Some video experiments

I've been working on running videos for a long time without any success. The double problem of the "bounce" and the extreme duration is tough to beat. There are stabilisation filters, but to get good results you need better cameras, and I suspect a better computer than I have. If you look at my Youtube videos you will see the Sydney ferry - auto-stabilised by Google - and the side effect of the pulsating bridge. So the point of view has stabilised, but it's only useful really if I crop out all the artefacts. Sometimes that will be OK, but running videos are usually going to be focussed on the distance/landscape.

Shortening the videos is achieved by editing; life is too short for that to be a practical solution. So ultimately I've decided on the tactic explored here, which is the auto-edit, multiple still approach.

With the canoe, the camera was shooting at 30 fps, which means an effective exposure of 1/30th sec. Because the canoe is slow, < 6 kph most of the time with me in it, and smooth-moving, the stills are reasonable quality. However, the auto-edit (1 frame per 7 seconds) means that the flash of an interesting bird is only caught by accident. Overall, the clip does give an idea of the creek though, so I'm pretty happy with it.

The first running video is my first real essay with this new technology. It has a lot of problems, but it's a start. I attached the camera to my head (not very comfortable after the 1st 90 minutes) and I forgot to orient it correctly. The raw footage has the sky rotated 90 degrees. Fortunately that can be more-or-less remedied in the editing suite, but the cost was the side black bars. Next, I can set the camera to record at 60 fps, which helps to reduce the blurring from the unstable gait, but at a cost to the image quality because the camera captures half as many pixels per frame to do it. Also, the aperture control is automatic, and slow to react to changes in shadow, so a lot of the frames are dark. At this stage I haven't worked out how to automatically discard them. But for all that, there are some nice images here - the sky/clouds towards the end, and there's a few more tricks I haven't tried yet. The great thing is - it takes 10 minutes to create this video from a 1 hour run, and it would take 10 minutes from a 2 hour run as well, which makes it at least feasible to video document the runs. some of the trails are well worth the trouble!

Here we have the result of a hand held camera. I think it goes for about 10 minutes, and I defy anyone to watch without nausea. This is what running would look like if your brain didn't have absolutely phenomenal stabilisation processing. I think I can make a case for the "art" of the artlessness of this, but not sure there's a career in it for me.

Finally, below, one more video recorded by accident while testing the new battery, and some new software that ended up as the basis for the two videos above. Nothing flash, but pleasing, nonetheless. (It's only 3 minutes, and yes, something does happen)