NHK+'s strange ARIB-in-SRT caption format

NHK Plus, NHKプラス

NHK+'s subtitles use a strange format

here is an example:

WEBVTT

21:15:03.927 --> 21:15:04.679
[CS][CS]

21:15:04.679 --> 21:15:12.054
[CS][CS][SWF_7][SDF_840_480][SDP_58_29][SHS_4][SVS_24][SSM_36_36][APS_7_2][WHF][COL_4][COL_5_1]♬〜

21:15:12.054 --> 21:15:15.425
[CS][CS][SWF_7][SDF_840_480][SDP_58_29][SHS_4][SVS_24][SSM_36_36][APS_5_2][CNF][COL_4][COL_5_1]うわ〜[MSZ]　[NSZ]すごい！[MSZ][APS_6_5][NSZ]うわっ[MSZ]　[NSZ]ぶつかっちゃいそうですね！

21:15:15.425 --> 21:15:18.828
[CS][CS][SWF_7][SDF_840_480][SDP_58_29][SHS_4][SVS_24][SSM_36_36][MSZ][APS_6_17][NSZ][CNF][COL_4][COL_5_1]いや〜[MSZ]。

21:15:18.828 --> 21:15:22.500
[CS][CS][SWF_7][SDF_840_480][SDP_58_29][SHS_4][SVS_24][SSM_36_36][APS_6_5][CNF][COL_4][COL_5_1]ここは川ではなくって[MSZ][APS_7_11][NSZ]海ということですか？

21:15:22.500 --> 21:15:26.735
[CS][CS][SWF_7][SDF_840_480][SDP_58_29][SHS_4][SVS_24][SSM_36_36][APS_3_2][YLF][COL_4][COL_5_1]そうです[MSZ]。　[NSZ]今日はですね[MSZ][APS_4_5][NSZ]海を旅するサケに迫っていきます[MSZ]。

21:15:26.735 --> 21:15:29.171
[CS][CS][SWF_7][SDF_840_480][SDP_58_29][SHS_4][SVS_24][SSM_36_36][APS_7_3][CNF][COL_4][COL_5_1]海の中の姿って見たことないですね[MSZ]。

21:15:29.171 --> 21:15:31.840
[CS][CS][SWF_7][SDF_840_480][SDP_58_29][SHS_4][SVS_24][SSM_36_36][SSZ][APS_8_4][NSZ][YLF][COL_4][COL_5_1]ですよね[MSZ]。[SSZ][APS_10_20][NSZ][COL_0][CNF]はい[MSZ]。

21:15:31.840 --> 21:15:39.618
[CS][CS][SWF_7][SDF_840_480][SDP_58_29][SHS_4][SVS_24][SSM_36_36][SSZ][APS_6_9][NSZ][GRF][COL_4][COL_5_1]川で生まれたサケの稚魚は[SSZ][APS_8_10][NSZ]春になると川を下り➡

21:15:39.618 --> 21:15:43.855
[CS][CS][SWF_7][SDF_840_480][SDP_58_29][SHS_4][SVS_24][SSM_36_36][SSZ][APS_12_11][NSZ][GRF][COL_4][COL_5_1]大海原へ旅立ちます[MSZ]。

21:15:43.855 --> 21:15:49.555
[CS][CS][SWF_7][SDF_840_480][SDP_58_29][SHS_4][SVS_24][SSM_36_36][APS_6_9][GRF][COL_4][COL_5_1]実に２万[MSZ]ｋｍ[NSZ]もの[MSZ][APS_7_19][NSZ]大回遊の始まりです！[TIME_57][CS]

21:15:49.555 --> 21:15:51.865
[CS][CS]

these timestamps are completely wrong for some reason
something something mpegts, idk exactly why it happens but if you mux them into the mp4 when you rip then the problem magically goes away
the first line seems to always start at 0:00:00.0 though, so you can resync manually if you need to

this is taken from サイエンスＺＥＲＯ　２万ｋｍの旅！“サケの大回遊”に迫る

they display on the site like this:

