Article URL: https://nothings.org/gamedev/ssao/ Comments URL: https://news.ycombinator.com/item?id=48979931 Points: 99 # Comments: 32

Please, feel free to gather better data yourself and prove me wrong. If you do so I'll link to it! Game developers are in love with screenspace ambient occlusion (SSAO). They're so in love with it that they do all sorts of crazy things. Amateur developers post screenshots where convex corners get brighter near the corner. Professional developers release games with bizarre halos and ugly artifacts (look at the floor behind the railing). Even when they avoid those flaws, games with SSAO often still look wrong. And an easy thing to pull out from it: corners of rooms look weirdly dark. I believe this is because people think the corners of rooms actually are dark, and they've allowed this belief to let them write code that produces unrealistic results. More specifically, I attribute the failure of SSAO in these cases to some combination of four factors: I am not an active graphics researcher, so I'm not going to set up a framework and show where these things went wrong. That's the job of people publishing SSAO papers and shipping SSAO in games. For the same reason, the following discussion is not the most scientific nor at all thorough. Perhaps collecting more data would suggest I am wrong. But my personal visual experience doesn't match what people are rendering with SSAO, so I wanted to share that experience; to which end I collected some data. Photographs were taken using a Canon 5D Mk II, exported as JPEGs, using exposures ranging from a 1/80 to 1/8 a second. The original large JPEGs were downsampled by a factor of about four in each dimension using Adobe Photoshop. All images had their chroma channels blurred with a 16-wide gaussian in Photoshop (before downsampling) to reduce chroma noise artifacts in the long exposure images.