Auto-caption a video

One prompt adds styled captions timed to speech. Pick from six caption skins, restyle instantly without retranscribing, and fix any line by hand.

Captioning is a one-prompt job in Pluged AI. The agent transcribes your timeline audio and lays styled captions timed to speech — a 5-minute video is captioned in the time it takes to make coffee, and switching styles afterwards is instant because it never re-transcribes.

What you'll make

A fully captioned video in one of six caption skins, with any transcription mistakes fixed by hand, exported with captions burned in.

Prerequisites

A video with spoken audio on the timeline (drag it from the Media panel onto the main track)

Step 1: One prompt

Open the Agent panel and type:

"Add captions to this video, TikTok Bold style"

Or let it choose the style for the platform:

"Caption this for a LinkedIn audience"

The agent extracts the audio, transcribes it, and places caption cues on a caption track — each one a real timeline element you can click, drag, and edit.

How the timing works (honest version): transcription is segment-level, not word-stamped. The agent knows when each spoken segment starts and ends, then paces the caption cues inside each segment by syllable weight — longer words get proportionally more screen time. The result reads as captions timed to speech; an individual cue can lead or lag the voice by a fraction of a second. If you need frame-exact timing, import a subtitle file (see Step 5).

Step 2: Pick a skin

Six caption skins ship with the app. Each sets the typography, background box, and screen position — nothing else (no strokes, no glow, no animations):

Skin	Look	Best for
TikTok Bold	Large bold white text on a dark rounded backing	TikTok, Reels, Shorts
Minimal Clean	Plain white text, no background box	Interviews, tutorials, clean brands
Boxed Contrast	Bold text in a near-solid black box	Busy or bright footage, accessibility
Editorial Highlight	Serif type on a left-aligned dark rail	Documentary, premium editorial
Neon Pop	Bright text on a deep blue box	Gaming, hype edits, music
Documentary Lower	Small quiet serif low in the frame	Long-form docs, staying out of faces

Name the skin in your prompt, or describe the vibe ("something quiet that stays out of my face") and the agent picks.

Step 3: Restyle without retranscribing

Changing your mind costs nothing:

"Switch the captions to Boxed Contrast"

This restyles the existing caption elements in place — the transcript is cached, so no re-transcription, no waiting, and any text corrections you've made are preserved. Try two or three skins against your footage; it's the cheapest taste decision in the app.

Step 4: Fix mistakes by hand

Transcription is good, not perfect — names, jargon, and mumbled lines miss. Captions are ordinary timeline elements:

Wrong word: double-click the caption in the preview or timeline, retype it, press Enter
Timing off: click the caption clip and drag it left/right; drag its edges to lengthen or shorten
Cue splits awkwardly: ask the agent — "merge the captions at 0:42" or "split that long caption into two"
Custom look: select a caption and use the Properties panel to override font, size, color, or position for that cue

Step 5: Already have subtitles?

If you have an .srt or .ass file (from a human transcriber or another tool), import it from the Captions tab instead of generating — you get its exact timing and text, and can still restyle with any skin. Full details in the captions deep-dive.

Step 6: Export

Export → MP4 (H.264) → High. Captions are burned into the rendered video — what you see in the preview is what viewers get.

Make it yours

"Caption this in Spanish" — the agent passes a language to transcription; leave it out and the language is auto-detected

"Caption it, then delete the caption cues over the intro — I want the first 5 seconds clean"

"Caption it, then move the captions higher — they clash with my lower-third"

The cached transcript also unlocks bigger asks in the same session: "now cut a 30-second highlight of the strongest argument" reuses the transcript to pick moments. See transcript highlights.

Troubleshooting

Captions drift from the voice in one section. Segment-level timing paces cues within each spoken segment, so a fast or mumbled stretch can drift. Drag those caption clips into place, or import an SRT for exact timing.
No captions appeared. The clip needs audible speech — check the clip isn't muted (Properties → Audio) and the volume isn't at zero, then ask again.
Wrong language detected. Say "re-caption this — the audio is French." Re-generating replaces the caption track.
Style change didn't apply. Skins restyle existing captions — if there are none yet, the agent needs to caption first. "Caption this with Neon Pop" does both in one go.
Captions collide with other text. Ask "move the captions clear of the title" — the agent's own quality checks flag overlapping text, but manual titles added afterwards can reintroduce it.