Forced Narrative Subs vs Full Subtitle Files
Burning English titles into a foreign master is not localization. Forced narrative, full subtitles, and SDH are three files that ride with different masters.
Burning English titles into a German picture and calling the file DE_final is not a German version. It is an English master with a confusing name. Forced-narrative, full subtitles, and SDH are three different timed-text objects, they ride with different picture and audio masters, and collapsing them into burned English is how you pay for a second localization when the first one "already shipped."
Three files, three jobs
The words get used as synonyms. They are not.
Forced narrative (FN) is the text the audience needs even when they have subtitles turned off. Amazon Video Central defines Forced Narratives as the translation of spoken dialogue and on-screen text that is in a different language from the primary audio, when creative intent requires that the viewer understand it. A French line in an otherwise English scene, a plot-critical shop sign, a "Three weeks later" card that is not already in the picture: those are FN. Netflix's timed-text style guides treat forced narrative as its own file (and tell you not to combine an FN cue with dialogue in the same subtitle). FN is a small file. Most rows are empty. That is correct.
Full subtitles are a complete translation of the spoken dialogue (and usually the FN events too) for viewers who do not speak the audio language. They are what you turn on. A German subtitle file against English audio is full subtitles. A German subtitle file against a German dub is usually the wrong object; the German viewer of a German dub does not need a translation of the German. They may still need FN for a third language inside the scene.
SDH (Subtitles for the Deaf and Hard of Hearing) is same-language (or dubbed-language) text that includes the dialogue plus speaker IDs and non-speech information: [door slams], [distant siren], [theme music]. Amazon's Video Central groups this with captions: "Captions/SDH: Timed text that includes both spoken dialogue and atmospherics for the deaf and hard-of-hearing." This is an accessibility deliverable, not a market-expansion deliverable. WCAG 2.2 Success Criterion 1.2.2 requires captions on prerecorded synchronized media. SDH is how a streamer usually meets that for the programme audio.
Closed captions in the CEA-608/708 sense are a broadcast-era cousin of SDH. On a streamer delivery they are often folded into the same requirement. Do not invent a fourth file unless the spec names .scc or IMSC captions as a separate ingest.
| Object | Who it is for | When it shows | What it contains |
|---|---|---|---|
| Forced narrative | Everyone, including subtitle-off | Always, even with subs off | Only the unintelligible bits |
| Full subtitles | Viewers who do not speak the audio | When subs are on | All dialogue, usually including FN |
| SDH | Deaf and hard-of-hearing viewers of that audio | When captions are on | Dialogue + speakers + non-speech |
ThreePlay's write-up of FN is consistent with the streamer defs: FN is delivered as a separate timed-text file, not burned into the video. Burning FN into the picture makes a textless impossible and makes a second-language FN a recut.
Which sidecar rides with which master
The matrix is the part teams skip, and it is the whole job.
| Master | Audio | Sidecars that belong with it |
|---|---|---|
| Original-language texted | Original | SDH in the original language. FN in the original language if the picture does not already carry those titles. |
| Original-language textless | Original | Same SDH. FN must now cover every plot-critical title you pulled out of the picture. Full subtitle files, one per market, against this audio. |
| Dubbed, market language | Dub | SDH in the dubbed language. FN only for languages still foreign inside that dub. Full subtitles in other languages against the dub, if you offer them. |
| Social, muted-feed | Original or VO | Burned-in captions of the speech, because the viewer will not toggle a sidecar. FN titles that are part of the design can be burned if you also ship a textless. |
Amazon is explicit on one trap: mezzanines delivered with Multi-Track Audio packages "must be semi-textless and must exclude embedded Forced Subtitles." They want the FN in the timed-text package, not in the picture. Netflix wants TTML (and IMSC as the profile IMF actually carries) for subtitles and SDH, not a pile of SRT as the archival original. SRT and WebVTT remain the right objects for a website <track> and for a YouTube upload. Burned-in versus SRT and VTT is the destination split; this page is the kind of file split.
