ah, nice to see people are interested in Polly =)
welly, right now I'm doing scientific non-Polly work at Caltech, California, and I won't be back in Norway until next month. So there really haven't been advancements in Polly during the past month.. as some of you might have noticed, there are still no terrain shadows, no animation support, the engine is still not very user friendly, and I have a zillion things to add / bugs to fix.
regarding the movie: i don't really know how to capture movies, I used Fraps but the video obvious doesn't work on anyting but VLC, I'm afraid. And another thing, I only have a laptop computer - so everything is developed/tested on a crappy graphics card, so things might seem slow. I'm going to buy a new computer at home this fall, to support better-looking graphics (and faster compile time.. damn source is getting big)
But yes, the engine supports bloom (and god-rays), if you have a fast enough graphics card. And, there will be support for anim8r animated scenes, and terrain shadows, but I haven't really gotten into that yet. I also plan to add proper volumetric clouds and dynamic weather. The user is also able to define the how everything should look - the Lua-scripts that controls a "game" defines which landscape textures to use, which objects are defined as "environment" objects (such as grass, shrooms and bushes), and the terrain editor controls which parts of the landscape should generate what kind of scenery (eg, mountainous landscape generates more stones and ferns, while a meadow-landscape generates more grass and stuff. It's all defined in the scripts by the user).
All in all, the current 3d objects in polly suck because I don't know anything about 3d design, but I generate objects with tree generators and hunt for textures on the net. So if somebody bothers to invest time into creating more proper objects and textures, things will certainly look better!
anyway. the trees you guys see now are like ~1000 polys each, but there is no upper limit on objects. But larger objects (like, 20000 polygons) will slow the engine down, naturally. However, I'm about to implement a normal mapper - firstly, for ground textures, but later for objects as well - and this will enable great-looking objects with few polys (as doom 3, etc). ALso, the engine supports (defined in scripts as well) a LOD-object world - for instance, trees that are close to the camera are rendered as a high-resolution mesh, while further-away trees
are rendered as low-resolution (or billoboarded) trees, making things way fast.
Anyways, developing everything from scratch takes time - and Polly will not be complete in many years. The first public release (hopefully in october) will probably be no more than a "technology" demo, where one can create simple games without much depth. But if people are interested, and if I continue existing without having much of a social life, Polly might actually get somewhere one day...
so, thanks for all the interest!