CG Effects for live action Feature Film

Here’s a little video, what my crazy creatures look like in the film.
(obviously downscaled and highly compressed.)

It’s actually great fun using cloth sim this way. It adds so much ‘life’ and chaos to the animation and makes it look much more ‘real’.

But on the way, to get there I encountered several issues, I guess most noobs bump into. :stuck_out_tongue_winking_eye:
Like having overlapping vertices in the cloth mesh, which completely freaks out the physics system.
Luckily this issue is easily solved, once you found out about, that is. (Just merge vertices by distance.)
Another super simple solution, once you know about it: the subdivision of the cloth has to be identical for the viewport and the render, when you bake the sim!! Otherwise your cloth will go bazooka.

In some less wild scenes I pinned parts of the cloth to the creatures. This was more of an emergency solution, though. It was the only way, to ged rid of jittering parts of the cloth in slow, calm animation. As soon as cloth is hanging it’s perfect. But the parts constantly laying onto a complex collision mesh (the creature), just keeps wobbling slightly. Obviously the higher the subdivision of the cloth the less obvious. But I could never get rid of it entirely. So I had to pin these parts.
If anyone has a solution for this jittering issue, I would be very thankful to hear about it.

In general, it would be amazing if someone could make a tutorial about all the issues you might run into when doing cloth sims (on characters especially). And about all the values and their actual effects in the cloth physics tab.
And hey…if you can fit all of it into one minute…even better! Ha, ha, ha.

I render the creatures in Blender as transparent PNG sequence - and then composit them in After Effects. I only use the real life footage in Blender as camera background pictures - as a reference for camera lenses, angles and lights.
To render the creatures I have to use Eevee, by the way. (As seen in this clip above).
Cycles render times would be almost manageable, but would slow down the whole process extremely. I prefer to keep moving and experimenting, which is only possible with Eevee.
Luckily there is a trick to get object motion blur in Eevee by using a vector pass from Cycles.
(Thank you Dylan Neill for the great tip and tut.)
This instantly lifts the animation on a different level.
Unfortunately the alpha channel is not blurred correctly and the motion blur is not transparent enough. But it works ok in my rather dark scenes. It actually creates a special kind of ‘look’.
But maybe somebody knows how to blur the alpha channel perfectly when using this trick?
Would be super helpful…please…anyone?? :wink:

So…now I wanna have some more fun with crazy, wild cloth frenzies!! :smile:

1 Like

Oh, one more thing!! :stuck_out_tongue_winking_eye:

As the longest render times are used for baking the cloth sim, I made the animation quite a bit longer than needed for one shot. Then I let the cloth bake over night, or during lunch. (Well…‘my’ lunch might be dinner for other people, ha, ha, ha.)
So I can use different parts of the baked animation for several, continues shots.
As it’s not possible to move the baked cloth sim, without breaking it, I rigged the whole scene (camera, lights, floor & ceiling bounce surfaces and so on) to an empty which is positioned exactly where the creature is. This way I can easily circle the creature with the camera without moving the baked cloth sim. The whole ‘set’ basically moves with the camera around the creature.
Easy and quick to do multiple angles and framings this way. (Obviously when the camera get’s closer, I only move the camera, not the whole scene.)


One Empty controls the whole scene (except the creature with the baked cloth sim).

This approach seemed much faster and easier than all the other solutions I read about (exporting the baked sim or turn it into shape keys). At least in this case here, with such a simple scene.

Btw, one thing I’m really, really missing in Eevee for this kind of work is a real shadow catcher function.
The workarounds (basically luma-keying a shadow on a white surface) often don’t work in my cases.
And pleeease…dear devs…don’t make me create a green screen set in CG! :rofl:

1 Like

Another little chapter in my quest for cinematic ‘gory’. :wink:

Unfortunately this time, I have to talk about the first real problem I ran into. An issue I could not solve so far. Just when I thought I grasped the basic challenges and possible issues of cloth simulation - I got slapped in the face and told otherwise.

So please, please, if anyone is experienced with cloth sim and could imagine what the problem might be - your help will be greatly appreciated!! My endless love and affection will be yours to behold!

