PeteGould
lvl.4
United States
Offline
|
JakeLikesStuff Posted at 2015-5-19 12:31
is there a way to easily and quickly trip Inspire vids before going in to Final Cut... my concern is that importaning multiple 10 clips creates a lot of render time before the files are ready to edit in Final Cut Pro.
There are several issues you can encounter when working with this kind of footage. In some ways it introduces complications we don't necessarily see in very high end (feature film, episodic television) products, which can be a lot easier to work in but fabulously more expensive. What I'm about to describe is not unique to the Inspire - it's true across all products at this level.
The most annoying is data compression. The Inspire saves videos using the H.264 compression algorithm and this is true regardless of whether you save to .mov or .mp4. Those formats are only the wrapper containing the file. The file data itself is H.264. And H.264 is simply a means of reducing the size of the file while retaining as much of the visual content as possible.
Let's talk about the resolution you're shooting in: 1080 progressive. First, consider the sheer size of the raster we're talking about - 1920 pixels wide by 1080 pixels high. That's 2,073,600 pixels. And since each pixel can be any of 16 million possible colors, each pixel requires three bytes of data to represent it - a byte for red, a byte for blue, a byte for green. So if you multiply 2,073,600 pixels by three bytes, you end up with 6,220,800, or roughly six megabytes. Per frame. At 60 frames per second. Clearly, something has to give here, because that's about 360 megabytes per second. Imagine the cost of the storage. Heck, imagine the cost of the electronics that can move and store that much data in real time. Then think of 4K which roughly quadruples that. Ouch.
Enter compression schemes. You know them from photography because if you save something as a .jpg file, the storage requirement is a tiny fraction of what it would be if saved to an uncompressed .tif, .tga or .bmp file. The idea is to throw away A LOT of data and come up with some way of ending up with an image that looks like the original (but is missing a lot of its actual data).
The problem with compression schemes is that the ones we're discussing here are lossy. In other words, when you open and read the file, you don't get the whole content back the way you do when you open a .zip or .bin file. You get a degraded representation of the image that is "good enough to look okay." If you save in a compressed format, then read the file and do something with it and save it again, it doesn't just save the original data. It recompresses that data. And often, the small visual artifacts created from lossy compression get mistaken by the compression engine, the second time around, for image content, so they get exaggerated when the file is saved again. Do this for a few iterations with a lot of compression added and the file will look really awful.
And video compression, as distinguished from still compression, adds a temporal element. H.264, like a lot of lossy video compressors, saves a "key" frame at fixed intervals, and then the subsequent frames only save the visual differences between that frame and the immediately preceding frame. That continues until the next "key" frame.
Now, there are degrees of compression. You can tell an H.264 compressor how big a file you're willing to live with in exchange for greater quality. If you compress too little you get a huge file. If you compress too MUCH, too much degradation occurs before the next "key" frame, so when the video player hits that frame you suddenly see a slight visual "hit." What you're seeing is more and more compression artifacts collecting over the course of a second or two (whatever the selected distance between key frames happens to be) and then suddenly a much cleaner frame, followed by the increasing collection of artifacts again, over and over, in a continuing rhythm. In the case of the Inspire, the folks at DJI have set the compressor so it's clobbering the daylights out of the video, and they've provided no user access to the compressor settings to allow us to back off on the compression ratio. This could be a choice on their part not to give consumers access to advanced controls, or it may be a hardware limitation: the compression settings COULD be set to allow the maximum data throughput the hardware is capable of. If so, that would be a real shame, because these files are currently way, WAY too compressed.
That's why some people are complaining about a rhythmic disturbance in their videos especially when shooting fine detail, like a field of grass, or a forest of pine trees. What they're seeing is the degradation over the course of the frames following a key frame, and then - SNAP - a new key frame, and then degradation again. If you're not accustomed to seeing it you may not initially notice. Once you're looking for it, it's obvious.
This is my longwinded lead-in to why I don't pre-edit Inspire clips. Because typically when you do that, the saved edited clips have been recompressed - and there's too much compression to begin with. I shoot judiciously (or at least I try to) and save the entire raw clip, which is what I use as source. You can elect to pre-edit and just save "buy" clips, of course. Usually when people do that, though, they save to a much less compressed target format in order to minimize recompression artifacts. But then you've lost the storage benefit of pre-editing, because a less compressed format can be ten times the H.264 format per running second of media. Bornish, in a post above this one, mentions tools that are intended NOT to recompress, but simply to extract and separately save chunks of video with the original compression intact. I haven't played with any of those tools. If there are confirmed tools that can extract and save H.264 video data while guaranteeing not to alter that data, that's a good solution.
NONE OF THE ABOVE, however, relates to what I'm seeing in your video, which is that the image has been scaled down and then back up again. I can only think of two places where that might be happening. One is in your Quicktime pre-edit pipeline. If it's getting 1080p/60 as an input, you need to make absolutely sure it's giving you 1080p/60 back out and not some other frame size due to a wrong or corrupted preset. The second place is in your FCP project. If the project is DV NTSC widescreen, then the imported 1080p media would have to be shrunk to fit the frame - if the output is then blown back up to 1080p when exported, you'd get what we're seeing in this video. There may be other ways this could be happening as well, but those two come immediately to mind. One thing is sure: the Inspire video output definitely looks better than that. |
|