clipify · Smart Long-to-Short Video Clipper

Automatically extracts highlight moments from long videos and cuts them into standalone short clips: Whisper word-level transcription locates the funny parts, each candidate is trimmed, optionally reformatted 16:9 → 9:16 (face-tracking pan or split-screen), and word-by-word captions are burned in.

Output is 10–25 second clips with highlighted captions — built for turning English interviews, podcasts, and stream recordings into TikTok/Reels-ready shorts.

Example invocation: "Cut this interview into 3 vertical shorts with word-by-word captions."

Full brief

Positioning

clipify is a long-video-to-short-clip pipeline: find the funny moments, cut the clips, reformat to vertical, burn in captions — fully scripted. It is purpose-tuned for English talking-head content.

Core capabilities

• Word-level Whisper transcription: tiny.en for English (~10× faster than small.en), base for non-English; word timestamps included.
• Funny-moment detection: 3–5 candidate clips (target 10–25s) selected on punchline, quotable one-liner, awkward-pause, self-roast, and audio-peak signals — each proposed with start/end, why-it's-funny, and a suggested title, cut only after user confirmation.
• Reformatting: 16:9 → 9:16; two-talker clips get either face-tracking pan (hard cuts following the speaker) or stacked split-screen (speaker on top); single-talker clips get a center crop.
• Caption burn-in: three presets — opus (big bold white, active-word highlight), karaoke (4-word chunks), minimal (clean type); custom ASS from a user reference is also possible.
• Delivery & iteration: outputs land in the source folder's clipify_out/; style, reformat mode, and caption timing can be iterated.

Workflow

1. Extract audio, transcribe the full source with Whisper, propose candidates for confirmation
2. Trim each chosen clip precisely
3. Confirm target aspect ratio (9:16 / 16:9 / 1:1)
4. Reformat: locate face ROIs → build speaker timeline → render pan or split-screen
5. Burn word-by-word captions
6. Deliver to clipify_out/ with per-clip duration, joke note, and path

Inputs & outputs

Output: captioned short-video files under the source directory's clipify_out/.

Boundaries with adjacent skills

• video-highlights: more general (Chinese/live content fine), steadier static vertical reformat; clipify specializes in English talking-head funny-moment detection plus dynamic face-tracking pan.
• video-editing: general editing primitives; no intelligent highlight-finding, face tracking, or caption burn-in.

Fit

• Cutting English podcasts/interviews into TikTok/Reels shorts
• Highlight edits from stream recordings and long videos
• Talking-head shorts that need word-by-word highlighted captions

Before you start

• Local ffmpeg, ffprobe, and whisper required; no API key.
• English content uses the fast tiny.en model; non-English uses base.
• The full source is transcribed only when hunting for funny moments; re-running Whisper on the trimmed clip gives more accurate caption timestamps.

clipify is part of the Aiglade Skill library. Invoke it from the Aiglade chat box in plain language.

← All 155 skills