I just quickly want to mention it here, as I explained it in detail in a separate topic:

The problem is, that a cloth sim would not bake nor cache. If I try to do so Blender either freezes (caching) or pretends to work… but actually does nothing (baking). This might be common in politics - but for a software that’s not a wonderful feat.
The strange thing is, that I tested an early version of the cloth, which worked perfectly fine. Then I extended and refined it, and actually made the mesh look much better - but it doesn’t work anymore.

Here’s a little video of what the test looks like:

Old version - No Problemo

New Version - Problemo!

Anyone has a clue what the problem might be?

EDIT:
The answer is…eh…yeahhhsomehow.
User Photox gave me a great tip for a workaround!! Kudos to him again. :pray:
(Append the cloth sim in a new project.)
So it works now!! Yippie! Bakerman is baking bread…again! :grin:
I still don’t know what the problem was - but…hey…who cares! :stuck_out_tongue_winking_eye:

1 Like

My dear friends - how the world has changed since my last post!!! :astonished:

But I guess we are all hearing and thinking way enough about it and we wanna talk about happier things here…so just this much:
When hell break loose and countries started to close their borders I was caught abroad…with an old 2013 Macbook Pro only. (At least it has 16GB RAM.)
As I didn’t want to waste the time, I still tried to work on the effects of my movie, though. Ironically the only scene I could work on was one of the biggest. (I did not have the material for the other scenes with me.)

It’s the final shot of the film. The camera follows a car from the top, driving along a highway, then tilts up and shows the landscape ahead.

We shot this scene with a drone on location, but unfortunately technical issues rendered the footage unusuable. It only served me as a rough guide.
So - Blender to the rescue, once again!

This is the basic setup…


A highway leading through an Asian swamp/lake landscape.
(I kind of like this side view. Unfortunately It’s not planned to use it in the film.)

As you can see on this picture I really, really had to try to keep it as lowpoly as possible. It was already a huge challenge to do this scene on such a limited machine. I think I counted every single vertex of the grass models. Ha, ha. They are super-rough, but from a distance they kind of work. I’ll never be as close to it as on this picture.

So this is the idea of the shot: We’ll start with such a kind of perspective…


…and then tilt up, while we still follow the car.


Finally the camera tilts up into the sky.

To keep the scene somehow manageable on my old Macbook, I tried to simplify and reduce the scene as much as possible. While still keeping it look ok.
For all the trees in the background I used flat billboards. (Some with arrays, some with particle systems.) For the grass around the water and near the shores I used simple grass models I made. Since the perspective is shifting quite a bit, flat billboards would look too cheap, I fear. I kept the grass models very simple though. And used even more reduced once in the distance.
My problem is, that I want to keep such a scene ready for some additional camera movements.
If I would decide on one specific move, I’m sure I could reduce the scene even more. Plus, I’m sure there are some more tricks I could use, which I’m not aware of. But I’m already quite pleased, that I managed to set up this scene under these circumstances.
The fine-tuning I have to do back home, on a more powerful machine, though. Eg. the water puddles on the road. Reworking the shore line. Improve the water, Or the finalizing of the car. Just to render a small render region of the road takes several minutes here. Not a very intuitive way of adjusting and finetuning values. :wink:

For the background I created several lowpoly assets. Like this Thai chicken farm.


Again, for how much we see of it in the final shot it’s too detailed…but…yeah…you never know.
For those little countryroad / farm lights I did not use a mesh light, by the way. I used a point light which lights a pure white ‘ball’. It renders faster and creates less noise than meshlights. I did a little test, and it looked that way.

During the work on this scene, there was one thing which really slowed me down. Some function I would have loved to have, but could not find. Maybe some of you know about it?
Is there a way to mute (toggle on and off) all particle systems of all objects in the scene with one push of a button? Like the viewable object types?
This would be soooo helpful to speed up viewport performance!
I don’t want to change the particle viewing options in each particle system of each object. I’d just like to click ‘show particle systems’ or ‘don’t show particle systems’ - in the viewport options.
Is there a way to do this?

So, now I wish you all a very creative time - get the best out of it…social distancing is literally another word for ‘blending’, my friends!! :stuck_out_tongue_winking_eye:

5 Likes

This work you’ve done is stunning. Thanks for sharing the journey so far

1 Like

Thanks a lot Gregg. That’s very kind. I really appreciate your comment.

1 Like

The closest I can find is the ‘Global child particles percentage’ setting, under Render Properties > Simplify > Viewport.

