Last major work I did for OoT UHD was work on low-poly assets and inject them into the rom with some community-built utilities. I've stress-tested the engine to see how it and the different graphics plugins respond to alpha textures and higher polycounts. The prospect is actually rather good, and I have some methods in mind for building unique areas in the game. Here is a screen of one of the variants of the 'Hilltest' stage model where I'm testing and building assets. These models are running natively in the engine with original textures in vanilla N64 format, they are not UHD assets and this is a test stage intended exclusively to test the limitations of the engine:
Things to note are that I do not necessarily intend to actually put a high quantity of dense, tall grass into Ocarina as much as I might like to, the engine simply is not made to handle such amounts of alpha transparency. Rather, terrain and foliage were the most obvious likely stressors I will have to deal with. In most instances, the grass effect at this height is actually quite jarring and tends to clip through at/around child Link's shoulder and face.
The engine itself has displayed some interesting characteristics in the degree of fidelity it can track vertices. All objects imported as a world model were restricted to a certain amount of decimal places, and anything finer in that is rounded to the next number, so smaller details such as the tree branches had a tendency to occasionally distort in ways that were more obvious ways. Overall, I'm satisfied with the way the engine performed and I believe I can build relatively detailed environments. Also, the general world model (including hills in the background) may appear modestly sized due to the angle and placement of the camera, as well as that of the textures which I only assigned lazily, but the world model itself is scaled to something like 2 or more times the size of Hyrule Field and the time it would take to walk across it reflects that.
Trees and other such objects in the vanilla game are handled as actor objects rather than part of the world model, and that is both likely because they can then be culled at different times from the rest of the world model while calling in repeating data for the purpose of compressing the model's data consumption, but also so they are not as limited by the vertex locking that I've seen when dealing with the world model.
(actually an older screen than the previous)
Noteworthy: The obelisk bricks are rendered with 2 polygon layers per side, and they do not repeat anywhere around the model. The base of it is surrounded by large candles, all of which have a unique appearance and shape but all of the candles use the same texture graphic mapped to different positions. Despite this, my approach to lighting has remained fairly convincing (even at low-res) and I'm pleased by this. The textures, like most others I've shown, were hand-drawn.