(thunderstorm passed with not a bang but a whimper)
Well, you make me optimistic; I thought this type of production is hard or impossible to sell. If there is a market for such videos, then I'm really interested and start looking for investors
that basically puts us in the same market as the biocursion dvd; i was thinking earlier if their revenue is enough to support such an endeavour, strings could be pulled and some hd-dvd / blu-ray fractal animation discs could be produced. i'm sure
someone will be interested in bundling such a disc with such a hd video player to show off its quality potential, and if it's done right the thing could be turned into a nicely cyclic/randomised display for parties or so - kinda like scott draves' "dreams in high fidelity" project (which isn't done at 1080p hd res iirc).
going further, i've been thinking a lot about achieving
realtime rendering of these 3d fractals using the latest generation of gpus (
http://techreport.com/reviews/2006q4/geforce-8800/index.x?pg=1 - mine should be arriving tomorrow); they have a great mix of general purpose and graphics-specific computational power that makes them ideal for the task. what's more, they are on an incredibly fast computational growth curve ("moore's law cubed" as nvidia coins it) and several gpus can be used in parallel. i'm doubtful that full hd resolution 3d fractals of the sort i showed previously can be done in realtime on a single g80, but perhaps with two of the next generation ones (which will be integrated into the cpu most likely, thanks to amd/ati teaming up) will be enough.
the remaining problems are only that of engineering a ridiculously efficient rendering scheme* that can be mapped to gpus, finding good music and convincing a sponsor that the 4 g80 cards are
not, in fact, for gaming!
* halfway there on that one, just need to map it to the gpu in a way that doesn't lean on the cpu so much. using a single 2.5ghz thread of execution that image i linked, which is 542 megapixels big at its full size (29310x18510 resolution), took just under a day to compute - and that's going
way over the quality requirements for an animation where 5x5 supersampling already looks completely awesome. this is somewhat complicated by the need for
temporal antialiasing (motion blur), which will increase the antialiasing quality but also take much longer - i'd say a day per frame per 2.5ghz core with outrageously high quality rendering at hd resolution isn't so bad though, and bodes well for a gpu implementation. those new intel chips (c2d) are a lot faster than my athlon x2 in several key areas and also support 64bit, which is totally necessary for my app; i can still do low level sse optimisation to get a 2-3x performance increase, also. shit now i've typed way too much again...