Nice work on that scene by the way, especially considering the current limiting circumstances.

1 Like

Thanks for the tip Cire. Yep, that would already help.
Just a bit here, as I’m not using a lot of children for this scene.
But it’s very useful to know for other scenes.

Looking great! Keep pushing through. :metal:t2:

1 Like

Just a tiny little scene today - a close up on a comic book.
Another shot, we had no time to shoot during principal shooting.
Or… maybe,we simply forgot to shoot it. I actually forgot, if we forgot… :wink:

Anyway, just a simple scene - took me about half a day to do it.
Big part of the work was for texturing and to match it to the existing footage.

This is how ridiculously simple the scene actually is, ha, ha, ha.
(Only the tiles with microdisplacement are missing in this unshaded screenshot.)

The next scene will be quite a bit more complex again…I promise… :grin:

3 Likes

Is that a real comic? Or a fictional comic that exists in the world of the film. :laughing:

And great work on this project, not sure if I ever commented here, but it’s been cool seeing the development of your project over the months. And seeing how you use Blender alongside live action footage.

1 Like

Thanks Avery for your motivating words.

The comic book only exists in the film. I wrote the storyline. But I only had a few pages drawn by a great artist friend of mine. Short parts will also be read in the film.

It’s a little cyberpunk story about a biotech company conducting experiments on poor people to create emotionless slaves for labour and pleasure. One of their ‘experiments’, YOYO, goes beautifully wrong…
Obviously not a very creative premise - but there is a unique, poetic twist to it.
I would love to get the whole book drawn. If only I had the time…

Funny thing, YOYO is not only a comic book in a movie…there is also a comic book inside the comic book inside the movie. :laughing:

1 Like

Sounds pretty neat! Hope finishing the rest of the project goes smoothly for you, especially in this time of quarantine. Everything is harder or easier to do right now!

:laughing:

1 Like

In my last post I promised a more complex scene…well…hmmm…not totally ready for that one.
So today I want to write about a jellyfish I’m working on for the movie.

Actually it turns out…it’s pretty complex as well. At least for me. :wink:
Creating and animating a jellyfish would be THE perfect tutorial subject. So many things involved. From basic modeling, uv, texturing, procedural nodes, shape keys, modifiers, to hair and cloth physics.

I’m not finished with the jellyfish, yet. My old macbook is simply not powerful enough to elaborate the details in animation and textures. But I think it already looks decent.

Well, here is how I did it so far:

I started gloriously by making a really bad choice. I chose an icosphere as the basic shape, cause I stupidly thought its shape would help create organic shapes and would be less perfect than a uv-sphere. Only later I noticed, that you can’t use loop cuts on icospheres!! (Cause they’re made of tris! Stupid me!! Now I would probably draw half of the outline and use a screw modifier.)
For the main animation of the bell of the jellyfish, I used 5 shapekeys. I first tried to do it with two or three wave modifiers. But I couldn’t get the “push upwards - then slowdown” cycles of actual jellyfish to work this way. So in the end I used shape keys and just one subtle wave modifier to add an organic, watery feel to the animation.
To loop the shape key animations I used the cycle-fcurve modifier. Extremely useful.
I’m actually quite happy with the animation I ended up with.
I parented the bell of the jellyfish to an empty and moved the empty along a curve.

The texture of the bell, especially the veins running along it, I hand painted in photoshop. Light will run along those veins, when animated. The rest is different layers of glass, some procedural nodes and a bit of layer weight light.

As it’s very chaotic and I don’t want to shock other people with it - I only show a tiny pic of the node tree. Ha, ha, ha. Feel it - don’t read it! :grin:

Then I created the inside of the jellyfish. Like everything in this jellyfish, the inside is fully animated - shapes and light/textures. I used 3 little copies of the bell with lights moving inside them (wave node) as central flowers and base for the three central tentacles. Around them I created two different colored ‘brains’, which I based on some strange shapes created with the ‘extra shapes’. (I don’t remember exactly which one I used. I think the ‘math’ one.) I animated those with displacement modifiers, and their texture also with the wave node. Finally I added two rings, which I animated with shape keys… the glowing spots on them are animated too…of course. :wink:
I parented all of the inside to another empty which I animated to match the shape keys of the bell. And then parented this empty to the main jellyfish empty.

