Can I configure AO somehow to make its gradients longer, like the right one? (it's usually looks like the one on the left).I use blender, but I can also try something else.
Yeah, play with the occlusion angle / occlusion distance (or whatever it's called in Blender), you'll get more or less contrasty maps. Also, make sure you are inspecting them in linear space.
>>644646There are only attenuation (which doesn't seem to do anything) and falloff (which makes everything evenly lighter/darker)>inspecting them in linear spaceNo idea what this means, I'm a complete noob
>>644699I don't know enough about Blender to help you there. Perhaps TexTools (http://renderhjs.net/textools/blender/) can offer you better control of the bake. Look also into this: https://polycount.com/discussion/comment/2613136#Comment_2613136>>inspecting them in linear space>No idea what this means, I'm a complete noobIn an AO map, 0.0 represents fully occluded, and 1.0 is completely exposed, 0.5 being exactly in the middle. This is because the map is stored in linear space, that is, the relation between occlusion values and how they are represented as rational numbers in the image file is linear.This is important to keep in mind when you inspect an AO map (or a specular, metalness, etc. -- many follow the same method). If you have your software interpret the map as non-linear (sRGB, for example, which you may have also seen referred to as gamma 2.2), it will expect the "half occluded" value to have been stored at about 0.7 (and other values accordingly), so it brings it down the whole image to compensate. The result of this is a dark, high-contrast looking map. But this is just an appearance, as the map is actually correct -- 0.5 is 0.5, although the program is showing you what would've been a 0.2.So, in summary, this is just a heads-up -- make sure you set your program to load those maps in linear space, or you may have been seeing a wrong interpretation of the data they hold.
>>644638you should write your own AO shader :^)https://youtu.be/Q1cuuepVNoY?t=46501 hour and 17 minutes in
>>644731>I don't know enoughstopped reading right there