Skip to content
SkillScript-led1,182 tokens in use

short-form-pipeline

Cut TikTok, Reels and Shorts clips from long YouTube videos: download, transcribe, pick moments with Claude, crop vertical and burn in captions.

3,862 stars on the repository MIT Licenseericosiu/ai-marketing-skills · short-form-pipeline
npx skills add ericosiu/ai-marketing-skills -s short-form-pipeline
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing short-form-pipeline costs

27tokens idle
1,182tokens in use
1,273tokens, up to

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Up to adds the files its instructions link. Counts are approximate, made offline with cl100k_base.

Model dependence Script-led

Script-ledMixedModel-led

Most of this skill is scripts, so it behaves much the same whichever model runs it.

This describes the skill's structure, how much of it is scripts. It does not say how well it works. Scripts: scripts/clip_sender.py, scripts/shortform_pipeline.py, scripts/video_clipper.py.

Touches

shelldetectedwrites filesdetectedcredentialsdetected

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Excellent
  • Frontmatter8/15
  • Structure23/25
  • Examples12/15
  • Output15/15
  • Size15/15
  • Resources12/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Good
  • Runs shell commands−5

    does this on the machine, and the frontmatter does not say so

    SKILL.md, line 12
  • Reads credentials−15

    does this on the machine, and the frontmatter does not say so

    scripts/shortform_pipeline.py, line 49
  • Writes files−8

    does this on the machine, and the frontmatter does not say so

    scripts/shortform_pipeline.py, line 69

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Good
  • Development activity88
  • Maintenance44
  • Community56
  • Maturity28
  • Momentum40

Short-Form Pipeline takes a YouTube URL and produces vertical clips ready to upload. It downloads the video and captions with yt-dlp, transcribes with Whisper, asks Claude for two to five moments of 30 to 60 seconds with a hook score of at least 7 out of 10, and checks that each clip ends on a complete thought.

FFmpeg then cuts each clip, crops 16:9 to 9:16 using a layout hint, and burns in word-highlighted captions. A second script clips without Claude, using heuristic scoring.

What You Get

  • Vertical 1080 by 1920 clips in H.264 and AAC with word-highlighted captions
  • Claude-selected segments with start and end time, title, hook sentence, payoff and layout hint
  • A second Claude pass that checks each clip ends on a complete thought
  • Layout-aware cropping, with face detection for talking heads when mediapipe and opencv are installed
  • Tuning points: voice patterns, hook strength minimum, target duration, zoom and face position
  • A troubleshooting list for FFmpeg, resolution, caption sync and upload problems

Common Use Cases

  • Turning a podcast episode into three short clips for TikTok, Reels and Shorts
  • Producing captioned vertical clips from a webinar or screen share recording
  • Running a no-API clipper with heuristic scoring for a quick pass
  • Batch-processing a channel’s videos while skipping ones already processed

What Makes It Unique

It covers the full path from URL to captioned vertical file in one script, and its crop step adapts to the video layout, with a different treatment for talking heads, screen shares, split screens and gallery views. A verification pass checks clip endings, and clips over 90 seconds are trimmed automatically.

Under The Hood

The scripts run as Python commands and shell out to yt-dlp, ffprobe and FFmpeg. They download videos and captions from YouTube, send transcript text to the Anthropic API using your key from the environment, and write video, caption and a processed-URL list into a workspace folder. A preamble at the start runs the repository’s version check and telemetry setup scripts, which log usage locally and make remote reporting opt-in.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers