I have been watching several demonstrations of DLSS 5, including people experimenting with it in Unreal Engine.
One especially interesting example uses DLSS 5 through OBS or a media source to enhance imagery displayed in the viewport. From what I understand, the enhancement may not currently carry through when recording the viewport directly, so this would need further testing.
The screenshots below show it being applied to the iClone 8 viewport with a Character Creator 5 character. It may also be possible to display a rendered scene or still image through the same process and capture improved screenshots for image tests, reference work, or selected production uses.
AI Studio provides its own 3D-guided generative workflow. I thought this was also worth sharing as an additional real-time image-enhancement experiment for CC5 and iClone users—especially for anyone interested in seeing how existing characters and renders respond to newer GPU-based processing.
I am still learning how the setup works and what can reliably be recorded or exported, but the visual result was interesting enough to investigate further.
I had a look at the DLSS 5 swapper but don’t want to risk installing stuff or messing with files since I use this to work etc.
It feels safer to have installed it into obs and then turn that on with display capture and then view it through that. Here is some more examples of how it looks.
You are correct that it will not have any effect on the CC/iC render output directly. It’s a filter over the top. There is a tool someone has made for converting stills and video to DLSS5 enhanced stills and video. See here:
You can also search for it on Youtube to find videos of it being used.
Looks amazing! I really hope we get DLSS 5 as a native render option inside iClone soon, not just as a filter on top of the viewport. That would interest me a lot more than AI Studio.
@Leowald Thank you very much. I tested it out. Works really well very impressed, improved my videos a lot. Have you tried using it at all with your images or videos? I tried it out on a few videos and images did not take very long for a 6 minute video either improved the look of it
For anyone wondering how it works, this is an open-source tool that runs locally on your computer. You start it using the included .bat file, which opens the interface in a browser tab. From there, you load your image or video, choose the style and strength settings, preview the result if needed, and render it. Once it has finished, you just save the enhanced image or video.
Although the interface opens in your browser, you are not uploading the files to an online service—the processing is done locally on your computer. Below are the original iClone render and the version I processed with the DLSS 5 Visual Enhancer.
Haven’t tried it yet. I thought it looked complicated, but your examples and explanation sold me. Just not sure if my hardware can handle it (Nvidia Geforce RTX 3060)
Thanks for this great find! It’s very easy to use and works perfectly on my Nvidia GeForce RTX 3060 graphics card. I’m blown away by the quality and speed! Even with HD videos. I’ll be following the project with great enthusiasm
The irony isn’t lost on how so many people just hated DLSS 5 and how “disgustingly” abhorrent it was in manipulating the artists’ original vision…you had people sending death threats to popular tech podcasters if they so much dared praising it in any way…
@Freshness34 The problem is that it isn’t unlikely that certain game developers will release games with much less graphical fidelity because DLSS 5 will “do the rest” and “make it look good” and everyone who doesn’t have a compatible GPU or doesn’t want to use it will experience much worse graphics compared to what’s available now, if the game is even allowed to run without DLSS 5. I mean, from a financial point of view it makes sense to simply use the available technology to make graphics look good instead of creating good looking assets.
@Sophus If you have sh*t geometry, DLSS 5 is not gonna help that at all. I think developers who get the kit and discover that it does nothing for the baseline geo will, at minimum, commit to insuring that it’s presentable. On a side note, in regards to Unreal Engine 5 now becoming ubiquitous, a lot of those inexperienced dev teams that fail to embrace performance/optimization from the start is the biggest problem. You have game devs I know personally that are tacking 4K textures onto everything, 200K+ polycounts for everything - those are the biggest problems I personally have witnessed. They then rush a game out then complain about the engine.
This is indeed an outstanding tool for those who wish to continue rendering inside iClone and bypass a time-costly future RTX render option (which would sacrifice some features such as PopcornFX). Thanks to both mstaylor and Rampa for pointing to this great tool.
Here is my example. You may see how Visual Enhancer instantly brought this render to life, giving it a decent light/shadow balance, reflections, depth, occlusion… etc.
One note though. On highly dynamic scenes and specially on scenes with dynamic lighting you may get artifacts and distortions in video output, those you usually see in some AI alternated/processed content.
It depends on what you use Unreal for. As far as I know, if the developers utilize technologies like Nanite, they must use 3D models with maximum detail, for example trees with individually(!) modeled twigs, leaves and needles. If they use old techniques like textured, flat polygons which are mapped with transparent tree textures, Nanite will not work correctly and they’ll get noticeable performance problems.
Using Nanite- you must be referring to video renders, not game dev that I currently work in. It’s apples n oranges. I’d definitely do 4K everything and max detail for video/trailers, etc. For games, trust me on this - you do NOT want to do that. This is someone who’s had to develop for both console AND PC.
