To begin, I shot this as a series of stills, JPEG, 21 Megapixel, fine resolution @ 1 shot ever 2 seconds. I did not shoot in RAW because the resulting files take much longer to process than the JPEG files do. If I was shooting for National Geographic I would – but I’m not.

Click here to view the embedded video.

My goal was to take a nice video of the sun set. As I was considering wrapping up shooting I noticed that the moon was rising over campus and was entering the frame. Unfortunately, I was shooting with a polarizing filter and that significantly cut down on the incoming light. I couldn’t remove the filter between shots because if the camera moved or the lens settings changed in any way the final video would have a jump in the final product. Of course, shooting at night, at f/8 with a polarizing filter causes other problems. The exposure time went from 1/800 in the daylight to 1.5S a dark. To compensate, I had to stop the time lapse at two points and bump up the ISO from 800 to 2000 to 3200. This added noise to the scene that I will address later.

Once I was done, I had nearly 20,000MB of photographs to work with. I imported them in to a new Aperture library and played with the settings until I achieved a look that I liked. I cropped the frame to 16×9 and applied the adjustments to all of the images. From there, I exported the frames as a series of images @2,000 pixels wide. 1080P HD video is 1920 pixels wide so this gave me a slight buffer when working with the sequence in After Effects.

In AE CS3, I imported the sequence as a new clip and setup a new composition for the video. From there, I used motion stabilization to analyze and remove most of the jitter caused by the wind. I selected two targets on opposite sides of the frame to calculate rotation and let AE do the rest. Next, I applied a noise reduction filter and I used a non-default temporal setting (5 frames, I believe) to help identify the noise in the higher-ISO sections. One of the benefits of shooting @ 21 Megapixels is that when you scale down the image, you are effectively oversampling the image and reducing the signal to noise ratio. The resulting frames were still pretty grainy so that is why I elected to use the noise reduction filter.

Next up, I enabled frame blending and remapped the time to 50%. The resulting video was too long to be enjoyable so I had to speed it up. This also had the added benefit of removing some of the flicker caused by the camera.

When you are filming a time lapse, you have a critical choice to make: do you manually set the exposure or do you let the camera? If you manually set the exposure, the resulting video will not flicker. This is great for clips where the lighting does not change dramatically. Unfortunately, in a sunset clip like this, you pretty much have to let the camera meter each and every frame. Each photo is slightly different: trees are moving, lights are reflecting off buildings and people are walking throughout the scene. This causes the camera to meter some frames brighter than others and the result is a video that flickers.

There are plugins to remove time-lapse flicker but I have not experimented with them too much. Apparently the color stabilizer filter can help with some of this but I’ve had limited success. Thankfully, speeding up the clip with blending helps to a degree.

Lastly, the resulting composition was exported as a high bit rate video and sent to Vimeo.

Click here to view the embedded video.

The net effect: this process takes a lot longer than dealing with JPEG images.  @21 Megapixels, there is plenty of data to play with, especially if the final video will be 1080P or smaller.  However, processing the images takes, on my machine, nearly 5 times as long.  If I was producing footage to be used in an IMAX theater I would still use this process, but for every conceivable circumstance, I can’t see any practical benefits to using raw files over JPEG.

Feel free to tell me I’m wrong in the comments

Click here to view the embedded video.

I’ve been experimenting with ways to take the best time lapse using the Canon 5d Mark II and I’ve come up with two distinct methods that are worth sharing.  The first method involves shooting normal video and speeding it up while the second involves a more traditional approach of shooting frames and combining them to a final video.  Each method has its benefits and I’ve tried to summarize them below.

In general, I prefer to shoot with an intervalometer and combine the frames in posts.  1080p video is about 2 megapixel.  My camera can shoot 21 megapixel native and this extra resolution allows for pans and zooms around a moving frame that you really can’t do when shooting at 1080p. A time lapse alone is interesting but I really like adding additional motion in post.

My workflow consists of shooting the image as 21mp JPEGs* and processing them against a common preset in Aperture 3.  The resulting images are exported and imported in to either Premiere or After Effects as an image sequence.  From there, the final pan/zoom effects are added.  The extra resolution of the 21mp images allows for subtle movement without causing the blur normally associated with pan/tilt/zooming video.

* I prefer JPEGs when shooting a time lapse because the resulting images process faster and take up less space.  When the images are scaled down to 1/10 the original number of pixels, the added benefits of RAW become negligible, in my opinion.  If I were shooting a video for National Geographic I would use RAW…but I’m not.

Method one: Use the HD Video recording  feature and speed up the clip in post processing to the desired speed.

Pros:

• It is easy to do
• you can see the images as they are being taken
• Produces fluid motion
• Allow for multiple output speeds

Cons:

• Resulting video size is limited to 4GB ~15 minutes
• Final results dependent on the software you are using
• Limited to 1080P video

Method Two: Use an intervalometer to shoot individual frames

Pros:

• The final images are much larger allowing for more manipulation in post (such as pan, zoom, etc.)
• It is easier to achieve the desired video look – you can batch process videos in Photoshop, Aperture, etc.
• There is no flicker in the final video output if manual mode is used

Cons:

• You need to reset your counter if you will be crossing the 10k image mark (on Canons, at least..)
• If you shoot RAW, you will need to process all of the images
• You need to carry either a laptop or an intervalometer to time the shots

Video Notes:

Scene 1: Method 2

Scene 2: Method 2

Scene 3: Method 1

Scene 4: Method 1

Click here to view the embedded video.

Lessons learned:

Even a small breeze can make shooting a time lapse a headache.  The jitter in the frame was caused from the wind.  I tried using After Effects to stabilize the motion but the nature of the frame make this difficult.  I’ve successfully used this method on bright daylight videos but I couldn’t get a good tracking on most of this footage.

Second, all of the frames were 21 megapixel JPEG images.  The average file size was about 7MB per frame which meant that After Effects did not like processing them.  It worked – but slowly.  To speed up the process, I used QuickTime Pro to import the frames, scale them down to 3000 x 2000 pixels and compress it as an Apple Intermediate Codec file.  This seemed to speed up the process quite a bit.

Lastly, use manual mode.  The camera will automatically adjust exposure for each frame which will cause a subtle but noticeable flicker in the image.  What I did to resolve this was to take a few test shots to measure the exposure.  Once I found a set of values I liked, I switched to manual mode and started the capture.