YouTube ads still want burn-in

YouTube ads still want burn-in. An .srt is a watch-page object: a CC button, a player, a viewer who sat down. Demand Gen and leftover TrueView-style inventory mix in-stream with in-feed and Shorts. Those last surfaces autoplay muted. Mute impressions do not enable your sidecar. Pixels in the file are the ad.

The watch page is not the ad graph

YouTube long-form is the exception that trained a bad habit. Landscape watch page has a player, a caption track, and an ingest path for SRT or Studio captions. Closed captions are correct there: searchable, restylable, language-swappable without a re-render. You can still burn-in if the look is the show. You should not only burn-in if you need a second language on that watch page.

Ads are not that room.

Video action campaigns — skippable in-stream and in-feed with a call to action — were upgraded into Demand Gen. First Pier’s YouTube ad-specs note, last verified 2 September 2026 against Google Ads Help, puts the final upgrades in April 2026. Demand Gen is broader than what it replaced: YouTube in-stream, YouTube in-feed, YouTube Shorts, Discover, Gmail, Maps, and Display from one campaign. You do not pick the surface per impression. Google does.

In-stream skippable can play in a player with sound. In-feed and Shorts ads start like the organic feed: muted, vertical or thumbnail-led, a silent first frame. Discover and Gmail are not a CC button either. If your campaign can serve any mute-first surface, the file needs burned-in captions. A sidecar you uploaded to a backing YouTube video does not travel into every Demand Gen slot as your art.

TrueView as a buying word still shows up in older briefs. Treat it as inventory that can include a player and a mute tile. Burn-in is the file that still talks when the tile is silent. The SRT is extra for the destination that can use it, not the delivery for the ad.

Platform captions are not your ad

YouTube can draw its own type.

Social Media Today, 10 February 2026, reported a mobile watch-page behaviour: captions switch on when the user mutes the device, then off when they unmute. YouTube’s own note in that write-up: not currently available for Shorts; on by default; can be turned off in caption settings. That is a convenience overlay on a watch page. It is not Demand Gen. It is not your font. It is not a skippable in-feed tile. It is not a reason to ship a silent ad master.

Shorts ads sit in a mute-first vertical feed. As of the 27 May 2026 YouTube Blog post (Improving AI labels for viewers and creators), photoreal / meaningfully AI-altered content gets an overlay on the Short. Long-form labels sit below the player, above the description. Labels alone do not impact recommendations or monetization eligibility. They do occupy pixels. Keep burned type out of that badge. Auto-captions stacked on your burn is a second collision. Keep the block compact.

In-stream can show player captions if a track exists and the viewer has them on. Many will not. The first five seconds before a skip are a mute-or-noise moment in a lot of living rooms. Burn-in still earns those seconds. A legal line and a price should not live only in a CC track the skippable viewer never enabled.

Do not trust “we uploaded captions in Studio” as the ad QA. Studio captions attach to a YouTube asset. Demand Gen will still serve a file. If the words are not in that file, the mute tile has no claim.

Burn the ad master; keep the sidecar for the watch page

Two objects, two jobs.

Ad master (always). Speech transcribed, names and numbers matched to the PDP, two-line BASIC for review, pixels in the MP4. Out of chrome, out of the Shorts overlay zone, off the pack label. DYNAMIC 2× only if you are testing type-as-creative on a mute tile and you still have a BASIC control.

Sidecar (sometimes). SRT or VTT for the watch-page upload, courses, accessibility teams, a second language on long-form. Convert with the VTT to SRT converter. Do not rename. Do not attach the sidecar to the ad and skip the burn.

If the same cut is both a Demand Gen asset and a 12-minute YouTube, you still burn the short and you still upload a track for the long. One pipeline is how a perfect SRT ships with a silent Shorts ad, or how three language burns get spent on a watch page that could have toggled.

Paid YouTube is a freeze-frame culture too. Policy review and brand screenshots pause the file. Karaoke mid-word is a half-lit offer. Static BASIC survives the pause. Same transcript as an organic Snap Reel if you have one; two burns.

Article 50 still wants a mark at first exposure on realistic synthetic media for EU viewers. That mark is not a subtitle cue and it is not an SRT. Put it in the picture. YouTube’s own AI overlay does not replace it.

What to actually export

Lock the ratio for the slot mix. Demand Gen will crop. Caption after the crop you will serve on Shorts and in-feed, not only on a 16:9 in-stream master. A 16:9 burn that dies in 9:16 is a silent vertical ad.

Then:

Transcribe the ad audio. 165 language codes. Code matches the tape.
Fix prices and SKUs to the PDP the ad lands on.
Burn BASIC, two lines, outline or backdrop. Preview on this plate.
Mute phone pass in a Shorts-shaped frame and in landscape. Both can serve.
Export MP4. That is the ad.
Export a sidecar if a watch page will exist.

The AI caption generator and add captions to a video are the burn path. They will not enable YouTube’s mute-to-captions feature for you. They put words in the file.

YouTube ads still want burn-in. The SRT is for the room with a CC button. Demand Gen is often not that room.

Burn the next YouTube ad master in Versely.

FAQ

If the campaign is only skippable in-stream, can I skip burn-in?

You can gamble. Many in-stream viewers still start loud, and a CC track can exist. Demand Gen will not stay “only in-stream” unless you constrain it, and the first seconds still have to work without hoping they enabled captions. Burn-in is the cheap insurance.

Will Studio captions on the backing video caption my Demand Gen ad?

Not as a controlled look on every surface. In-feed and Shorts tiles show the file. If the words are not in the pixels, the mute impression has no type. Keep Studio captions for the watch page.

Do Shorts ads work like long-form CC?

No. Shorts ads are mute-first vertical. Burn-in is the format. As of 27 May 2026, photoreal AI labels overlay the picture as well. Keep type out of that zone.