I first heard about this camera during the Adelaide Film Festival domefest event earlier this year, and yesterday attended a seminar on current 2k-4k video recording technologies, including hands-on on a few of these cameras, such as the tiny SI2k, as well as the Red One.
For those that don’t know, the Red One camera is a 4000×2000 pixel video camera that records in Raw, like professional DSLR’s do for photos. It records up to 30fps at that resolution, but can record at a higher frame rate at lower resolutions (120fps at 2k from memory).
The digital workflow seems simple enough, and the resulting footage from it is absolutely breathtaking. RED’s latest 35mm Film Demo Reel was decidedly jaw dropping to say the least. I’d love some proper testing time with wide fisheye lenses for fulldome / regular cinema presentation.
My journey of seeking what’s artistically possible at the edge of current technology (at a cheap price), is always fun. While my 360 degree photos and little planets have been coming along nicely over the past year (yet to be officially posted and written about), here is something else I’ve been working on.
Inspired by Rožnov | The Timelapse, recently I’ve been experimenting with timelapse animation techniques with my still camera. Usually a timelapse is performed from a static location, using a some sort of tripod. Traditionally moving the camera results in terrible stuttering from the camera’s changing point of view, unless the tripod is on a dolly, some kind of rail, a vehicle with suspension / or has some kind of movement buffering. As tripods are not allowed in many places in Japan, and me always trying something insanely difficult, I’ve been working on a purely handheld technique.
My tests so far have used no stabilisation or weight added to the camera, but during post-processing in Virtualdub I’ve used Thalin’s Deshaker to remove any unwanted shudder. After many tests now, it comes really as no shock that using some sort of weighted mini-tripod or steadicam apparatus is truely beneficial for good results.
Using a fisheye lens, far away objects don’t move as much within the frame as with normal size lenses, so naturally, starting off with my 8mm peleng lens seems like good place to begin. For all of these I haven’t worried about final output fps, so all these are at the standard NTSC video frame rate of around 30 frames per second. I’m thinking 15 or 12 fps might be better, perhaps with some interframe temporal smoothing, but as youtube is as it currently is that’s something I can work on.
My first attempt was at Nagoya Station just after Christmas, when the lights were still up. I tried a photo every metre sideways while I walked around the lighting area, keeping the twin towers in the center of frame as much as possible, with extra shots at the beginning and end for an accelerating and decelerating ‘dolly’.
The motion was very good, but a few stuttering moments caused the tower to ’sway’. Youtube for some reason wants my video to be upside down, so apologies about that.
My second test was many months later at Suzuka circuit, watching the Super GT. Being a test this was a very very short timelapse, perhaps only 20 seconds of real time (at one photo a second). Longer is always better, but it’s all a learning process. Although the original file is in HD quality, youtube really overcompresses the video, resulting in a very blocky image. For higher quality open the video in another window and click high quality. Perhaps I’ll reupload these to another video hosting site, perhaps vimeo.
My third test was naturally over a longer period of time, this time at the Nagoya Dome. Unfortunately I ran out of memory on my SD card and didn’t have any spares on me. As these photo sequences are geared towards animation, I realise that I don’t have to shoot them at full camera resolution for HD output. Additionally, as I walked around the dome, the security guards kept asking me for my ticket to see where I was sitting, so I had to stop and tell them I was sitting ‘over there’, (where my ticket is), and they let me through. These sudden stoppages messed up the flow of the video point of view, as well as the on-pitch action. A good test nonetheless.
Next I tried using my Dad’s 50mm f/1.8 prime lens. It’s a really short clip, as the rest of the photos had too much foreground action to successfully attached for motion. This was taken in the Meitetsu express train somewhere between Ichinomiya and Nagoya, Aichi.
At Sobue Junior High School, this next one was a test for a looping video, but near the end a loose softball came into vision and I thought it interesting to follow the student picking it up. Also being very sunny, the peleng lens got a big lens flare as I didn’t have an umbrella on me. I might try this one again another time.
I went to Hiroshima on the sixth of August for the A-bomb memorial ceremony. While I was there I took a walk around the entrance of the museum/park, taking a photo with every step.
The next day on the island of Miyajima, I took two timelapse sequences on the gondola - from the top of the mountain to the first station, and from there to the bottom. In the top gondola there was a window that could be opened so I rested the camera on the gondola frame, so this one is not handheld. I should have started filming a few seconds earlier, but as the cabin was filled with people, I did the best I could being merely a tourist. Although there is a bit of jitter (which can be taken out) the raw result, I think you can agree, is very very nice.
As for the bottom half of the mountain, the gondola cabin had a window that couldn’t be opened, so I had to shoot through the glass handheld as still as possible. This is the longest timelapse I’ve done so far, a good ten minutes of holding the shutter down. I forgot I had my external shutter release in my bag. It looks really nice in full HD, but for some reason Vimeo’s only showing it in a low resolution.
The last attempt was again with the 50mm prime lens, but this time in the shinkansen (bullet train). Here you can see three minutes of real time in about seven seconds. As the clouds were rushing by very quickly, I thought it might turn out as an interesting animation.
The result was not the best, but making the motion smooth search based on excluding pixels less than 60% white made an alright result.
That concludes the current progress of my timelapse animation experimentation. More to come as time and new experiences are exposed.