Unable to extract CC3 closed captions

[Jessecar96]

Hello,

I have a video file that was extracted from a Fairly OddParents DVD that contains closed captions only on CC3. The CCExtractor GUI does not find these, and I cannot find any combination of CLI options that will extract them either.

I've uploaded a 30 second example of the file here:
https://weather.horse/samples/cc3.mkv (24 MB, right click and save as)

I can confirm with VLC that the captions do exist on CC3:

Is there any way I'm missing to extract CC3 or is this something that needs to be added?

[bbgdzxng1]

A useful little trick is to launder the file through FFmpeg and copy to a ts, nut, m2v or h264 container. Avoids some of the quirks of MKV containers.

$ ffmpeg -i "./cc3.mkv" -map '0:v:0' -c:v 'copy' "./cc3.ts"  

$ ccextractor --quiet --output-field 2 --out "ttxt" --stdout "./cc3.ts"

00:00:01,936|00:00:04,003|POP|♪ TIMMY IS AN AVERAGE KID ♪
00:00:04,004|00:00:06,906|POP|♪ THAT NO ONE UNDERSTANDS ♪
00:00:07,208|00:00:09,075|POP|♪ MOM AND DAD AND VICKY ♪
00:00:09,310|00:00:10,877|POP|♪ ALWAYS GIVING HIM COMMANDS ♪
00:00:10,878|00:00:12,379|POP|    BAD TWERP!
00:00:12,380|00:00:14,080|POP|        ♪ DOOM AND GLOOM
00:00:12,380|00:00:14,080|POP|        UP IN HIS ROOM ♪
00:00:14,081|00:00:15,648|POP|    ♪ IS BROKEN INSTANTLY ♪
00:00:16,150|00:00:19,219|POP|♪ BY HIS MAGIC LITTLE FISH
00:00:16,150|00:00:19,219|POP|WHO'LL GRANT HIS EVERY WISH ♪
00:00:19,487|00:00:20,520|POP|    ♪ 'CAUSE IN REALITY... ♪
00:00:20,521|00:00:24,424|POP|    ♪ THEY ARE HIS ODDPARENTS,
00:00:20,521|00:00:24,424|POP|    FAIRLY ODDPARENTS ♪
00:00:24,425|00:00:26,760|POP|            WANDS AND WINGS.
00:00:24,425|00:00:26,760|POP|FLOATY, CROWNY THINGS.
00:00:26,994|00:00:28,895|POP|    ♪ ODDPARENTS,
00:00:26,994|00:00:28,895|POP|    FAIRLY ODDPARENTS ♪
00:00:28,896|00:00:30,397|POP|    ♪ REALLY MOD, PEAPOD,
00:00:28,896|00:00:30,397|POP|    BUFF BOD, HOT ROD ♪

Other output options...

$ ccextractor --quiet --output-field 2 --out "g608" --stdout "./cc3.ts"
$ ccextractor --quiet --output-field 2 --out "scc" --stdout "./cc3.ts"
$ ccextractor --quiet --output-field 2 --out "srt" --stdout "./cc3.ts"
$ ccextractor --quiet --output-field 2 --out "webvtt-full" --stdout "./cc3.ts"

And a neat trick when outputting a report is to specify both fields (report is actually misleading if you don't) and use -o - rather than -stdout, since this works around the Error: You can't extract both fields to stdout at the same time in broadcast mode.

