No filming:An AI avatar or clone yourself
    Versely
    Guides

    Merge clips into one UGC file

    Generate hook, demo, and CTA as three keepers, then merge. Do not prompt one take for all three.

    Versely Team••7 min read

    Hook, demo, and CTA are three generates. Merge them after each one is a keeper. One prompt that asks for a greeting, a pour, and a URL will jump, invent a cousin bottle, or swallow the offer. Merge videos (merge_movie_scenes) takes a list of clip URLs and returns one file, with a shared transition and a shared audio-handling setting. It does not rescue a miss by putting it between two others.

    This is the three-clip UGC stitch. It is not a hook-pack and it is not a movie-scene essay.

    Three generates, three jobs

    Write three jobs before you open a model. If two of them share a verb, you do not have three jobs.

    Beat Job What you lock Typical length
    Hook First-second claim Face, result, or mess — plus a burned line 1–2s
    Demo Proof on the real SKU Image-to-video from the approved still, or filmed hands 3–6s
    CTA What to do next Overlay URL, offer, or next step. Face optional 1–2s

    Procedure on paper:

    1. Write the hook sentence. Mute-readable. Not a logo.
    2. Write the demo verb. One verb. “Unscrews” or “pours” or “clicks.” Not “unscrews, pours, smiles, and shows the URL.”
    3. Write the CTA as overlay copy. If it is spoken, it still gets a burn.
    4. Lock one product still for the demo. Crop to 9:16 if that is the ship ratio.
    5. Name words you will not claim.
    6. Name the model family per beat. They do not have to match, but grain will not match if they do not. Plan for that.

    Generate rules:

    • Hook: picture that carries the claim. Silent Kling 3 Turbo at 12 credits is enough when TTS or overlay will speak. Do not put the catalog pack in this prompt unless the pack is the proof.
    • Demo: attach the still. Image-to-video. Text-to-video will invent a cousin SKU.
    • CTA: a hold, an end card, or a last look at the pack with type. Do not generate a new face that never appeared in the hook.

    The UGC video generator and the AI video generator are the doors. Merge is not a generate. Do not ask merge to “add a CTA.” Type is a later burn.

    If you already have a single-take UGC you like, you are not on this page. This page is for the slate that was always three beats.

    Lock each clip before you stitch

    A merge of maybes is a longer maybe with a crossfade.

    Keeper checks, per file:

    1. Hook: pause frame one. Claim readable muted. No logo hold.
    2. Demo: pack matches the still. Label letters. Hands travel through the verb. No teleport.
    3. CTA: overlay copy fits two lines. Offer matches the spoken line if any.
    4. Ratio matches across the three. Do not merge 16:9 demo into 9:16 hook and hope.
    5. Duration is already close to the beat. Trim handles on each child first if the generate over-ran. Trim a video per clip, then merge. Merge is a shared transition, not a per-clip trimmer. If you need per-item start_sec/end_sec on mixed media, that is compose, not this tool.
    6. Audio intent is named: keep each clip’s sound, drop it, or replace later. Native audio on the demo plus silence on the hook is a jump in the mix. Decide now.

    Reject a clip that fails its job. Regenerating scene two after a “nice” crossfade is how temp films become the master. Keep the sources. The merge is a child.

    Do not caption the three files separately unless you like retiming three burns into one timeline. Caption the stitch.

    Do not upscale the three files separately unless you will also upscale the child, or unless only one plate is the 4K hero. Grain mismatch across a merge is obvious on a phone.

    Merge in order: hook, then demo, then CTA

    video_urls is an ordered list. Wrong order is a different ad.

    Procedure:

    1. Put URLs in beat order: hook, demo, CTA. Not “the best looking file first.”
    2. Pick transition_type and transition_duration once for the whole join. A hard cut is the UGC default. A long crossfade is a brand film. Do not dissolve a talking head into a pour unless you meant to hide a bad splice.
    3. Set audio_handling. If the hook is silent and the demo has native audio, dropping clip audio and laying TTS later is cleaner than a mix that pops at the seam.
    4. Run the merge. estimate_cost before a long list. Three short clips are the point of this page; a twenty-clip movie is still this tool, not this brief.
    5. Watch the child at 1×, muted, then with sound.
    6. Inspect each seam: three frames before, three after. If a seam pops, you needed a trim on the incoming clip or a B-roll cover, not a fancier transition.

    Checkable seam tests:

    • Hook → demo: the burned hook line has either cleared or still tells the truth over the pack.
    • Demo → CTA: the verb finished. Do not cut off the pour to hit a time cap. Shorten the hook instead.
    • Last frame: not a black hold. Looping feeds replay. A logo card as the last frame kills the loop.

    If you need picture-in-picture on the demo (face over pack), composite that before the merge, or compose instead of merge. Merge concatenates whole clips. It does not corner-mount a talking head.

    If one clip is a still (end card), merge is the wrong tool. Compose accepts images and video. This page assumes three videos.

    Caption the joined file once

    One timeline, one transcript, one burn.

    Finish order:

    1. Picture lock the child. If you replace demo, you remake the merge. That is expected. Keep the three parents.
    2. Audio lock. TTS or native, not both fighting. Replace video audio if the stitch should be one VO bed.
    3. Transcribe and burn on the child. Add captions to video: 165 language codes, BASIC 1× or DYNAMIC 2×. Preview on the first seconds of the joined file. The hook’s first line is the mute test.
    4. CTA type is overlay if nobody said it. Do not hope auto-captions will typeset a URL that was never spoken.
    5. Recheck safe area on the 9:16 child. Right rail and bottom chrome eat CTAs you parked too low.
    6. Export the child. Archive the three parents and the merge settings.

    If you localise later, keep the silent or picture-locked child and swap audio, then recaption. That reuse is why the demo was image-to-video from a still, not a baked English native-audio hero you cannot recast.

    Stitch the three keepers in Versely. Bring three URLs in order. Do not bring one paragraph prompt.

    FAQ

    Can I generate hook, demo, and CTA as one 15-second take?

    You can. You will usually get a jump, a cousin pack, or a swallowed CTA. Three keepers plus a merge is the procedure on this page.

    Do I have to use a crossfade?

    No. Hard cuts are the default for UGC. Transition is a shared setting for the whole merge, not a per-seam design.

    Should I caption each clip before merging?

    No. Caption the joined file. A later remake of scene two would orphan the first two burns anyway.