causing it to shake violently.
However, I need to ensure proper collision handling so that the ball doesn’t pass through the tree, and the leaves and branches do not clip through the ball.

How can I achieve this?

I've just started to learn apex/kinefx, but ran into some obstacles. The goal: though apex ragdoll is quite promising, I need some more control for my ragdoll animation, so I've extracted the ragdoll controls from electra and made a separate ragdoll setup with ragdoll solver. After that I would like to feed it back into the apex context without retargeting as its the same skeleton that I've extracted. Somehow I managed to work it with retargeting, but then I messed it up, and I'm not even sure if it was the proper way. Now it doesn't work at all and I'm totally clueless what am I missing or what Im doing wrong at this point. Thanks for any help!

Greetz,

Krisz

retarget_recovered_recovered.hiplc

I have literally tried everything and by everything I mean every single option that is available in vellum solver and vellum cloth conf.

• Substeps and constraint iterations don't help at all.
• Collision passes, don't help at all.
• There is no forces except gravity. Changing friction doesn't help.
• Tweaking mass and thickness don't help.
• Tweaking bend and stretch stiffness/damping/rest length don't help!
• Changing cloth scale doesn't help.
• Remeshing doesn't help.
• Changing objects scale up to 100 doesn't help.
• Cloth itself is a one sided piece of geo without intersections.
• There is a gap between cloth and collision source, so there is enough room for vellum.

I'm running out of options here.

Has anyone experienced that? Could it be a bug?

Someone suggested the capture override sop might do what i'm after but i can't find any real information on how to use it.

Attribute combine definitely doesn't work and using a capture attrib unpack and lerping the boneCapture_data values in a wrangle doesn't seem to work either because the amount of bones influencing each point seems to change from capture to capture.

Anyone got any ideas on how i can pull this off?

I've been following Max Rose and sideFX tutorials for APEX and loving it so far.
I understand it's Beta and I may not find an answer yet, but does anyone know how I could setup something like on the image attached, where the toes of my character could collide with a ground or some custom shape and adapt to it...
I'm really novice with APEX, and I don't even understand yet if it would be an APEX setup or something done down the road after the APEX setup...
So far I think the toes should have an IK setup, and I was thinking I should feed some surface VDB to those toe's IK controllers to stay "out" of that VDB, with some volumesample/gradient operations, but perhaps anyone have had a similar thing to tackle (or have the sufficient knowledge!) and could point me in some direction!

Thanks to whoever reply or read this!
François.

Next idea is to use Houdini's agent's skeleton and transfer/remap MIXAMO's mesh and cycle animation. This is because the agent's skeleton is promising to be used for the both of crowd and ragdoll systems. I bebelive we just need to use Kinefx properly in order to set a FBX charactor&animation as an agent. However, the both methods have not worked yet correctly.

I appriciate if you could demonstrate how to set up MIXAMO's FBXs to be sim ready for crowd and ragdoll sims. 

Thank you so much for your assitance!

fbx_to_agent.zip

Anyway, if anybody can help me here I would be grateful, thanks!

Best, Christian