Over the years of creating videos from various formats and sources, and especially from various cameras, there’s a chain of command I like to implement in the workflow to get the best quality out of all assets. Especially when working with DSLR footage (when the imaging sensor was not particular designed for shooting video), part of the process is just “cleaning” the footage to make it look better. Here is my system, particularly for Adobe Premiere Pro and After Effects (though these filters work on any other editing system).
Initially, when I worked with Final Cut Pro 7.0 (FCP) for editing, DSLR footage was not natively compatible with the program. This means that you would import footage from your camera card, and the system had trouble “reading” the footage (i.e. playing back in realtime, working with some basic filters without the need to render, etc.). This is a real headache if you have to constantly render your project every few minutes, and it really slows down both the computer and your day. Additionally, the raw files that come out of the camera aren’t really meant for post-processing, so when you start color-correcting your footage, the effects are really poorly rendered. So for these two reasons, it was necessary to transcode to a “better” format.
The ProRes codec for Quicktime video files (Quicktime is native for Macs) is about as good as you can get, especially when you’re working with assets from mid-range cameras, and which will be publishing for the web (where it will be compressed heavily to be fit for streaming). ProRes worked natively with FCP, so you get the benefit of ease in the editing system, and the flexibility of being able to color-correct in a more robust format. This comes at a cost of file sizes being 3-4x what they are out of the camera, which is acceptable, and the speed of transcoding up to ProRes isn’t too bad on a good system (though there still is a time-loss, nonetheless). The current iteration of Final Cut Pro X has fixed this problem, though there are cases when you still want to have this practice down (for example, when you have a little extra time, when there are a lot of effects needed, or when you’re dealing with assets from multiple types of cameras and sources).
Adobe Premiere Pro has been able to handle raw DSLR files for many years without a problem, and if your project doesn’t need a lot of effects, it’s nice to just work right away. Now my system is purely to import card media to the computer for editing, and create a duplicate for backup on an external drive. Once you’ve got your edit locked, and lets say the graphics/visual effects are minimal and just simply compositing, the order of steps upon rendering out the video are:
- Warp Reduction
- Reduce Noise
- Visual Effects/Cleanup
- Color Correction
- Titles & Graphics
1. Warp Reduction
In an ideal world, you don’t use this effect. At all. It’s ugly, it reduces quality and adds blur, it’s slow to work, and it it’s almost always noticeable. Try to use it as little as possible (through good photography), try to use the effect on the lowest settings, and only use it on the shots where the effect won’t be noticeable. Think of it like a cast for a broken arm: your arm will be useful, but you’re not hiding the fact that you broke your arm.
2. Reduce Noise
Reducing noise is essential when working with DSLR footage. Even when working with low ISO’s on mid-range cameras, there can still be noise, especially if you’re working with lower-end codecs. Red Giant offers Magic Bullet Denoiser II, but I found better results with Neat Video (both for $99).
Anything from laser battles to fixing a blemish on the talents’ face, if the effect shouldn’t look like it’s added in, than it should look as close to the source material as possible.
4. Color Correction
It would be wonderful if color correction wasn’t necessary, but even just trying to make shots look consistent requires a lot of work. Especially for documentary work in which you’re in a real, natural environment, lighting is hard to keep consistent from room to room. Fluorescent lights shift colors. And if you’ve set your camera picture style to be as neutral as possible, to get the full range of highlights and shadows, you’ll need to push in contrast to not make your image look like mud. And coloring is fun.
I like to use the Unsharp Mask, Radius of 1.1 and Amount between 40 and 120. You’ll need to add a “crispness” to the picture, even when the final video will be compressed and streamed through small devices. But go too far, and you’ll bring out the limitations of the file codecs.
6. Titles & Graphics
These are the non-diegetic elements in your video, like titles, which are supposed to look artificial. These should already be crisp, if you add sharpness to these, you’ll bring out the aliasing and pixelation.
The noise removal and sharpening do add time to your final render, so be wary of your limitations (like if delivery is more important than quality), but otherwise this is a good outline of how to get the best out of your video pipeline. However these tools can be great assets to have to make your video stellar.