I've been working on getting dynamic flow maps working in the UDK. Just finished getting a huge post on it over on my blog (to big to copy over) but here is the link for those interested!
Looking at this picture, the waves around the moving object look rather unnatural.
Some ideas for improving the effect and the performance:
If you had some temporal accumulation (combining and building on several of the previous frames from the depth camera) combined with generating some normals/waves for the moving objects (like cryengine does for when the player walks through water) the effect might look a bit better in the case of fast moving objects. You migt also be able to make it cheaper by amortizing the cost of rendering the heightmap across multiple frames, although I don't know if this kind of stuff is possible/exposed in UDK.
Try sampling a lower mipmap or rendering a lower resolution depth map. I've found that flow maps can often look more natural with lower resolutions. The change in directions will be less sharp and should help making it look less like lava. But it would be pretty cool to see this effect with a lava shader
Looking at this picture, the waves around the moving object look rather unnatural.
Some ideas for improving the effect and the performance:
If you had some temporal accumulation (combining and building on several of the previous frames from the depth camera) combined with generating some normals/waves for the moving objects (like cryengine does for when the player walks through water) the effect might look a bit better in the case of fast moving objects. You migt also be able to make it cheaper by amortizing the cost of rendering the heightmap across multiple frames, although I don't know if this kind of stuff is possible/exposed in UDK.
Unfortunately almost none of that stuff is exposed at all in the UDK, though the idea had crossed my mind. I should add on the visual end the weird delay comes from the fact that I recorded this in editor (where SceneCaptureActor2D's are super delayed). When running in a proper build, the delay is gone completely and helps in many ways.
Try sampling a lower mipmap or rendering a lower resolution depth map. I've found that flow maps can often look more natural with lower resolutions. The change in directions will be less sharp and should help making it look less like lava. But it would be pretty cool to see this effect with a lava shader
The current resolution of that height map is 256x256 for a relatively large area. I found that offsetting the values before the blur gave "larger" influences than lowering the resolution. I should also add that what was recorded in the gifs, was recorded for visibility, versus quality (I had a hard time showing what was going on). This is why the water doesn't look at all like water, but more like goop
Changing the flow intensity, flow speed, and other wateresque properties (reflections, specular, wave normals) is a different matter all together
Replies
Looking at this picture, the waves around the moving object look rather unnatural.
Some ideas for improving the effect and the performance:
If you had some temporal accumulation (combining and building on several of the previous frames from the depth camera) combined with generating some normals/waves for the moving objects (like cryengine does for when the player walks through water) the effect might look a bit better in the case of fast moving objects. You migt also be able to make it cheaper by amortizing the cost of rendering the heightmap across multiple frames, although I don't know if this kind of stuff is possible/exposed in UDK.
Unfortunately almost none of that stuff is exposed at all in the UDK, though the idea had crossed my mind. I should add on the visual end the weird delay comes from the fact that I recorded this in editor (where SceneCaptureActor2D's are super delayed). When running in a proper build, the delay is gone completely and helps in many ways.
Thanks for re-uploading the gifs!
The current resolution of that height map is 256x256 for a relatively large area. I found that offsetting the values before the blur gave "larger" influences than lowering the resolution. I should also add that what was recorded in the gifs, was recorded for visibility, versus quality (I had a hard time showing what was going on). This is why the water doesn't look at all like water, but more like goop
Changing the flow intensity, flow speed, and other wateresque properties (reflections, specular, wave normals) is a different matter all together
Thanks for the feedback though!
Alex