No filming:An AI avatar or clone yourself
    Guides

    Captions on carousel covers are a still

    The first carousel frame is a poster. Burn a line into that still. Do not ship a timed track nobody plays.

    Versely Team••7 min read

    The first frame of a carousel is a poster, not a timed caption track. People see a still (or a one-frame hold) before they swipe. A sidecar SRT attached to “the video” does not letter that cover. Karaoke on slide 4 does not help slide 1. Treat the cover as a thumbnail: extract or design a still, put one readable line on it, then decide whether later slides are silent graphics or actual video with real captions.

    This is the cover-still procedure. It is not a slideshow-feature tour.

    Name the cover as a poster

    Instagram carousels, LinkedIn carousels, and in-app slideshow exports all have a first tile. That tile does the hook job the first second of a Reel does. Mute autoplay may not even apply if the tile is a JPEG.

    Procedure:

    1. Write the destination: IG carousel, LinkedIn document, in-feed slideshow. Aspect is usually 4:5 or 1:1 on IG, sometimes 9:16. Do not design a 16:9 YouTube poster and upload it as slide 1.
    2. Write the one line the cover must carry. Same test as a first-second hook: a reason to stay (here, a reason to swipe). Not a logo. Not “swipe for more” as the only copy.
    3. Decide whether slide 1 is:
      • a still you designed,
      • a frame extracted from a video keeper, or
      • a one-second video loop.
    4. If it is a still or an extracted frame, you are in overlay-on-image territory. There is no word clock. Karaoke captions have nothing to highlight.
    5. If it is a loop with speech, it is a tiny video and can take burn-in — but the feed still is often the first frame of that loop. That first frame still has to read as a poster with the line already on it.

    Reject a cover that is a talking-head mid-phoneme with no type. That is a bad thumbnail in a swipe skin.

    Reject a cover that is only a burned caption track exported from a 15-second VO, freeze-framed at 0:00 before the first cue appears. You will post a face with empty lower-third.

    The AI slideshow maker can sequence stills. It does not magically caption a JPEG. Letter the cover on purpose.

    Letter one line into the cover still

    Overlay, not ASR.

    Procedure:

    1. Get the still. If it must match a video later in the carousel, extract a frame from that keeper. Do not text-to-image a face that is not in the later slides.
    2. Write the line. Two lines maximum. Near 32–42 characters per line for Latin script is a readability guideline, not a format law. You have no CPS clock because the still sits until someone swipes. They can read slowly. They cannot read a paragraph.
    3. Place the type in the safe area for that surface. Carousel chrome is not identical to Reels: username, page dots, sometimes a shop sticker. Keep the line out of the bottom and out of the dots.
    4. Burn the type into the image (image text overlay, or a designed slide). This is text overlay logic: you authored it. Nobody had to say it.
    5. Do not run add captions to video on a still. There is no speech to transcribe. You will pay a caption pass for silence or for a hiss.
    6. Export the cover at the slot size. Do not screenshot a timeline.

    Checkable cover test:

    • Pause on slide 1 as it will appear in the grid and in the swipe. Can you read the line at phone size.
    • Cover the rest of the carousel with your hand. Does slide 1 still make a claim. If it only makes sense after slide 3, it is not a cover.

    Logo plus “New drop” with no product is a brand sting. Put the pack or the result in the still if that is what you sell.

    Do not attach a timed track to a JPEG

    Sidecars need a player. A carousel still is not one.

    What will not letter slide 1:

    • An SRT you exported from the talking-head you plan to put on slide 5.
    • Platform-native captions on a video you did not upload as slide 1.
    • A DYNAMIC preset. There is no token to animate on a still.

    If slide 1 is a video tile:

    1. Picture lock that short file.
    2. If someone speaks, burn captions on that file. Language codes, BASIC or DYNAMIC, same as any other clip.
    3. Still freeze frame one and confirm the first caption cue is already on screen at 0:00, or add overlay so the poster-state has type before ASR would have fired.
    4. Do not rely on the viewer enabling CC. They are swiping stills.

    Later slides:

    • Silent graphics: overlay on each still. Repeat the claim if they will be screenshotted out of order.
    • Talking-head video: real caption burn on that tile only. The cover’s line and this tile’s transcript should not contradict.
    • Pack shots: forced narrative rules if you localise; the SKU wordmark may stay.

    A slideshow export that becomes one MP4 is a video, not a carousel. Once it is a single file in Reels, you are back to timed captions on speech and overlay on authored headlines. Do not confuse turn-a-slideshow-into-a-video with an IG swipe. Different object, different clock.

    QC the swipe, then the still in the grid

    Two surfaces.

    Checklist:

    1. Grid / thumbnail. The cover still at tiny size. Line still readable. Face or pack recognizable. This is why you did not put 40 words on slide 1.
    2. Open state, slide 1. Full size, dots visible, username visible. Line still in safe area.
    3. Swipe through. Each slide’s type is authored or transcribed on purpose. No empty talking-head. No SRT you forgot to burn.
    4. Alt text on the cover still (and on other stills if the platform allows). Alt describes the picture. It is not the caption track. Different page, still do it.
    5. Mismatch. If slide 1 says “20% off” and slide 4’s VO says “30% off,” you have two ads. Fix the still or the speech.

    If you will also post the same story as a Reel, that Reel is a new file: timed captions, first-second hook, no carousel dots. Do not upload the cover JPEG as a video.

    Archive the cover still next to the carousel project. A later price change is a new lettering pass on the still, not a recaption of a track that never existed.

    Letter the cover as a poster in Versely. Bring a still (or a frame from the keeper) and one line. Do not bring an SRT.

    FAQ

    Can I use add-captions-to-video on a carousel cover?

    Only if that cover is actually a video with speech. A JPEG has nothing to transcribe. Overlay the line.

    Should the cover line be the same as the Reel hook?

    If it is the same campaign, yes — same claim. The object is still a still. Do not ship a timed track to do a poster’s job.

    Do later video slides inherit the cover’s type?

    No. Each tile is its own file. Burn or overlay on that tile. Keep the claim consistent by hand.