Subtitle or dub, per market is the coverage question (Versely captions across 165 language codes, dubs into 25). Once you have chosen subtitle, you still have to choose FN / full / SDH. Dubbing a market does not retire SDH for that market. It creates a new SDH in the dubbed language.
Keep captions in sync with a dub is the timing follow-up: a full-subtitle file timed to English will not sit on a Spanish dub. Retiming is a sidecar job. It is not a reason to burn.
Why burn-in fails this job
Burn-in is the right tool for a muted social feed, where captions have to survive with the sound off and no one is opening a CC menu. It is the wrong tool for a localizable master.
Once English FN is pixels:
- A German picture department cannot restamp it. They cover it or they recut.
- A streamer that wants IMSC FN on a textless has to be told the FN is not available as text.
- SDH cannot add
[whispering]next to a line that is already a bitmap. - You cannot turn it off. A hearing, English-speaking viewer of the "German" master still sees English.
Versely's caption tools are burn-in tools. Automatic captions transcribe and composite styled captions into the frames. That is correct for the social cut. For the streamer or agency master, start from a plain transcript, take the words out as text, and author FN / full / SDH as sidecars in the format the spec names. Do not run the burn-in tool on the only copy and then try to recover a textless from it.
Audio description is a fourth object (a spoken track, WCAG 1.2.5). It is not a subtitle file. Do not put AD notes into SDH and call it done.
A concrete spot
A 30-second English ad, one line of on-screen French, a legal super, a VO, shipping to the UK and to France.
UK, original audio
- Picture: texted (legal super in English) and textless.
- FN: the French line, in English, timed to the line. Not the legal super (that is picture on the texted, and a title-sheet event on the textless).
- Full subs: none required for English-speaking UK, unless you choose to offer them.
- SDH: English dialogue +
[espresso machine hiss]+ speaker if two voices overlap. - Social cut: burned English captions of the VO. No bars. No FN file.
France, original audio (subtitled)
- Picture: the same textless, plus a French texted if they restamp the legal super.
- Full subs: French, including the French line (which is no longer foreign) and the English VO.
- FN: probably empty against French audio; against English audio with French subs off, the VO is unintelligible, which is a full-subtitle problem, not FN. FN is not "whatever we forgot to dub."
- SDH: not a substitute for French full subs.
France, dubbed
- Picture: same textless, French legal super.
- SDH: French, against the dub.
- FN: only if something in the dubbed version is still foreign.
- Full subs in French against a French dub: usually omitted.
If someone burned FREE DELIVERY and the French line in English onto the only picture, none of the French rows above are possible without a recut.
FAQ
Is forced narrative just "the titles"?
No. Titles that are part of the designed picture (a brand end card, a legal super in the original language) belong on a title sheet and a texted/textless pair. FN is timed text for things the audio language does not already cover. A location card you built in English is a title. A Japanese line spoken in an English scene is FN.
Can one TTML file contain FN, full subs, and SDH?
Some platforms fold FN events into the full subtitle and SDH files so that turning subs on does not drop the FN. Amazon says any subtitle, caption, or SDH file in an MTA package must contain a full translation of spoken dialogue and narrative text, including Forced Narratives. Netflix still talks about forced narrative files as their own object in the style guides. Follow the spec you are ingesting to. Do not assume one file satisfies a spec that named three.
We only post to social. Do we need any of this?
You need burned captions of the speech on the social file, and a clean master without them. You need SDH or a sidecar the moment the same cut goes to a website, a connected-TV buy, or a streamer. You need FN the moment a language the audience does not speak is load-bearing and you have not dubbed it. Social-only is a destination, not a permanent exemption.
Why not burn the English FN "for safety" and also ship a sidecar?
Because the burned FN is now in every texted and cannot be removed for a textless or a restamp. "For safety" is how you double-author and still fail QC on a semi-textless mezzanine. Pick pixels for social, sidecars for masters, and keep the picture clean enough that both are possible.