Hi team! Very intrigues by the tool so far and very excited about its potential based on the changes already made in beta 2. I had a project recently that required a very quick turn around, and due to bugs on my cloud rendering service I had to render it here on a single workstation. This seems like the perfect application for something like Lavina, so I did a comparison. You can see a direct video comparison at the link below:
https://youtu.be/Bpgj5aO6A3o
For the final deliverables at the time (1 30 second clip & 1 20 second clip), I used vray progressive at 1/100 with Brute Force/Light cache/nVidia denoiser. For the 30 second clip seen above, each off the 900 frames ran for 1.5 minutes, so I let it go overnight for about 22 hours. There's some obvious flickering issues, but it was good enough for this task. Once I updated to a RTX 2080 and downloaded Lavina beta 2, I exported a vrscene file and rendered that way to compare (default settings of 100 samples). Here's my thoughts on the comparison:
PROS
- WAY faster...3 hours compared to 22
- improved animation experience (Scene) made it super simple to export (compared to beta 1 issues we were having with only single frame exported)
- easy to tweak LUT, bloom, DOF, material glossiness to get an acceptable look
CONS
- WAY more flickering than low progressive CPU settings...don't think this quality would have been acceptable in this case
- the glossy floors just aren't getting the right kind of contact shadows that they do in standard rendering...everything feels like it's floating, probably because of the next item...
- anything other than 100% denoiser creates an unacceptable amount of noise that never resolves away (in fact, gets progressively worse)... attached image
- some stronger halo artifacts around various objects
- grey objects had unsupported texture maps (animated or Advanced Wood...not a big deal to replace here)
All this said, I think Lavina will be great for sending pre-render reviews for clients, unless it's just a matter of tweaking settings to get the bigger noise issues to go away (higher samples?)
Thanks...I'll keep checking in with tests on new versions.
John
https://youtu.be/Bpgj5aO6A3o
For the final deliverables at the time (1 30 second clip & 1 20 second clip), I used vray progressive at 1/100 with Brute Force/Light cache/nVidia denoiser. For the 30 second clip seen above, each off the 900 frames ran for 1.5 minutes, so I let it go overnight for about 22 hours. There's some obvious flickering issues, but it was good enough for this task. Once I updated to a RTX 2080 and downloaded Lavina beta 2, I exported a vrscene file and rendered that way to compare (default settings of 100 samples). Here's my thoughts on the comparison:
PROS
- WAY faster...3 hours compared to 22
- improved animation experience (Scene) made it super simple to export (compared to beta 1 issues we were having with only single frame exported)
- easy to tweak LUT, bloom, DOF, material glossiness to get an acceptable look
CONS
- WAY more flickering than low progressive CPU settings...don't think this quality would have been acceptable in this case
- the glossy floors just aren't getting the right kind of contact shadows that they do in standard rendering...everything feels like it's floating, probably because of the next item...
- anything other than 100% denoiser creates an unacceptable amount of noise that never resolves away (in fact, gets progressively worse)... attached image
- some stronger halo artifacts around various objects
- grey objects had unsupported texture maps (animated or Advanced Wood...not a big deal to replace here)
All this said, I think Lavina will be great for sending pre-render reviews for clients, unless it's just a matter of tweaking settings to get the bigger noise issues to go away (higher samples?)
Thanks...I'll keep checking in with tests on new versions.
John
Comment