Then I created the three central tentacels. Just some spiraled and deformed long planes with noise and solidify modifiers. And finally a cloth modifier. Took me quite a while to find the right physics settings for the cloth, to make them move the way I wanted to. I also kept the settings slightly different for each of the three tentacles, so they move more organically.
I made them look ‘glassy’ and have light running along them. As for the bell I also used some layer weight to enhance the glow on the outlines in the lit parts.

And finally i added two hair particles system to the bell. Short, thin hair just under the bell, which beautifully extends the bells movements. And long hair for the outer tentacles.

Kudos to everyone still reading…I think this post got much longer than I planned. Sorry for that.
You know…it was my first jellyfish. Everyone likes to talk about their first jellyfish… :kissing_closed_eyes:

ps: apart of the first one, pictures here are just viewport screen grabs - not renders.

12 Likes

Beautiful work! Each update makes me more curious and eager to see the final film.

2 Likes

I’m in love with your work. :heart_eyes:
Pls keep it up, I’ll like to see the end result.

1 Like

Thank you Cire.
You guys are great - so motivating and supportive.
It’s a delight posting work here.

1 Like

Oh wow, Paul! Super happy to hear that. :blush:
Merci beaucoup.

1 Like

And up to the next scene…social distancing creates lot of time for CG work. :wink:

The purpose of this shot is an animated bird’s eye view of the protagonists car driving through a street in town. The further part of the street being under water. Driving into the flooded part of the street will serve as metaphor for moving from reality into a surreal world.

As with the last few scenes I can only prepare this shot, due to my old computer here. I will only be able to elaborate the details and render when I’m back home later this year.
But I thought it’s still interesting to show, what I got so far.

So these is are two low quality pictures of what the scene will look like roughly…


The final result will have much better render quality - and several compositing effects (in after effects) like smoke/haze, lens flares, light leaks, lens distortion and so on.
I also plan to add more details/objects on the sidewalks and facades. But this is for later…

The most interesting part of the shot is the camera move. Even with basic viewport shading it looks really cool. (Sorry I can’t show this here. But I will show a final render once it’s done.)

A basic overview of this scene can be seen here…


As you can see, I only modeled what’s really seen in the shot. Facades which are not seen, I left out.

What I’d like to talk about in this scene is my quest for lowpoly solutions.
I did this scene here on my old Macbook - so I knew I had to save polys wherever I can.
Lowpoly meshes can work surprisingly well, if the textures are containing a lot of details and have a good quality. Additionally, the light has to be used the right way, in order not to make the simply shapes of the meshes stand out.
Textures I all shoot by myself, by the way. So I can make sure, that they work for my purposes.

One good examples for this lowpoly approach are cars.
I did another topview scene before. And a big issue I had, was the amount of polygons good car models use. To make a topview shot of a street look believable, it needs several cars parked at the edge of it. But the car models I have (bought) all use hundred thousands of faces. If I remove all inner parts and decimate the remaining meshes I still end up with several 10K. If I decimate more, I get strange shading errors. So I hardly used any prop cars in the scene. But this got me thinking - I was building lot’s of low poly facades - why not do the same with cars. If they are only seen from one perspective, and only from far away, shouldn’t that work?
So I did a quick test to create ‘photorealistic’ (from really far away, ha, ha, ha) car models with max 100 faces.

And I thought, yeah…this could work. I might give them a few more faces. Extrude the windows for example. But it’s definitely the right way.

So a scene like this… (this perspective will never be used - the car models are not made for it, but it’s a good example)


can be achieved with rather few polys, I think…

Here’s an example from an even lower angle.


The first facade on the left I actually made just for very distant buildings. But it might work in this birds eye view shot as well, I think…I hope. :wink:
(For closer shots I have much more detailed versions of similar facades.)

When the car drives into the flooded part of the street, I will use dynamic paint to create the waves. I’m currently experimenting with it, but am not happy yet with the results. It looks kind of ok, but is pretty far off reality. I suspect, that Blender’s dynamic paint system is not made to create realistic water ripples needed in such a case. But I will tell more about this in a later post…

I’d also like to mention that the trees in this scene are not modeled by me. They are from the ‘Realistic Trees Asset Pack’.
The Mercedes hero car, I also bought, but remodeled and retextured it to fit my purposes.

Cheers, my friends.

9 Likes

Nice work! Looks amazing.

1 Like