Welcome to the week nine round up of the Hangar 45 student project! Here you will see some of the progress I have made with this task on a weekly basis and see me talk about some of the challenges and (more importantly) solutions that I deal with.
Project Phase 2 COMPLETED
This week I’ve managed to get all models essential for the final project textured in-engine with sufficient materials.
The process took roughly half a day per mech once all material and detail templates were ready – in this state the characters fulfill the minimum project criteria, which is why I pushed to get the all to this state. I do hope to have some extra time to polish a few parts on each character at the very end, but until then all other efforts are going towards building the final scene.
Project Phase 3
Having wrapped up my characters it is time for Phase 3 of the project which is presentation and optimization.
After EXTENSIVE testing I have decided not to include the Nvidia developed real-time global illumination tool. This is mostly due to the fact that in my current set up it makes little visual difference… that is due to the fact the VXGI is by default mounted to the player position in the level (very efficient!) and has a very limited range (also efficient), however this means that the effect rarely covers an entire character in the scene, let alone the entire level.
To fix this I tested both anchoring the effect to a blue-print actor on the level, which allowed for GI within a limited area, but produced horrible artifacts
The next test was to try and expand the GI range both from the anchor and it’s character mount state through console comands… this killed performance to around 10 FPS…
To get around this I edited the post-process settings to get a balanced result between performance and visual quality and managed to get a performance of 20 FPS.
Next I decided to do some on and off tests that revealed that in the end due to the lighting set up (mostly a static HDRI and a skylight) the end results showed very little improvements at the cost of 10-20 FPS – and that was that the effect was cut.
The extensive testing however revealed that HBAO+ (a component of the VXGI UE4 branch) had a great effect on visual quality at the cost of 1-3 FPS. With a lot of option editing (VXAO proved difficult to control and edit at a higher performance cost and was also cut) I managed to get a GI-like effect form HBAO+, while overlaying a light UE4 AO effect on top of it. This produced great results at a fraction of the VXGI performance cost.
Next steps are to start collecting all visual effects from my old engine tests into one scene and continue to optimize. So far My goal is to get an FPS of around 30 for the project so I will try to cram in as much as I can.