Thanks for your input, Lucius. It's very much appreciated. I'm not quite ready to post about lighting yet, I just wanted to clarify the screenshots in my latest post aren't from Daggerfall Scout before anyone got too excited. Still, your advice is most welcome and I'm grateful for the expert tips.

You're spot-on the simple geometry can handle extra passes without much cost. So far performance is still good with an out-of-the-box shader and all lights in place. The challenge has been finding the right effect to simulate Daggerfall's charming lights. I can't be confident about the performance impact until I reach the desired result, mainly because I am a shader novice. I'll definitely post more about it when I'm ready.

I was looking at your latest post on DFWorkShop. Great job deciphering the climate map by the way. Looking at the woods.wld map compared to the climate map - certain borders and shapes match exactly. I wonder if the climate.pak identifies textures and plants and the woods.wld identifies terrain shape and perhaps plant distributions in the world (outside of cities and other similar areas). I haven't looked into it in detail, just a guess right now.

And looking at your images I think that the cities look great! But there are a couple of things you can do to make the images look "cleaner".

*For sprites it looks like you're doing alpha testing and not doing any post processing on the original textures (other then converting from paletted to RGB[A] it looks like). I'd suggest doing two things to make the edges look cleaner and get rid of the dark outlines. 1) After generating the alpha channel (generated based on the color key value I assume), where alpha = 1 (opaque pixels) copy the color value into neighboring transparent pixels. This is "dilation" which is often used when unwrapping uvs on models into charts to help minimize seams between charts. In this case when the opaque pixels around the edges blend in the transparent colors (via bilinear/mipmapping), you'll have colors that match instead of dark borders. Now, if you're willing to sort transparencies, you can enable alpha blending and get nice softer edges. You can combine alpha testing and alpha blending so that where the alpha = 0 it doesn't write to the zbuffer, limiting sorting artifacts to only the edges which will be small except when close. Once you start blending you might notice the sprites being cutoff at the tops and/or bottoms (maybe even the sides if the sprites are really tightly contained), so you may have to make the textures slightly bigger to avoid the problem. You can use non-power of 2 textures on hardware that supports it to limit the memory cost. If I failed to explain myself well or any of my assumptions are incorrect, let me know.

*This one is simple: anisoptropic filtering for fast enough hardware. Since the view distance is much greater then the original game as well as much higher resolutions you've exchanged visual "hiss" and fog for overly blurry textures as the surfaces become more perpendicular to the view plane. This can be fixed using anisotropic filtering - it'll make the large flat areas look much better.

Sorry if you had already planned on fixing these things and just hadn't got to it yet. Just a few suggestions to make your Daggerfall scenes look very nice and clean.

No worries, mate. Good advice is always welcome. I wasn't aware of that dilation trick until you explained it (very clearly too). Just keep in mind the purpose of my journal is to show something under development, not a finished bit of software. There will be a polishing pass eventually. For now, you'll just have to deal with the rough edges . Please drop me an email sometime, I'd love to talk graphics in more detail without derailing the thread.


Certain borders do match, but the rest of the map is undefined. If you want to look into it in more detail http://www.dfworkshop.net/wp-content/uploads/2009/05/texturearchivemap.jpg screenshot was made by writing out bytes 6-7 from http://www.uesp.net/wiki/Daggerfall:WOODS.WLD_format#Large_Map_Data to a RAW image file. There may be more to it, or it could just be a relic from development. It's possible that time will suggest how this data might be used, but for now it goes into the big box of loose jigsaw pieces to be picked and worried at another time.