Merge timespan in between scenes list

[gphg]

I have this in development bash script that take arguments.

__input="$1" # Required
__phase="$2" # Aka controller

__video_file=$(basename -- "$__input")
__video_name="${__video_file%.*}"
__video_ext="${__video_file##*.}"
__csv_name="${__video_name}-Scenes.csv"

# (1)
# Detect content and write it to be revised later.
# Saving into a file is important for consistency!
if [ ! -f "$__csv_name" ]; then
  echo "$__csv_name doesn't exist."
  scenedetect -i "$__input" \
  detect-content \
  list-scenes -f "$__csv_name" --skip-cuts
fi

# (2)
# Split everything as it is for human revision.
# Do it manually: remove lines to be merged, eg it was too short.
# To speed up, eg it will be split again later, you can lower the quality.
if [ -f "$__csv_name" ]; then
  echo "$__csv_name exist, process video spliting."
  scenedetect -i "$__input" \
  load-scenes -i "$__csv_name" \
  split-video -crf 32 -p ultrafast
fi

UPDATE: Fixed error on file loading. Any further updates go to this gist.

Step (1) produces list of scenes in CSV.
Step (2) process the CSV from source into several scene videos.

After the script is executed, list scene file produced and scene videos. This is the untouched CSV:
input-Scenes.csv

My idea: I want to combine several scenes into one from the saved list of scenes

Normally, I merge the processed videos by scene ranges into one video using ffmpeg. It is too much brain-power and the video is lossy because reencoding.

My expectation is,

1. From saved list-scenes, I removed, manually lines that need to be merged:
2. Removed lines simply merge to previous one until it hit next one.
3. However, at the moment, on current version, it skips. It produces same results, unlisted scenes are skipped.

This is the manually selected scenes:
input-stagged-Scenes.csv

Scene Number,Start Frame,Start Timecode,Start Time (seconds),End Frame,End Timecode,End Time (seconds),Length (frames),Length (timecode),Length (seconds)
1,1,00:00:00.000,0.000,202,00:00:03.367,3.367,202,00:00:03.367,3.367
4,1710,00:00:28.483,28.483,1915,00:00:31.917,31.917,206,00:00:03.433,3.433
12,4335,00:01:12.233,72.233,4623,00:01:17.050,77.050,289,00:00:04.817,4.817
17,6357,00:01:45.933,105.933,7151,00:01:59.183,119.183,795,00:00:13.250,13.250
23,8684,00:02:24.717,144.717,8993,00:02:29.883,149.883,310,00:00:05.167,5.167
31,10862,00:03:01.017,181.017,11095,00:03:04.917,184.917,234,00:00:03.900,3.900

Above is expecting to produces six videos:

1. Scene 1 to Scene 3 are merged into one video
2. Scene 4 to Scene 11 are merged into one video
   and so on.

[Breakthrough]

However, at the moment, on current version, it skips. It produces same results, unlisted scenes are skipped.

What do you mean by skipped? Could you provide the output of list-scenes -n after you load your modified CSV?

The load-scenes command works only on cuts based on the start time column by default, so you should already be getting the expected results based on the CSV you edited.

---

Testing locally, as a concrete example, if I take the following CSV and remove scenes 2-3:

Timecode List:,00:00:03.754,00:00:08.759,00:00:10.802,00:00:15.599
Scene Number,Start Frame,Start Timecode,Start Time (seconds),End Frame,End Timecode,End Time (seconds),Length (frames),Length (timecode),Length (seconds)
1,1,00:00:00.000,0.000,90,00:00:03.754,3.754,90,00:00:03.754,3.754
2,91,00:00:03.754,3.754,210,00:00:08.759,8.759,120,00:00:05.005,5.005
3,211,00:00:08.759,8.759,259,00:00:10.802,10.802,49,00:00:02.044,2.044
4,260,00:00:10.802,10.802,374,00:00:15.599,15.599,115,00:00:04.796,4.796
5,375,00:00:15.599,15.599,480,00:00:20.020,20.020,106,00:00:04.421,4.421

After removing scenes 2 and 3, I see the correct scenes - scenes 1-3 have been merged into scene 1:

-----------------------------------------------------------------------
 | Scene # | Start Frame |  Start Time  |  End Frame  |   End Time   |
-----------------------------------------------------------------------
 |      1  |           1 | 00:00:00.000 |         259 | 00:00:10.802 |
 |      2  |         260 | 00:00:10.802 |         374 | 00:00:15.599 |
 |      3  |         375 | 00:00:15.599 |         480 | 00:00:20.020 |
-----------------------------------------------------------------------

This seems to match what you expect, so some more information would be helpful.

[gphg]

Another question: in case I want to make a bit of change, eg a specific scene is too early, which column should I edit?

[Breakthrough]

Another question: in case I want to make a bit of change, eg a specific scene is too early, which column should I edit?

By default, the load-scenes command uses the "Start Frame" column for placing cuts. You can use the -c/--start-col-name option to use another column if you would prefer, e.g. to use "Start Time" column instead:

scenedetect -i video.mp4 load-scenes -i scenes.csv -c "Start Time"