Caption AI means captions, not video
People searching Caption AI want auto captions on a video. That is a burn-in job, not a text-to-video model.
People searching "caption ai" want auto captions: speech on a video, timed, usually burned into the picture so a muted feed can read it. They do not want a new video model. They do not want a caption-themed generate. Caption AI, as a query, is a captions job.
Search Console still shows "caption ai" (46 impressions, position 47, seven days to 4 September 2026) in the same weak-CTR cluster as "ai caption generator" and "auto captions." The tool that matches the intent is the AI caption generator. This page exists so that query stops landing on a generic AI explainer.
The query is a burn-in, not a brand essay
"Caption AI" is a product name in the market and a generic string in Search Console. Either way, the job the searcher brings is: I have a clip, I need words on it, I do not want to type cues by hand.
That job has three clocks, and collapsing them is how "the captions are bad" becomes unactionable.
- Words. Speech-to-text. Product names and numbers fail first. Read the transcript before you burn.
- Time. Caption timing is when each line appears. A uniformly late track is an offset (you trimmed an intro after captioning). A track that drifts is often a frame-rate mismatch. Burn-in freezes the clock into pixels.
- Look. Preset, font, safe area. The safe area is the part of a 9:16 frame the platform UI does not cover. A line on the true bottom of the export is invisible in the Reel.
The Versely path is transcribe and burn: speech becomes timed lines, a style is applied, the result is composited onto the frames. Burned-in captions cannot be switched off, which is why they won silent autoplay. A sidecar .srt is a different object for YouTube long-form and course players. Mixing the two is how a TikTok ships with no words on screen and a perfectly good caption file in the project folder.
Closed captions are the sidecar switch. SDH is speech plus non-speech. Auto captions are not SDH unless someone authors [music] and speaker IDs. If you searched "caption ai" because a client said "make it accessible," ask which file they mean. The generator will give you a speech burn. It will not invent a broadcast 608 packet.
Text overlay is not captions. Overlay is copy you wrote — a hook, a price, a CTA — and nobody has to have said it. Auto captions transcribe what was said. Short-form usually needs both: the spoken line burned in, and one authored line that states the claim in the first second.
The generator is not text-to-video. It will not invent a talking head from "caption this script." If you do not have a video yet, you are in the AI video generator or image-to-video, then you caption.
What to do in the tool, in order
Lock the cut. Burned-in captions are pixels. A later trim is a recaption, not a ripple.
Transcribe. Pick the language that is actually spoken. The captioner does not translate. A Spanish VO with the language left on English is a bad English transcript.
Read the names. Versely will not know your SKU. Fix the transcript.
Split long clauses. Latin-script lines are commonly kept near 42 characters. A 90-character cue is a paragraph on a face.
Preview the style on a few seconds of this clip. The editor's 480p preview pass is free and carries a short per-user cooldown. Contrast against this background. Classic white-on-stroke still wins most talking-heads. Animated families bill at 2× the caption render for the same transcription work; use them when the caption is part of the performance.
Watch on mute, on a phone. If you cannot finish the line, reading speed is too high. If the line sits in the chrome, move it.
Export the MP4. Do not also attach a sidecar and assume Instagram will draw it.
Add captions to video is the editing-task map: 165 language codes, 1× basic / 2× dynamic presets. Caption styles is the catalogue. Captions for silent autoplay is why the burn exists. What captions cost across a month of clips is the 1× / 2× cadence. Use those when you already know you are captioning. This page is only the disambiguation: caption AI means that stack.
A script timer is not this job. The script timing estimator divides word count by WPM in the browser and spends nothing. It does not transcribe and it does not burn. If you searched "caption ai" with a script and no video, you still need a picture row first — or you only needed duration, and the free tool is enough.
What the query is not
Not a video model named Caption. There is no reason to run Veo because the search contained "AI." Veo 3.1 writes clips. Captions are a pass after a clip exists.
Not a substitute for SDH or forced narrative. Those are authored files. See CC vs subtitles vs forced narrative.
Not unlimited free magic. Full caption burns cost credits. The 480p style preview is the free look. In-browser converters and checkers on free-tools are actually free and do not caption the video. What free AI caption generator means is that honesty in full.
Not translation. Dub, then caption in the new language, or caption the original and ship a second-language sidecar for long-form. One auto-caption pass will not do both.
Not "Captioning the prompt." Putting a slogan in a text-to-video prompt paints letters into generated pixels, badly. Overlay and burn-in are typesetting. Models that "render text in the video" are a different, unreliable job.
Article 50 (applies 2 August 2026) wants a visible or audible mark on certain realistic synthetic media for EU audiences. That mark is not a caption cue. Put it in the first frame or the first second of audio, then caption the speech as you would any other cut.
If a vendor named Captions AI is in the brief, you are in a competitive comparison. The intent of the query on this site is still auto captions. Point the work at the generator. Do not write a novel about a logo.
The test
Do you have a video with speech, and do you need the words on screen for a mute feed? AI caption generator.
Do you have a video and a player with a CC button? Transcript plus sidecar, maybe SDH. Burn-in is optional.
Do you have no video? You searched the wrong noun. Generate or shoot, then caption.
Do you have a script and a clock question? Script timer, not caption AI.
The 46 impressions at position 47 are people using a product-shaped phrase for a task. Serve the task.
Caption the locked cut in the caption generator. Sign up when the file already has speech and the next click should be words on picture, not another model.
FAQ
Is Caption AI a video generator?
Not as this query is used. The searcher wants captions on an existing clip. Text-to-video is a different tool. Caption after the clip exists.
Will auto captions appear on Instagram if I only export an SRT?
No. Reels need the words in the picture. Burn in. A sidecar is for players that draw a track.
Are auto captions the same as SDH?
No. Auto captions transcribe speech. SDH adds speaker IDs and non-speech such as [music]. Author those, or you shipped subtitles under a captions name.