No filming:An AI avatar or clone yourself
    Versely
    Guides

    DYNAMIC captions on paid are a test

    BASIC is the paid default. DYNAMIC costs 2× and is a variant, not the master. Keep a BASIC control.

    Versely Team••7 min read

    BASIC is the default on paid. DYNAMIC costs 2× and is a variant you test, not the master you pin in the ad account. add_veed_captions bills 21 BASIC looks at 1× (simple, plain, beans, corpo, and the rest) and 9 DYNAMIC looks at 2× (glass, whisper, glide, fusion, terminal, handwritten, backdrop, and the rest). The multiplier is motion, not better transcription. Karaoke on an offer plus a legal line is two graphics fighting. Keep a BASIC control.

    This is the paid-test procedure. It is not “paid versus organic” as a format essay — that split already has a home. Here the rule is: do not let Glass become the paid source of truth.

    Lock BASIC as the paid master

    Paid is a compliance-plus-clarity job that happens to run in a 9:16 frame. The platform already expects a quieter surface: offer, legal, sometimes a paid-partnership mark, sometimes an Article 50 label. Phrase-level type is the right clock. The whole cue arrives together. The creative experiment is the hook, the face, the product — not the caption preset.

    Procedure for the master:

    1. Picture lock the ad cut, not the organic Reel. Paid often has a different first frame, a different crop, a different lower-third.
    2. Clean the transcript once: names, SKU, price, claim. That document is shared with any variant.
    3. Burn BASIC via add captions to video. Language code = spoken language. 165 codes.
    4. Two lines, high contrast, outline or a small backdrop. Off the legal row. Off safe area chrome. Paid chrome can include a CTA bar and a sponsored chip the organic file did not have.
    5. Author legal as overlay. Auto-transcribe will not write it. The legal row is not a caption.
    6. Preview BASIC on the first seconds of this ad cut. Contrast is against this plate.
    7. Export. This file is the control. Name it like one: sku_hookA_basic. Do not name it final until the test is over, but do treat it as the default that ships if the test is inconclusive.

    Reading speed still applies. A 1:1 BASIC block fails if the cue is a paragraph. Split at a clause. Do not “energise” density by switching to DYNAMIC. Motion does not add time. On a 6-second unit this is load-bearing.

    Speaker IDs and [music] belong on paid only when the sound is the joke or the proof. Most UGC ads are already mute-designed. A wall of brackets is a second script.

    The AI caption generator will burn Glass if you ask it to. It will not know this is an ad account. You do.

    Treat DYNAMIC as a named variant at 2×

    DYNAMIC is nine animated looks. Word-level. The active token punches. That is useful when the first seconds are the words and you are measuring whether type-as-creative lifts results. It inherits karaoke’s audio gate: dirty mix makes the highlight miss the syllable. Isolate or use a dry VO before you pay 2× to advertise drift.

    Procedure for the variant:

    1. Copy the same locked cut and the same corrected words. If you recut the hook, it is a different test (creative), not a caption-tier test.
    2. Pick one DYNAMIC preset. Name it in the file: sku_hookA_glass. Shopping all nine on paid is nine double-rate renders and no control.
    3. Preview with preview_caption_style on the first seconds. Full-captioning a look you have not seen is how 2× becomes a surprise.
    4. Place the animation off the mouth, off the pack letters, off the legal overlay. Karaoke sitting on a disclaimer is a fail regardless of CTR.
    5. Burn. Watch muted. If you cannot finish the line, you did not get a better ad. You got a faster flash. Cut words or drop back to BASIC.
    6. Confirm the bill: this render is 2× the BASIC master for the same transcription work. Thirty paid clips on Glass are sixty base caption renders. What captions cost across a month of clips is that arithmetic.

    Do not save Glass as the brand-kit default “because organic uses it.” A default charges every later paid render at 2×. Organic can keep DYNAMIC. Paid’s kit is BASIC until a test wins.

    Do not run DYNAMIC on the legal line. Legal is overlay, static, readable.

    If the paid placement is a 6-second bumper, DYNAMIC is an even harder sell. There is no time for a word clock to perform. BASIC plus a short overlay is the default; DYNAMIC is still allowed as a variant if the hypothesis is “type is the hook,” and the control still exists.

    Run the test with a control, then stop

    A test without a control is a vibe.

    Procedure:

    1. Hypothesis in one sentence: “Glass on this hook lifts thumb-stop versus Plain, same picture, same words.” If your hypothesis is actually “new hook,” test hooks, not caption tiers.
    2. Same campaign, same audience, same bid, same destination. Split the caption file only.
    3. Ship BASIC control and one DYNAMIC variant. Two cells. Not eight.
    4. Measure the thing you named: hook rate / CTR / hold, depending on the platform’s actual metric. Do not invent a Versely lift percentage.
    5. Watch both files muted. If DYNAMIC loses on readability, it is not a winner even if a dashboard wiggles. Unreadable type is a defect.
    6. If DYNAMIC wins on the metric and still reads, it can become the master for this hook, not for the whole account. The next SKU starts on BASIC again.
    7. If it loses or ties, keep BASIC. Delete Glass from the paid kit. You already paid 2× to learn that.

    What is not a test:

    • Replacing the master with Glass after one internal preference review.
    • Running DYNAMIC because the organic Reel used it, then uploading that file to Ads Manager with a sticker for the disclaimer.
    • A/B of Glass versus Fusion with no BASIC cell. You compared two 2× looks and learned nothing about the default.

    YouTube and Meta will still apply their own AI info / labels where those rules say. Caption tier does not discharge disclosure. Captions for silent autoplay is the mute-feed rule that both cells still have to pass.

    Do not let the variant overwrite the kit

    Finish like a test system, not like a buffet.

    Checklist:

    1. Brand kit / saved style for paid: a BASIC look. Simple, plain, or corpo. Phrase-level.
    2. Brand kit for organic: may be DYNAMIC if mute-hook type is the job. Separate save. Separate default.
    3. Filenames and the ad-account creative names include basic or the DYNAMIC preset, so nobody duplicates the winner onto the wrong SKU by accident.
    4. Legal overlay lives in the paid edit, not in the organic file you reused.
    5. Recaption if you trim. A DYNAMIC burn on an old in-point is a late highlight and a wasted 2×.
    6. Archive both renders with the hypothesis note. Next quarter’s “we always use Glass on paid” should have to survive this note.

    Credits never turn into USD here. The only number that matters on this page is the 2× multiplier on DYNAMIC versus BASIC, and the clip count it compounds across.

    Burn BASIC as the paid master, Glass only as a cell, in Versely. Bring a locked ad cut and a control name. Do not bring a brand kit that saved DYNAMIC for everything.

    FAQ

    Is DYNAMIC forbidden on ads?

    No. It is a variant. The default is BASIC. If the hypothesis is type-as-creative, run it against a BASIC control. Do not pin Glass as the paid master because organic likes it.

    Does DYNAMIC transcribe more accurately?

    No. Same speech-to-text. You are paying for motion. Accuracy is still names, numbers, and a clean mix.

    Can I use the organic Glass file in the ad account?

    Only if you are willing to skip the control, crowd the legal row, and pay 2× as the default. The procedure is two burns: BASIC master, optional DYNAMIC cell.