Duration: 1:25

1,198 views

14

This is an update to the "Meerkat Demo" (Created by Weta Digital) project that is available from Unreal Engine Market that provides an excellent example of 4.2x animation, hair and screen capture GI, but updated to Unreal Editor 5.1 (GitHub 12/10 UE5 Main Build) and lumen.

In doing so many of the materials had to be modified, and a number of lights were either turned off or attenuated to allow Lumen's excellent GI to do its magic. It is notable that I had to make a few tweaks to the eagle rotation on some frames to prevent a light saturation effect to the material. And if you look closely as the eagle is coming out of the sun, there is frame I left to show what happens. This was done on a Github UE 5 build from mid December so maybe this will be fined tuned, but I think it has more to do with not tweaking every material. I should also be noted that like standard reflection captures sphers and screen capture reflection modes, character meshes don't appear. If you look closely in the original project in the eyes of the Meerkat as the eagle is almost upon it, there is no reflection. Of course this would only be for a few frames, but nevertheless I decided to create a static mesh of the character mesh at that moment and add it to the sequence for those frames to capture the reflection and appropriate lighting. To add even more realism, I moved the static model towards the Meerkat for even greater effect. Of course you will never notice unless you do a frame by frame analyisis, but it was worth doing. Of course I could have just enable RTX reflections that does allow character meshes and provided even better reflections, but this was test with "software" lumen only on a standard Intel i9 and 128 GB memory.

I also took that time to add spatial audio by attaching the audio track to the Meerkat and using modified 'Sound Attenuation" to spread out the 2-channel original track to four channels of a 5.1 channel layout adding more depth. It would have been more convincing if they had laid down each foley or creature audio track separately, but it appears it was mixed outside of the Unreal Engine, which sadly the Unreal audio developers seem to think this is the best way to do this and spending more time in turning Unreal Engine 5 turning it into a synthesizer instead working with the Sequencer developers to allow full mix downs to a full spatial track series using Microsoft Spatial Audio API. Which I have found I can do, but only by creating 2 passes with the camera turned 90 degrees on its side to capture up/down channels and using Premiere to layout a conversion to AmbiX. If you are looking for stereo, 5.1 or 7.1 audio, you can export it as a multichannel WAV file, but I found I needed to export it and import into a Audacity and export it again for it to be muxed properly with the ProRes LT video file to create a 6 channel AAC file in FFMPEG and uploaded to YouTube. The original MOV was 3.7 GB for less than two minutes of animation. Not the most efficient, but I have found it to provide the best transfer to YouTube. It should be noted using "-copy" is the best way to use FFMPEG since it will do a direct transfer without transcoding the output. If you need to transcode, I highly suggest x.265 with nvidia GPU transcoding. Super fast and the results are very good.

One more note: Even though this could have been rendered in real-time with my RTX-3080 with Lumen at '4' quality. I needed to wait a few frames to allow Lumen to settle down before the next 'shot camera' was rendered. Even so, it only took 10 minutes to render both the 4K MOV & WAV tracks using the legacy 'Movie Scene Capturer'. I tried Render Queue, but it needed a set up for each shot, which wouldn't have given me more quality unless I wanted to use a lot of frame samples for anti-aliasing which would have taken much longer to render. I really hope they don't pull out the old 'Movie Scene Capturer' since 'Render Queue' doesn't do full mixdown audio and cannot capture orthographic camera output.

Oh, and happy new year.