SDH on paid social is a mute brief

SDH on paid social is a mute brief, not a broadcast deliverable. The ad is already watched with the sound off. Speech captions are the soundtrack. [music] and [door] matter when the sound is the joke or the proof. Otherwise SDH is a long-form and course job. Do not dump a labelled mix onto a six-second UGC ad.

Paid is already mute

SDH means subtitles for the deaf and hard of hearing: speech, plus who is speaking when that is not obvious, plus non-speech audio that carries plot or tone. Regular subs assume you can hear the mix. SDH assumes you cannot.

Paid social assumes you cannot too — for a different reason. Autoplay is muted. The impression is silent whether or not the viewer has hearing loss. A speech-only burn that the mute scroller can finish is already doing the main SDH job for that surface: the words are in the picture. That is why burned-in captions won the format.

What paid social does not automatically need is a second script of labelled effects. [upbeat music], [whoosh], [clicks tongue] on every cut is a radio play sitting on an ad that already has a legal line, a CTA, and two lines of speech. The mute viewer did not come for the Foley log. They came for the claim.

The exception is load-bearing sound. If the joke is a door slam, the mute viewer misses the cut unless you write [door slams]. If the proof is a timer beep, write the beep. If a song lyric is the hook and you cannot license the lyric as type, you have a different problem — but if the presence of music is the gag (“wait, that’s our hold music”), label it. Sound as joke or proof: write it. Sound as bed: skip it.

Speaker IDs follow the same test. A single on-camera creator does not need JESS: on every cue. A two-voice testimonial where the crop hides a face does. Off-screen VO over product: ID once. Spam IDs spend CPS the offer needed.

When the sound is the joke

Run a mute pass and ask, at each cut: does this edit still make sense if I never hear it?

• Yes, because the mouth said the claim. Speech captions are enough. No brackets.
• No, because the cut is a sound. Write the label. Keep it short. [glass breaks], not a sentence describing the glass.
• No, because two people talk and I cannot see who. Speaker ID. Then speech.
• No, because the legal meaning lives in a tone. That is unusual in performance ads and common in some demo spots (a heart-rate beep, an alarm). Label it, and still put the claim in words. Do not rely on the bracket alone.

Auto-transcribe will not invent these labels. ASR is a listen on speech. SDH is an author pass after that. If the brief is “full SDH,” someone types the brackets. A DYNAMIC karaoke preset is not SDH. Motion is not [music].

Paid review will read the brackets as copy. Keep them factual. Do not write [epic drop] on a UGC ad. Do not write [sad violin] unless you are prepared to defend it like a claim. [music] is enough when the point is that music is happening.

Article 50 marks and paid-partnership rows are not SDH. They are disclosures. They still occupy the same safe area. Sequence them. A [door] cue on top of “Paid partnership” on top of two-line speech is three jobs in a band that had room for one.

Full SDH belongs on courses and long-form

A course player, a webinar archive, a YouTube long-form, a broadcast master: the viewer may sit, may use a CC toggle, may need the mix described because they are not looking at a six-second hook. That is the SDH job as standards mean it.

Deliver a sidecar there — SRT or VTT — with speech, IDs, and non-speech labels. Closed captions are the object; SDH is the flavour of content in that object. The player can restyle. You can patch a miss without a re-render. Convert with the VTT to SRT converter rather than renaming files.

Do not confuse that master with the paid cut of the same footage. The ad is a clip: insight, offer, mute. Burn speech. Add a bracket only for load-bearing sound. Keep a BASIC two-line look (21 presets at 1×). DYNAMIC 2× word-pop is the wrong tone for a labelled mix and the wrong height for a legal row.

If a buyer’s accessibility rider asks for SDH on the ad, ask which sounds carry meaning. A rider that means “captions on mute” is already satisfied by a readable speech burn. A rider that means “describe the mix” is a longer file, usually not the 9:16 paid unit. Put full SDH on the landing-page video and the course. Put mute-brief captions on the ad.

Deaf viewers still see the ad in the feed. That is why the speech burn has to be finishable, high-contrast, two lines, out of chrome — the same bar you should already be meeting. Full labelled SDH on a six-second clip is how you fail that bar by filling it with brackets.

Mute brief, then burn

1. Cut the ad. Picture lock.
2. Transcribe speech. Fix names, prices, SKUs.
3. Mute pass. Circle the cuts that die without sound.
4. Author only those labels, plus IDs where the face does not do the work.
5. Burn BASIC, two lines, outline or backdrop. Leave room for legal overlay.
6. Recheck mute. If a bracket pushed the claim off the hold, the bracket lost.

The AI caption generator and add captions to a video transcribe. 165 language codes. They do not write [door]. Overlays and an edited transcript do.

Paid social is a mute brief. Write the sound when the sound is the joke. Leave full SDH to the room with a player and a longer sit.

Label only the load-bearing sound, then burn in Versely.

FAQ

Is a speech-only burn “SDH” on a muted ad?

It is the mute brief: words in the picture. It is not full SDH until you add speaker IDs where needed and labels for sounds that carry meaning. Most UGC ads need the first and only a few of the second.

Should I caption bed music on every paid cut?

No. Bed music is not the joke. [music] on every file is a second script. Label music when its presence is the gag or the plot.

Where does full SDH go if not on the ad?

Course players, webinar archives, YouTube long-form, broadcast masters: sidecar SRT or VTT with speech, IDs, and non-speech labels. The ad keeps a readable speech burn plus brackets only for load-bearing sound.