A deep-zoom to 1.9e-78 showcasing a new frame interpolation system.
This is a very old project I started exploring before the quad-core system was in use, and before frame interpolation was in the software. Back then, it was impossible to render this video in a reasonable amount of time--it would have taken about 8 months to a year. With the current hardware and software, it takes only about 9 days!
|MP4 Files (QuickTime player)|
|Fast Download||360x240 300 Kbps 19 MB fast start|
|Fair Quality||720x480 1 Mbps 47 MB fast start|
|DVD Quality||720x480 4 Mbps 167 MB|
|Hi-Def Quality||720x480 8 Mbps 331 MB|
|WMV Files (Windows Media Player)|
|Fast Download||360x240 300 Kbps 19 MB|
For the first time on this site, four video quality options are offered for an animation. The Fast Download files are small, highly compressed, and should be able to play as they are downloaded on most high-speed internet links with minimal interruption. The Hi-Def quality files are compressed at high-definition video bit rates (well, the low end of HD anyway--see below). Two intermediate options are also available.
|Date Generated:||10-Oct to 7-Nov 2009|
|Final Image Size:||1.9e-78|
|Video Length:||5:00 of fractal, 5:25 overall|
|Rendering Time:||219.2 hours|
|Audio:||Custom with Acid Pro 7|
This is the first animation since CanyonDeepTrial1 that uses frame interpolation, which is explained in detail in the Technical section. Essentially, this is a method for generating video frames more efficiently by calculating a set of master images that are much larger than the video output size. Digital imaging techniques are then used to zoom in to the master images to make the video frames. Although the master images are much larger than the video frames, there are far fewer of them, and an optimal size and number can be found that minimizes the total calculation time needed to make an animation.
This project is the debut of a major overhaul of the frame interpolation system in my software. A major problem has been fixed, and some significant improvements have been added. The problem had to do with lateral motion while zooming; the previous system would not determine the correct master image region to use. The improvements mostly relate to how the optimal master frame size and number of master frames is determined. This version of the program can now pan and zoom, and finds the optimal set of master frames for a video.
Like many projects on this site, GroupB uses music created from sound loops that are carefully merged together to create a final piece. This one uses some bass lines with very deep, low frequencies, and I have to admit it has pushed the limits of my rather modest audio hardware. I experimented with different mixings and tried to make it sound reasonably decent on inexpensive computer speakers without being too booming on audio equipment with better bass response. It might be time to upgrade to some better speakers.
Since the frame interpolation technology overlaps segments of video derived from different master images, I tried to capture that thematic element in the music, having different voices fade in and out as the music progresses. Near the end, the video gets kind of monotonous, and the music parallels this as well.
In terms of motion, this animation is very simple, with no acceleration at the beginning, no deceleration at the end, and no pausing, panning, or any other motion effects. It just starts abruptly, zooms in steadily, and stops abruptly. I tried to make the music incorporate that thematic element as well, with an abrupt stop and change of direction at about 2:50, and a rather abrupt end.
The large, 300MB files here were encoded at about 8 Mbps, and the resolution is 720x480. Technically, these are not really high-definition video since the resolution is only 480 vertically. I chose 720 horizontally not because it matches the DV format (which has rectangular pixels, by the way, not square) but rather because it gives an aspect ratio of 3:2, which is a little bit wider view than 640x480 (4:3 aspect ratio) but not quite as radically wide as a 16:9 aspect ratio.
The bit rate is "only" 8 Mbps, which is about half what a true high-definition format bit rate ought to be, but the number of pixels in this animation is also about half what a real 1280x1080p HD video would be, so this seems reasonable. Given the video frame size (which is determined by how much time I am willing to spend calculating video frames for the project) 8 Mbps delivers enough video information that the pixel resolution is the limiting factor in the quality of the product, not the bit rate. Sort of. A fair amount of color artifact can be seen even in the 300 MB files, but the level of detail is close to what the uncompressed raw video files look like.