LOD is displeasing me much

[gernot66]

i guess that's the very last time i will post this

it's not yet possible to find good values.

as example, the eagle, to compare take the updated model from my latest build.

the pilot should disappear at LOD2, it's set rather low (lod_pixels = 30), the pilot vanishes when the ship is still relatively big. of course NOT 30 pixels, something 300 i would assume, there is no lower i can use since i reserve LOD1 for the collision mesh. that is needed, to prevent a textured collision mesh which lowers the performance drastically, on bigger ships such makes them unusable.

on the other hand i don't like to set LOD2 lower then 30 else it keeps to many divisions for a relative small object.

how i should i solve this problem?

since that is changed it displeases me every time i like to make it better, it's not possible, that makes me only angry then.

the ship is still good to see and the pilot is gone, you can see when he vanishes and it's already enough when you view it from behind. additionally there is no more possibility to set a specific LOD for a submodel.

i can't work with that, sorry. really sometimes i think i better stop making models, it's frustrating when i can't make them pleasing to my eye. i'm steady going in circles with that problem.

it was good and i never understood why it had to be changed, additionally it seems there is no interest in taking this change back, like that wouldn't count. if that didn't counts i stop modeling, it makes no more fun, it's only frustrating to see you can't get it work right.

it was really good before imo, and i usually managed it to get a proper LOD from biggest to smallest in good steps. now everything is either high or low, no more mid, no more in between is possible,.

simplified you could say you could forget the LOD, it's has getting useless. it's not possible to keep a pilot visible until it nearly vanishes for the eye, i have to keep him until lowest. neither it's possible to have less divisions for a acceptable medium distance then. so i could leave that simply away, if it appears anyway always the same.

i know for "static" models this didn't count's much except you have different models or certain parts that will appear only on higher LOD's. but all this didn't works no more or not in the way i expect it, if all happens in between the first or last "meters" of distance, if i can't tell this part has to appear then and this later or earlier or i have the advantage of LOD controlled divisions, i don't need that.

sorry if i sound angry and loud, but this really makes my angry everytime i pick up a model to fine tune it and get no result afterward. all the typical coolness of the scripted models is gone, you could take as well stupid XA models like for FFED3D. that special difference we had compared to that is (by near) gone. you wouldn't need that damned script, animations and materials could be setup in the CAD, many modelers would appreciate that (except perhaps my little self).

maybe i really give up modeling for pioneer, it's frustrating me. it's very simple no? i can't achieve something i like, it's like cooking a meal and no one comes to eat and it looks like all i've cooked so far is tasting bad now.

[gernot66]

excuse my wild behave please, i hope you understand.

fortunately every time i'm feeling so upset, perhaps i go to the shop to get some milk and bread, the radio's playing eurythmics "sweet dreams..." calms me a bit down always.

radio is funny, no?

[PinBlacniP]

The models that I see Vipers etc. I am assuming are yours? ... are excellent! For LOD ... in my ancient Flight and FPS models etc. A Rule of thumb for scaling I have found useful is. 1). Max Distance scale texture colours to 1 pxl of dominant colour ... the 1 pxl being the current max. physical resolution of expected display size. Then using a rough series to increase detail models. vis Colour only 1, 4 ,9, 25, 36, ... Perception to very rough shape is still mainly colour map. Think that you don't see much detail at distance, this is inferred by movement, colour and context. As an object approaches/is approached detail follows the same rule. ... Camera and eye work very differently! A number of LOD models, can normally be derived from the Full Detail Model. Do NOT despair, Do Not Give up, I like what I see. ... Field of view, zoom, telephoto effect are all secondary to the perception. Human sight is an emotion not a science and is a grossly personal experience. We can only hope for a pleasing effect. As a general rule I have found that above the 5 colour pxl display approach, the eye has difficulty distinguishing between more than 7 LOD ranges (or plains of change) Note that even for persons with exceptional depth perception at approx. 10M (30 feet) 3D effect is lost. Note: to range a distance a sniper for example must use a focal plane device to read the distance. (in context say ... is size X at distance Y the brain will provide an illusion of perspective.) Judging the 7 LOD changes, is dependent on the grossness of the model. A rivet is nothing till about 1M (39 ins.) where as an oil tanker or Panther in good light with NO fog would still show detail at several Km. A useful exercise for this is percentage of FOV (field of view) or COV (Cone of View), mapped to max. expected res.
Don't Despair It's good work and mostly cool! (in my view anyway)

[robn]

Duplicate of #72.