The font is Hiragino TV Sans Rd S, you can pinch it from the dev tools quite easily
Interestingly they also have an option to display the subs below the video, which also disables a lot of the formatting
i don't think i've ever seen that before

looking at the raw line, we can see that it is composed of text+tags, quite similar to ASS in fact

[CS][CS][SWF_7][SDF_840_480][SDP_58_29][SHS_4][SVS_24][SSM_36_36][APS_6_9][GRF][COL_4][COL_5_1]実に２万[MSZ]ｋｍ[NSZ]もの[MSZ][APS_7_19][NSZ]大回遊の始まりです！[TIME_57][CS]

but there are no styles with defaults to be reused here, everything is re-specified every time, because >broadcast

the tags are listed in the ARIB STD-B24 standard, which defines the binary subtitle format used in actual broadcasts
https://www.arib.or.jp/english/std_tr/broadcasting/std-b24.html "Fascicle 1"
starting from Table 7-15 on page 89

referencing this, I came up with a hacky solution to "convert" them manually:

find and replace
\[APS_\d+_\d+\] -> \\N (regex) (position change -> newline)

[WHF] -> {\c&HFFFFFF&}

[YLF] -> {\c&H00FFFF&}

[CNF] ->
      {\c&HFFFF00&}

[GRF] -> {\c&H00FF00&} (white, yellow, cyan, and green coloured text respectively)

search for [DRCS_ and replace with the corresponding character from nhk's lookup table (needs jp ip/vpn)
you will only have this if you have strange characters like the

in Osamu Tezuka's name
something to do with character sets not having certain characters i think, idk

for example:

[CS][CS][SWF_7][SDF_840_480][SDP_58_29][SHS_4][SVS_24][SSM_36_36][MSZ][APS_4_5][WHF][COL_4][COL_5_1]＜[NSZ]フルカラーでよみがえった[SSZ][APS_10_7]てづか　おさむ[APS_12_6][NSZ]手[DRCS_36_36_2_AADwAAAAAAAAAADwAAAAAAAAAADwD/////8AAADwD/////8AAADwDwAAAA8AAADwDwAAAA8AAADwDwAAAA8AAADwDwAAAA8AAADwAP////AAAADwAP////AAAP//8AAP8AAAAP//8AAP8AAAAADwAADwDwDwAADwAADwDwDwAADwAA8PDw8AAADwAA8PDw8AAADwD/AA8PAAAADwD/AA8PAAAADwAAAP8PAAAADwAAAP8PAAAADwAPDw8PAAAADwAPDw8PAAAADw8A8A8PAAAADw8A8A8PAAAAD/D/Dw8A8AAAD/D/Dw8A8AAP8AAAAP8A8AAP8AAAAP8A8AAAAAAADw8A8AAAAAAADw8A8AAAAAAP8A8ADwAAAAAP8A8ADwAAAA/wAA8AAAAAAA/wAA8AAAAAAAAAD/AAAAAAAAAAD/AAAA]治虫の名作をお楽しみ下さい[MSZ]＞

find that mess in the lookup table

and the line becomes:

[CS][CS][SWF_7][SDF_840_480][SDP_58_29][SHS_4][SVS_24][SSM_36_36][MSZ][APS_4_5][WHF][COL_4][COL_5_1]＜[NSZ]フルカラーでよみがえった[SSZ][APS_10_7]てづか　おさむ[APS_12_6][NSZ]手塚治虫の名作をお楽しみ下さい[MSZ]＞

finally:
\[[^]]+\] -> empty string (regex) (strip all other tags as we don't want to deal with them)

^\\N -> empty string (regex) (remove leading \N left over from the initial position tag)

this gives you the bare minimum of SRT but with colours
and it proves the concept of converting it to a usable format

idk why they would do it like this
i can only imagine that ISDB is so deeply embedded in their processes that they weren't able to let go of it

initially i thought of writing my own script to fully convert it to ASS for rendering
but it occurs to me that since it maps directly to the ISDB binary format, it may be better to convert it to a bitstream like in the actual broadcasts.
And then you can take advantage of the many ARIB -> something-else converters that already exist

-garret1317, April 2025