$ ccextractor --output-field "both" --out "report" --report-format 'json' -o - "./cc3.ts" 

{
  "schema": {
    "name": "ccextractor-report",
    "version": "1.0"
  },
  "input": {
    "source": "file",
    "path": "./cc3.ts"
  },
  "stream": {
    "mode": "Transport Stream",
    "program_count": 1,
    "program_numbers": [1],
    "pids": [
      { "pid": 256, "program_number": 1, "codec": "MPEG-2 video" }
    ]
  },
  "programs": [
    {
      "program_number": 1,
      "summary": {
        "has_any_captions": true,
        "has_608": true,
        "has_708": false
      },
      "services": {
        "dvb_subtitles": false,
        "teletext": false,
        "atsc_closed_caption": true
      },
      "captions": {
        "present": true,
        "eia_608": {
          "present": true,
          "xds": false,
          "channels": {
            "cc1": false,
            "cc2": false,
            "cc3": true,
            "cc4": false
          }
        },
        "cea_708": {
          "present": false,
          "services": []
        }
      },
      "video": {
        "width": 720,
        "height": 480,
        "aspect_ratio": "02 - 4:3",
        "frame_rate": "04 - 29.97"
      }
    }
  ]
}

[Varadraj75]

I dug into this and the root cause is that CCExtractor's MKV demuxer only
recognizes V_MPEG4/ISO/AVC and V_MPEGH/ISO/HEVC tracks — V_MPEG2
tracks (common in DVD-sourced MKV files) are silently skipped entirely,
which is why no captions are found regardless of which CC channel you target.

I've tested with your sample file and confirmed the fix works:

ccextractor cc3.mkv --output-field 2 -o output.srt

Produces the full CC3 caption output correctly.

PR: #2152 — adds V_MPEG2 track detection and processing to the MKV
demuxer using the existing process_m2v() infrastructure already used
by mp4.c.

[bbgdzxng1]

@Varadraj75 ✅ Great catch on the mpeg2video/H.262 in MKV, especially given the popularity of makemkv for DVD-Video extraction. Great enhancement.

@Varadraj75 interesting observation from OP's source that --out report does not pick up CC3 by default. This is off-topic and probably deserves an entirely separate ticket - and testing with a sample containing both CC1/3 to test thoroughly.

Thoughts...

• Either cc3/4 should be excluded from report, if report is only --output-field 1 by default
• or, report should be more promiscuous by using --output-field "both" by default, but this gets messy with Error: You can't extract both fields to stdout at the same time in broadcast mode

What's your opinion? I guess it is because report is technically an "output" rather than a holistic input analyzer. Off-topic for this ticket.

ccextractor --out "report" -o - "./cc3.ts"  
File: ./cc3.ts
Stream Mode: Transport Stream
Program Count: 1
Program Numbers: 1 
PID: 256, Program: 1, MPEG-2 video
//////// Program #1: ////////
DVB Subtitles: No
Teletext: No
ATSC Closed Caption: Yes
EIA-608: Yes
XDS: No
CC1: No
CC2: No
CC3: No <<<<<<<<<<<<<<<<<<<<<<< Kinda misleading for CC3 to be included in the report, because we know CC3s are present in this file
CC4: No
CEA-708: No
Width: 720
Height: 480
Aspect Ratio: 02 - 4:3
Frame Rate: 04 - 29.97

MPEG-4 Timed Text: No

ccextractor --output-field "both" --out "report" -o - "./cc3.ts" 
File: ./cc3.ts
Stream Mode: Transport Stream
Program Count: 1
Program Numbers: 1 
PID: 256, Program: 1, MPEG-2 video
//////// Program #1: ////////
DVB Subtitles: No
Teletext: No
ATSC Closed Caption: Yes
EIA-608: Yes
XDS: No
CC1: No
CC2: No
CC3: Yes
CC4: No
CEA-708: No
Width: 720
Height: 480
Aspect Ratio: 02 - 4:3
Frame Rate: 04 - 29.97

MPEG-4 Timed Text: No

[Varadraj75]

Great point @bbgdzxng1 — the report mode behavior is genuinely misleading.
CC3: No when the captions are actually there is worse than not showing CC3
at all, since it could make users conclude the file has no CC3 content.

I agree this deserves a separate ticket. The cleanest fix is probably to have
--out report internally use --output-field both for its analysis pass,
with the broadcast mode conflict handled gracefully (just suppress the error
when in report-only mode since there's no actual stdout conflict).

I'll open a separate issue for this once #2152 is settled.