I recently bought a drone for photography. But the video looks so good that I end up taking a lot of video, too. Video is new to me, so the raw files keep piling up in my "in" tray. I will have to learn to edit video, one day, or send them to a pro to edit for me.
I decided that the least I could do is watch the videos, organize them, and trim them to just the interesting parts. Saves on disk space.
I wrote a bash script to help me:
- Loop over all videos in a dir.
- Play each video.
- Extract the clips I want.
- Tag, rate and organize the clips.
The script opens each video in two MPV players. One is full screen and unscaled (watching 4k on a 2k screen means the video appears zoomed in to a 2k region). This is for pixel peaking. I can quickly check, at a glance, the raw video quality. Another MPV window acts like a PIP, taking up a quarter of the screen, and showing the whole video scaled down.
If a video is DLOG, a LUT is applied to MPV to show the video in a more natural colour (raw LOG video looks grey before it is processed).
Hacking this together, without a plan, I use simple msg boxes, on top of the playing videos, to control the process. Better than having to flick back and forth to a terminal window.
When I see a good place in the video to start my cut, I press the "Start" button. An input box pops up, prefilled with the current time of the player, e.g. 00:00:09 if the video is 9 seconds in.
I watch some more of the video and notice some messy, jerky camera movement starting at 38s. I press the "End" button, and another input box pops up to capture the end time of the clip. I change it to 00:00:37 to exclude that jerky part.
Now, in the background, ffmpeg is called to extract the section of video between 9s and 37s. I use keyframes so that video does not need to be re-encoded. It sets the real start time to the nearest keyframe before the start, and the real end time to the next keyframe after the input end_time. This means the output video is always a bit longer than I chose. I can trim those few extra frames when I use the clip. Because we don't re-encode, the extraction time is near instantaneous.
A preview of the clipped file is played back at high-speed.
If the source video is long, and contains more content I want, I continue playing until I see the next clip I want to extract.
When I finish with a source file, I am asked to give a star rating (1-5) for the videos and then to choose tags. For these I make use of the rating and file tagging extended metadata (xdg). I can select any number of pre-existing tags, and add new tags. Some metadata tags will be added automatically, such as frame_rate, resolution, and colour_profile.
Now the clips are in the output dir, and I choose to send the original file to the wastebin. The next video in the source dir starts playing, and the process continues until the dir is empty.
The result is a neat pile of trimmed and catalogued video clips. Ready to be thrown into some YouTube video.
Only problem, now? I'm more interested in refining the bash script, than I am in learning to use Resolve to make a finished video.