"I just tested with a quick recording from 111.1 KTTV.
It looks and plays fine in Channels DVR, on the webUI as well as Channels app. Of course"
Playback in Channels is not quite perfect
I do not see closed captions and (not exactly playback, but playback-adjacent) I can't tell if commercials are being correctly detected. The comskip files all appear to be generated at the correct times (ie half an hour or so after recording finishes), and the comskip .log shows that the content is apparently being understood (somewhat...), but somehow it doesn't get properly integrated.
For example captions do not seem to be detected, likewise volume transitions are not being detected. Even logo's are not being detected (weird!), or black frames.But it's not like comskip just says "unknown codec" and gives up; every frame is apparently decoded, just not "understood".
To give a flavor of the output:
Distribution of Volume cutting: 87 positive and 84 negative, ratio is 1.0357
Confidence of Volume cutting: 84 negative without good logo is too much
Confidence of Volume cutting: 0 out of 0 are strict, too low
Black Frame List
---------------------------
Black Frame Count = 1
nr frame pts bright uniform volume cause dimcount bright type
0 114944 1917.616 0 0 0 br 0 0 B
Combining CC Blocks...
CC's detected on the following frames - 1 total blocks
--------------------------------------
0 - CC start - 0 end - 114944 type - NONE length - 0:31:57.61
Caption sums
---------------------------
Pop on captions: 0: 0.00 - 0:00:00.00
Roll up captions: 0: 0.00 - 0:00:00.00
Paint on captions: 0: 0.00 - 0:00:00.00
No captions: 114945:100.00 - 0:31:57.66
The NONE type of closed captions were determined to be the most common.
Start - 1 End - 114944 CCF - 1 CCL - 0 Type - NONE
Fuzzy scoring of the blocks
---------------------------
Block 0 has twice excess length.
Block 0 score: Before - 1.00 After - 0.00
Dictionary not processed successfully
Threshold used - 1.0500 After rounding - 1.0500
---------------------
Initial Commercial List
---------------------
No change
Logo fraction: 0.0000 Logo disabled
Maximum volume found: 7618
Average volume: 540
Sound threshold: 500
Silence threshold: 100
Minimum volume found: 0
Average frames with silence: 45
Black threshold: 19
Minimum brightness found: 28
Minimum bright pixels found: 0
Minimum dim level found: 0
Average brightness: 318
Uniformity level: 700
Average non uniformity: 161
Maximum gap between logo's: -1
Suggested logo_threshold: 0.7750
Suggested shrink_logo: 0.00
Max commercial size found: -1
Dominant aspect ratio: 1.7800
Score threshold: 1.0500
Framerate: 59.940
Average framerate: 59.941
Total commercial length: 0:00:00.00
Cut codes:
F: scene c: change
A: aspect t: cutscene
E: exceeds l: logo
L: logo v: volume
B: bright s: scene_change
C: combined a: aspect_ratio
N: nonstrict u: uniform_frame
S: strict b: black_frame
r: resolution
----------------------------------------------------
Block list after weighing
----------------------------------------------------
# sbf bs be fs fe ts te len sc scr cmb ar cut bri logo vol sil corr stdev cc
0:+ 0 0 0 1 114944 0.00s 1917.62s 1917.62s 0.00 0.05 0 1.78 7 { E br 3180 0.00 5400 450 0.000 31 NONE
114944 Frames Processed
BTW is anyone using all this fancy neural network hardware we will have in all our Apple Silicon Macs going forward to detect commercials? That seems like a natural fit, likely to work a lot better than the rigid techniques of comskip, and a nice little project for some youngster wanting to make a name for themselves and to get started with modern SW development...