today’s news:
i recently merged the video
branch. it adds support for various real time video grading applications. i discovered that my old 5d mark II can in fact record 10, 12, and 14-bit raw video through magic lantern, so i integrated support for this. new key features:
input:
- real time webcam input (v4l2) (maybe i should also add a v4l2-loopback output device, so i could grade my webcam video for lockdown video calls.)
- load magic lantern raw video files (.mlv) with audio
processing/gui:
- press space bar to start/stop video in darkroom mode, parameter changes apply immediately
- use real time clock to time the av (i’m using 1:2.35 24fps obviously). this mostly means waiting after processing finished until it’s time to start with the next frame.
- play audio that comes with mlv video (alsa, need to make this optional)
- implemented raster/drawn masks with varying radius, opacity, and hardness (need to provide screenshots/streamline it a bit)
- improved support for graph blocks, i.e. sub-graphs that can be added at once, as a sort of template (the drawn masking + guided filter consists of a few modules and is best added this way for instance)
output:
- added the
o-ffmpeg
module that is wired to thecli
to quickly convert/export graded mlv to mp4 (https://jo.dreggn.org/vkdt/src/pipe/modules/o-ffmpeg/readme.html)
some screenshots as proof:
lighttable mode with a bunch of mlv raw videos. you can sort, filter, label, and tag them just like regular raw files:
darkroom mode showing the node graph and the i-mlv
module. all modules work as for regular raw files but are applied per frame. in particular i hope to be reusing the noise profiles for this camera. i am doing this and it works, but because of the pixel binning i should actually do a correction step i think.
darkroom mode showing the frame counter and a random different frame from the video (will work on a better timeline ui soon):
hope this is useful for some video geeks if anyone has higher res/more heavy weight .mlv raw video samples to battle test my pipeline i’d be very interested in trying your files/hearing your experiences in case you try yourself.
some features don’t work yet with videos (like copy/paste the processing graph), but maybe once this is ironed out a bit i’m thinking i might do a first release.