Deliver Texted and Textless on Every Spot

A texted-only master is a localization dead end. The day a market needs a different legal line, a different end card, or the product name in another language, you either composite type onto a file that already has English burned in, or you go back to the grade. Both are slower than shipping a textless the first time.

Netflix's own IMF glossary treats this as a named picture version, not a nice-to-have: a Textless Version has a timeline identical to the texted delivery, with no titles, credits, narrative text, location call-outs, or other supportive type. They want it after the texted package has already passed QC, and they want it as picture only. You do not need to be delivering IMF for the idea to apply to a :30.

What "textless" actually means

Textless is not "a version without captions." Captions are a subtitle problem. Textless is a picture problem: every bit of type that lives in the image (supers, pack shots with a claim, lower thirds, end cards, forced in-picture translations, location cards) is gone, and everything else is identical.

Identical means:

• Same cut. Same frame count. Same first frame of action.
• Same grade. Same grain. Same resize.
• Same audio if the audio does not contain a language-specific VO you are about to replace. For a music-and-effects bed under titles, keep the bed. For a VO that reads the super, the textless picture still matches; the VO is a different deliverable.

A "textless" that is an earlier offline, or a grade from Tuesday, or two frames shorter because someone trimmed the hold on the end card, will not take a restamp. The new title will drift, or the last card will sit on the wrong take.

The UGC delivery spec already asks for a clean master next to a captioned master. That clean master is the social cousin of a textless. Agency and streamer work needs the same split, plus a title sheet, because the type is not captions of speech. It is designed titles.

Keep titles on a video track

The only reliable way to export a matching pair is to never let titles become pixels in the picture until the last print.

In an NLE that means:

1. Lock picture on V1 (and V1 only). Grade that. The textless is V1.
2. Build every super, end card, pack claim, and location card on V2 (and V3 if you have a legal line that versions separately).
3. Export texted with V2/V3 on. Export textless with those tracks muted. Do not re-render V1 in between.

If a title was generated into the footage (a model that drew "SUMMER SALE" on the storefront, a Veo clip that arrived with garbled burned-in subtitles), it is not a title. It is scenery. You cannot mute it. You either inpaint it, recut it, or accept that this shot cannot have a textless. Catch that at generation, not at delivery.

Versely's overlay tools composite type into the export. A fixed text overlay burns a single line you supply. Timed text overlays burn a sequence you time yourself. Neither is a subtitle transcript, and the three tools people call "captions" are easy to mix up. For a texted/textless pair those overlay tools are how you print the texted. The textless is the same timeline with those overlays not applied. Do not run them on the only copy of the file.

Animated titles belong on the same rule. Title sequences that carry a story still have to sit on a track you can mute, or you are handing the next market a baked English open they cannot restamp without a new grade.

The title sheet

A textless without a title sheet is a blank plate and a guessing game. The sheet is a table, one row per title event, and it is the document a local market actually uses.

Column	Example
Event ID	T04
TC in	01:00:08:12
TC out	01:00:12:00
Duration	3:13 at 25 fps
Role	Super / end card / legal / location / pack claim
Source language copy	Free delivery over £40
Font, weight, size	Brand Sans Bold, 64 px at 1080
Color	#F5F0E6 on 80% black bar
Position	Lower third, 10% inset, inside the crop-safe band
Animation	Fade 8 frames in, hold, fade 8 frames out
Legal required?	Yes — must survive translation
Notes	Do not cover the pack logo

Frame rate on the sheet must match the master. A sheet in 24 fps against a 25 fps textless will put every title late. Titles are not captions, but they share a clock.

If a font is licensed only for the original market, say so. What OFL permits is not a substitute for the brand's own licence. A market that restamps in a substitute font still needs the position and the duration, which is why those columns exist even when the typeface cannot travel.

Ship the sheet as a spreadsheet, not a PDF of a spreadsheet. Local producers have to sort, filter, and paste.

How a market restamps without a new grade

The happy path:

1. They take your textless. They do not touch the grade.
2. They take the title sheet. They translate the copy, or swap the legal line, and they keep TC in, TC out, position, and animation unless the translated string physically will not fit.
3. They composite their titles onto V2. They export a local texted.

That is the entire point of the pair. Localizing brand videos can then be a dub or a subtitle job on a picture that still has room for type. If you only shipped a texted English master, they are either covering your type with a patch (which looks like a patch) or asking you to regrade a new textless you should have printed on day one.

Two failure modes show up constantly:

• The pack shot is the title. The claim is designed into the label on a generated product. There is no V2. You cannot restamp a bottle. Shoot or generate a clean pack, and put the claim on a super.
• The end card is a rendered poster. A model made a beautiful lockup with the logo and the URL. Pretty, and unlocalisable. Build the end card in a layout tool from the logo file, then composite it. The textless holds on the last picture, or on black, for the same duration.

Netflix's ordering is worth copying even on a spot: QC the texted first, then print the textless from the same timeline, then let markets version. A textless you grade after the texted "just to make it nicer" is a second master, and the titles will no longer sit on the picture they were designed against.

FAQ

Do social ads need a textless too?

If any type is in the picture other than burned captions of speech, yes. A unique end card, a price, a URL, a legal line: all of those version. Captions of the VO are a different object; ship a clean master and a captioned master as well, per the UGC spec. Do not treat those two pairs as the same folder item.

Can I generate the titles into the shot if they look better?

Only if you are sure no one will ever change the words. The moment a lawyer, a retailer, or a market manager edits a line, baked type is a reshoot. Designed type on V2 is a re-export. Looking better in the original language is not a reason to destroy the second language.

What if the textless is longer because we held for a title?

Then the texted and the textless do not match, and a market cannot restamp. Hold the picture for the title duration on V1 with the titles muted. The hold is editorial. It belongs in both versions. The title is what changes.

Who owns the title sheet, production or the market?

Production writes the source-language sheet and the timing. The market writes the translated copy into new columns, and they send that sheet back with the local master. If they change a duration, they say so on the row. A sheet that only exists in someone's project file is not a deliverable.