Make a TikTok from raw footage

One prompt turns long-form footage into a vertical TikTok with a hook, punchy captions, and music. Then refine it conversationally.

Raw 5-minute talking-head video to a posted TikTok in about 10 minutes. You give the agent one prompt; it finds your best line, cuts the reel around it, reframes vertical, captions, and mixes — then checks its own work. Your job is the last 20%: taste calls, made in plain language.

What you'll make

A 30-second vertical clip that opens on your strongest actual line, with TikTok Bold captions timed to speech, a music bed under the whole cut, and pacing tight enough to survive the feed.

Prerequisites

Any talking footage: interview, podcast, vlog, lecture, pitch
Pluged AI open with the footage imported (drag the file into the editor)

Step 1: One strong prompt

Open the Agent panel and type:

"Turn this into a 30-second TikTok: hook from my best line, punchy captions, vertical reframe, music."

That single prompt triggers the full edit. Resist the urge to split it into steps — the agent plans better when it owns the whole outcome.

Step 2: What the agent does

Watch the panel — each stage shows up as it happens:

Looks at the footage first. It analyzes the video and transcript before cutting anything; you can't direct footage you haven't watched, and neither can it.
Posts a storyboard card above the chat: thesis, 30-second target, and beats mapped to moments in your footage. The hook beat quotes a real line from your transcript — the agent finds hooks, it doesn't invent them. (See Storyboard.)
Cuts the reel: keeps 3–6 whole-thought segments that serve the thesis, opens on the strongest line (often mid-sentence), cuts on breath and sentence boundaries — no machine-gun fragments.
Reframes vertical with the subject framing driving the crop, and adds motivated movement like zoom punches at cut points.
Captions timed to speech in the TikTok Bold skin, kept inside the vertical safe area, never colliding with other text.
Lays a music bed under the entire cut, ducked beneath your voice, ending when the video ends.
Checks its own work: critique, verify, then an AI watch-through of the rendered cut — and fixes what fails before reporting done. (See How the agent checks its work.)

Every change lands on your real timeline and is undoable with Cmd/Ctrl+Z.

Step 3: Refine conversationally

Play the result, then direct like an editor giving notes:

Different caption look:

"Restyle the captions with Boxed Contrast"

Restyling swaps the skin without retranscribing — it's instant. The six skins are TikTok Bold, Minimal Clean, Boxed Contrast, Editorial Highlight, Neon Pop, and Documentary Lower (see caption skins).

Tighter middle:

"The middle drags — tighten it. Cut anything that restates the point."

Swap the hook:

"Open on the line about losing the first client instead"

Point at a real line from your footage. The agent grounds every hook in the transcript — if you ask for a generic hook like "You won't believe this", it will push back and offer your actual best lines instead, because stock-phrase hooks are against its rules.

More energy:

"Add zoom punches on the two biggest lines and a whoosh on the first cut"

Step 4: Export

Click Export (top-right), choose MP4 (H.264), High quality, and export. Captions are burned in. Upload straight to TikTok, Reels, or Shorts.

Make it yours

Template route: "Apply the TikTok Highlight template" combines the Viral Cut style pack and TikTok Bold captions in one move. See creator templates.
Longer format: "Make it a 60-second podcast clip with Documentary Lower captions" — or follow podcast to clip.
Product angle: selling something on camera? Try the UGC ad tutorial.
Visual depth: "Add stock B-roll where I mention the warehouse" — the agent pulls stock into the bin and overlays it at the exact transcript moment. See B-roll and overlay cards.

Troubleshooting

Captions look slightly off from the speech. Transcription is segment-level, not word-stamped — cues are paced within each segment by syllable weight, so an individual caption can lead or lag by a beat. Say "nudge the captions at 0:12 later" or drag the caption clip on the timeline. For frame-exact subtitles, import an SRT/ASS file (captions deep-dive).
Too long. Say "trim to exactly 30 seconds" — the agent re-cuts to hit the number rather than speeding everything up.
The crop misses your face. Say "reframe with face focus" — the reframe is heuristic, guided by the footage analysis, and a focus hint fixes most misses. See smart reframe.
The hook feels weak. Don't ask for a punchier phrase — ask for a different moment: "what are my three strongest lines? Open on the best one."