Guides

    Level Five is authored overlay, not transcribed captions

    add-text-overlay burns the finish spec you wrote; auto-captions transcribe speech you do not have on a silent sanding plate.

    Versely Team6 min read

    Add a text overlay burns the finish spec you wrote; auto-captions transcribe speech you do not have on a silent sanding plate.

    Level Five is a string a crew already decided: skim, sand, light, the job that hides fasteners and joints under a critical light. It is not a word someone said into a mic while the pole sander moved. A mute plate of a wall does not become "Level 5" because you ran the caption generator. Type the spec. Burn it. Do not ask a transcriber to invent a finish grade.

    /for/contractors is the GC page: demo-day time-lapses, behind-the-wall problems, a final walkthrough. This post is the drywall finish callout on a sanding still. Process transparency is a different job. The spec on the plate is this one.

    The spec is a line you already know

    "Level 5 skim," "ready for gloss," "joints knocked down" — those are claims you will stand behind with the homeowner. They belong in a string you typed, proofread, and burned after picture lock. They do not belong in a video model's bezel, and they do not belong in a transcript of silence.

    Add a text overlay to a video is add_video_captions: one fixed line, whole clip, position top / center / bottom. Font size, colour, and an optional background block are controllable. It does not listen. Silence, a sander, a shop vac — none of that changes the string.

    That is the correct object for a finish grade. The wall does not speak. You do.

    Do not prompt the generate to "letter LEVEL FIVE on the wall." Prompted type is a misspelled grade you cannot edit against the spec sheet. Keep the plate clean. Burn the real line. Crop to 9:16 before you overlay; top and bottom sit in a different safe zone after a reframe.

    If the spec has to change at timestamps — "hang," then "tape," then "Level 5" — that is timed overlays, not one persistent badge, and still not auto-captions. You write every line. You set start_sec and end_sec. Nothing gets heard.

    Auto-captions need speech. A sanding plate has none.

    The AI caption generator does both jobs the word "captions" covers, and keeps them separate. Subtitles are transcribed from speech and burned in a chosen style. Overlays are text you wrote. The tool page is honest about that split. Using the transcribe half on a mute wall is how you pay for a blank track.

    Add captions to a video is VEED: speech in, styled timed text out, 165 language codes, BASIC presets at 1×, DYNAMIC at 2×. Accuracy depends on audio clarity. A pole sander is not a voiceover. Overlapping noise degrades a transcript you did not need.

    If a superintendent later records a walkthrough and says "this is Level Five," then captions are the right second pass on that file. The sanding plate you posted at noon is still overlay. Do not caption the mute cut hoping the grade appears. Do not overlay a paraphrase onto a talking walkthrough and call it captions. You will misquote the take.

    The caption generator will not rescue a plate that never had a mic. Preview a style on speech you actually recorded. A DYNAMIC look on a silent file is a double-rate burn of nothing.

    The plate is the wall, not a talking taper

    Drywall content that holds is a still of the surface under raking light, then motion that does not invent a second room. A talking taper explaining "what Level Five means for your resale" is a different generate, and it needs a real named person if it tells a homeowner what to do. A constructed tradesperson is a presenter you did not hire.

    Keep lips out of the sanding plate. Score later if you need a bed. The overlay is the spec. The wall is the proof. Mixing them in one text-to-video paragraph is how you get a cousin kitchen and a misspelled grade in the plaster.

    Nearby contractor content — contractors — can still be demo-to-walkthrough. Do not rewrite that page as a finish-grade overlay. A GC reel that shows framing is not a Level Five callout. A drywall reel that typesets the grade is not a full-gut documentary. Name the object.

    If you needed before/after of the same wall, that is first-last: textured or jointed as frame A, skimmed as frame B, same standing spot. Overlay still happens after interpolation, on the file you keep. Captions still need someone to have spoken.

    Lock the wall. Then print the grade.

    People overlay early because the plate looks empty. Empty is information. It means you can still see whether the light, the joints, and the motion are right. Type hides that. Type also bakes a finish claim into pixels you may recut.

    Generate or film the plate. Decide the crop. Decide whether a later walkthrough will carry speech (captions then, on that file). Then burn "Level 5" or the line you actually use. Offers change. Spec language changes. Copy lock can be later than picture lock. If either is open, do not burn.

    One string, whole duration, your words. That is overlay. Several authored lines on timestamps is timed overlay. Speech on a mic is captions. A silent sanding plate is the first of those three. Running the other two on it is a wasted render.

    FAQ

    Will the AI caption generator read "Level Five" off the wall?

    No. Transcription hears speech. Overlay prints a line you typed. A mute sanding plate has nothing to caption. Type the spec.

    Can the overlay change from "tape" to "Level 5" mid-clip?

    Not on add-text-overlay. One string, whole file. Timed overlays are the other named job, still authored, still not a transcript.

    Should I record a voiceover so captions can typeset the grade?

    Only if you wanted a voice. The grade does not need a mouth. Overlay the string on the silent plate. Captions are for speech you actually recorded.

    Is this the same as the contractors audience page?

    No. /for/contractors is GC process — demo, framing, walkthrough. This page is the finish spec as type on a drywall plate. Do not swap them.