Append Sponsorblock Metadata (marks) with Python API?

[nickv2002]

I'm trying to use the yt-dlp version 2021.10.10 with the Python API to download YouTube videos and append the SponsorBlock data as chapter markers, but I can't get the sponsor chapters to get written out.

Here's my debug output from where I've gotten so far. Note that it does check with SponsorBlock and correctly find 3 segments.

[debug] [youtube] Extracting URL: http://www.youtube.com/watch?v=xaQJbozY_Is
[youtube] xaQJbozY_Is: Downloading webpage
[youtube] xaQJbozY_Is: Downloading android player API JSON
[debug] Sort order given by extractor: quality, res, fps, source, codec:vp9.2, lang
[debug] Formats sorted by: hasvid, ie_pref, quality, res, fps, source, vcodec:vp9.2(10), acodec, lang, filesize, fs_approx, tbr, vbr, abr, asr, proto, vext, aext, hasaud, id
[debug] Testing thumbnail 41
[debug] SponsorBlock query: https://sponsor.ajay.app/api/skipSegments/584d?service=YouTube&categories=%5B%22outro%22%2C+%22selfpromo%22%2C+%22preview%22%2C+%22interaction%22%2C+%22sponsor%22%2C+%22music_offtopic%22%2C+%22intro%22%5D
[SponsorBlock] Found 3 segments in the SponsorBlock database
[info] xaQJbozY_Is: Downloading 1 format(s): 22
[debug] Invoking downloader on "https://r4---sn-nx5s7n7z.googlevideo.com/videoplayback?expire=1634370143&ei=_y1qYaPEKYvEj-8Pzd-z-Aw&ip=97.115.134.231&id=o-ACdFqtEZcoU6i8H46WkEYCKChE66HZzp-MCI120VPrFQ&itag=22&source=youtube&requiressl=yes&mh=Yr&mm=31%2C29&mn=sn-nx5s7n7z%2Csn-nx57ynls&ms=au%2Crdu&mv=m&mvi=4&pl=20&initcwndbps=962500&vprv=1&mime=video%2Fmp4&cnr=14&ratebypass=yes&dur=700.963&lmt=1634129774082154&mt=1634348223&fvip=4&fexp=24001373%2C24007246&c=ANDROID&txp=5516222&sparams=expire%2Cei%2Cip%2Cid%2Citag%2Csource%2Crequiressl%2Cvprv%2Cmime%2Ccnr%2Cratebypass%2Cdur%2Clmt&sig=AOq0QJ8wRQIgCvgMnmbhNDcZG3CN1DAxCvc8QBKCgH-CM2i-Mhwkqt8CIQCf_x2-buOrQB2tdJFq1f1oZ1-Ni07pLnkPO86STczS-w%3D%3D&lsparams=mh%2Cmm%2Cmn%2Cms%2Cmv%2Cmvi%2Cpl%2Cinitcwndbps&lsig=AG3C_xAwRAIgaPAtabAyTKRkXQfBfIKeongvSUgB046bIiea1huClRMCIFtCXqhz0ZlO16imLq-ioVewGknvCsYbp6kX-MN6GRD7"
[download] Destination: /youtube-dl/What Dinosaurs ACTUALLY Looked Like [xaQJbozY_Is].mp4
[download] 100% of 91.62MiB in 00:08 
[Metadata] Adding metadata to "/youtube-dl/What Dinosaurs ACTUALLY Looked Like [xaQJbozY_Is].mp4"
[debug] ffmpeg command line: ffmpeg -y -loglevel repeat+info -i 'file:/youtube-dl/What Dinosaurs ACTUALLY Looked Like [xaQJbozY_Is].mp4' -i 'file:/youtube-dl/What Dinosaurs ACTUALLY Looked Like [xaQJbozY_Is].meta' -map 0 -dn -c copy -map_metadata 1 'file:/youtube-dl/What Dinosaurs ACTUALLY Looked Like [xaQJbozY_Is].temp.mp4'

However when it writes out the file, I just get a mp4 video with the Chapters from the YouTube metadata but not the Sponsorblock.

Here's a simplified version of my code to show what I'm doing:

import yt_dlp
postprocessors = []

postprocessors.append({
    'key': 'SponsorBlock',
    'categories': ["outro", "selfpromo", "preview", "interaction", "sponsor", "music_offtopic", "intro"],
    # Run this immediately after extraction is complete
    'when': 'pre_process'
})
postprocessors.append({
    'key': 'FFmpegMetadata',
    'add_chapters': True,
    'add_metadata': False,
})

ydl_options = {
    'postprocessors': postprocessors,
}

with yt_dlp.YoutubeDL(ydl_options) as ydl:
    ydl.download(['http://www.youtube.com/watch?v=xaQJbozY_Is'])

---

BTW to get the correct output this works via the CLI (with some extra quality markers thrown in that don't matter just to make the download quicker).

