Embergen by JangaFX - First thoughts

General / 12 October 2020

After a few months getting familiar with Houdini, I came across an interesting program called EmberGen. A couple of days later I'm using it in production. Here are my thoughts:

Background

Embergen example renders (https://jangafx.com/software/embergen/)

Here's the rundown if you're feelin a bit too lazy to check out the website. 

EmberGen enables rapid iteration of volumetric fluid simulations by running on the GPU.

There's nothing much else to it really, except it runs in real-time...

Thoughts

Honestly speaking, I was sceptical at first. Who can honestly believe that there's a program out there that can do what Houdini does but instantly?

Here's the thing, it kind of does what it says on the box. I'll get to cons in a bit but here's some of my perceived benefits after a couple of weeks with the program.

Pros

  1. Iteration. The fact that you can slide sliders around and get instant feedback is amazing. Not only that, but the sliders are a bit more intuitive than Houdini (in my subjective opinion).
  2. Game dev tools appear great. I haven't touched them, but there seems to be a few decent options for the game dev side of things. You can export sprite sheets with multiple passes including: scattering, emission, depth, and normals. Simulations can also be very easily looped.
  3. Damn, can't think of any more benefits sorry 😐 

Cons

So I've mentioned what I reckon is pretty sweet about EmberGen, time to shit on it real quick:

  1. Is it just me or is it very difficult to set scene scale and match things up in external software? God damn I thought this would be easy but nope. EmberGen not only imports at weird scales, it also exports at weird scales. There's no difference between increasing the resolution of the volume and upscaling. Both options just scale the volume up. There's no option to set pivot points so when you export to other programs, you have to go through a process of manually aligning the volume both in translation and in scale.
  2. No sparse solving (yet)
  3. No undo? It's a funny one but yeah, a very useful feature to just not have yet?
  4. Limited support for animation and camera imports.
  5. I could go on but it's beta software so maybe not?

Conclusion

EmberGen is not production ready by a long shot. I say this, yet I am using it in production 🤨? Well I'll tell you why. For some certain situations, it is honestly a heck of a lot faster to use EmberGen than Houdini. From a film point of view, it comes in real handy when you have a small number of relatively low complexity simulations to pump out. Depending on your GPU power you can ramp this up to higher complexity and that's where it starts to get really powerful. As an example, I run a Quadro RTX 6000 with 24 GB of VRAM and damn, like I certainly max it out but that's almost enough! 

EmberGen definitely has it's place in my small setup I've developed over the past week. With the help of Houdini and Redshift, I've been able to get a system going where all I need to do is spend my day making different simulations, and let the computer automatically render them all out over night. EmberGen smashes out sims, Redshift smashes out renders, and Houdini glues everything together.

Where EmberGen falls apart however, is when you need a LOT of different simulations or you need massive complexity. Houdini is better in those cases because you can set it up to automate thousands of different variations. In the case of high complexity sims, well, Houdini is more accurate plus you have access to system memory.

A lot of what I mentioned is addressed in the EmberGen roadmap. Apparently they plan to introduce sparse solving which has the potential to really alleviate memory related pain, not to mention introduce performance related gain. Seems like undo is a planned feature too!

On the Houdini side of things, we have 18.5 coming soon and rumour has it they plan to introduce some EmberGen-like features. What a time to be in this industry! Catch ya's next time!