$ yt-dlp -v -f '(bv*+ba/b)[vcodec^=avc1] / (bv*+ba/b)' "http://www.youtube.com/watch?v=xaQJbozY_Is" --sponsorblock-mark all
[debug] Command-line config: ['-v', '-f', '(bv*+ba/b)[vcodec^=avc1] / (bv*+ba/b)', 'http://www.youtube.com/watch?v=xaQJbozY_Is', '--sponsorblock-mark', 'all']
[debug] Encodings: locale UTF-8, fs utf-8, out utf-8, pref UTF-8
[debug] yt-dlp version 2021.10.10 (zip)
[debug] Python version 3.9.7 (CPython 64bit) - macOS-11.6-arm64-64bit
[debug] exe versions: ffmpeg 4.4, ffprobe 4.4, rtmpdump 2.4
[debug] Optional libraries: sqlite
[debug] ANSI escape support: stdout = True, stderr = True
[debug] Proxy map: {}
[debug] [youtube] Extracting URL: http://www.youtube.com/watch?v=xaQJbozY_Is
[youtube] xaQJbozY_Is: Downloading webpage
[youtube] xaQJbozY_Is: Downloading android player API JSON
[debug] Sort order given by extractor: quality, res, fps, source, codec:vp9.2, lang
[debug] Formats sorted by: hasvid, ie_pref, quality, res, fps, source, vcodec:vp9.2(10), acodec, lang, filesize, fs_approx, tbr, vbr, abr, asr, proto, vext, aext, hasaud, id
[debug] Testing thumbnail 41
[debug] SponsorBlock query: https://sponsor.ajay.app/api/skipSegments/584d?service=YouTube&categories=%5B%22sponsor%22%2C+%22interaction%22%2C+%22outro%22%2C+%22preview%22%2C+%22selfpromo%22%2C+%22music_offtopic%22%2C+%22intro%22%5D
[SponsorBlock] Found 3 segments in the SponsorBlock database
[info] xaQJbozY_Is: Downloading 1 format(s): 22
[debug] Invoking downloader on "https://r4---sn-nx57ynls.googlevideo.com/videoplayback?expire=1634546559&ei=H99sYfPTMYiDj-8P55S4uAM&ip=97.115.134.231&id=o-APPgWnAK37V8JXrWPqpaDtLJDusBRvHRe8i6DgOOIPv9&itag=22&source=youtube&requiressl=yes&mh=Yr&mm=31%2C26&mn=sn-nx57ynls%2Csn-n4v7sn76&ms=au%2Conr&mv=m&mvi=4&pl=20&initcwndbps=951250&vprv=1&mime=video%2Fmp4&cnr=14&ratebypass=yes&dur=700.963&lmt=1634129774082154&mt=1634524843&fvip=4&fexp=24001373%2C24007246&c=ANDROID&txp=5516222&sparams=expire%2Cei%2Cip%2Cid%2Citag%2Csource%2Crequiressl%2Cvprv%2Cmime%2Ccnr%2Cratebypass%2Cdur%2Clmt&sig=AOq0QJ8wRgIhANoZc1OMCunh--mKrIxRGu4Hq-yPvSFu6eZVTIfOge71AiEA0f3-B-_QncL_FBtq6BLVnBfK7rVoPCGvh2bg0RkJ4wc%3D&lsparams=mh%2Cmm%2Cmn%2Cms%2Cmv%2Cmvi%2Cpl%2Cinitcwndbps&lsig=AG3C_xAwRQIgOehfUqeOtXjlD9BKKVms_ffq9gRTjco7PC2XMpGxZwMCIQC13SFu6D9yJ8KDJPrbUsZWMcCGU3LKr60reqgAsPJhUg%3D%3D"
[download] Destination: What Dinosaurs ACTUALLY Looked Like [xaQJbozY_Is].mp4
[download] 100% of 91.62MiB in 00:01
[debug] ffprobe command line: ffprobe -hide_banner -show_format -show_streams -print_format json 'file:What Dinosaurs ACTUALLY Looked Like [xaQJbozY_Is].mp4'
[Metadata] Adding metadata to "What Dinosaurs ACTUALLY Looked Like [xaQJbozY_Is].mp4"
[debug] ffmpeg command line: ffmpeg -y -loglevel repeat+info -i 'file:What Dinosaurs ACTUALLY Looked Like [xaQJbozY_Is].mp4' -i 'file:What Dinosaurs ACTUALLY Looked Like [xaQJbozY_Is].meta' -map 0 -dn -c copy -map_metadata 1 'file:What Dinosaurs ACTUALLY Looked Like [xaQJbozY_Is].temp.mp4'

This results in the following chapters

What extra processing steps do I need to add in so that the SponsorBlock chapters get written out to the file?

[pukkandan]

See #1238 (comment)

tl;dr: add ModifyChapters PP

Note: It must be added before FFmpegMetadata

[nickv2002]

Thanks, that got me 90% of the way. I didn't want to remove them just mark them so I ended up with:

import yt_dlp
postprocessors = []

postprocessors.append({
    'key': 'SponsorBlock',
    'categories': ["outro", "selfpromo", "preview", "interaction", "sponsor", "music_offtopic", "intro"],
    # Run this immediately after extraction is complete
    'when': 'pre_process'
})
postprocessors.append({
    'key': 'ModifyChapters',
})
postprocessors.append({
    'key': 'FFmpegMetadata',
    'add_chapters': True,
    'add_metadata': False,
})

ydl_options = {
    'postprocessors': postprocessors,
}

with yt_dlp.YoutubeDL(ydl_options) as ydl:
    ydl.download(['http://www.youtube.com/watch?v=xaQJbozY_Is'])

With the key being to call the ModifyChaptersPP before the FFmpegMetadataPP as you stated.