video-essay
Video-essay and documentary-style YouTube editor: script-first, voice-over-led edits with archival footage, B-roll, quote cards, maps, timelines, charts and chapter cards, cut at the pace of an argument (4-8 s shots), filmic grade, sourced on-screen attribution and YouTube chapte
Install
npx skills add https://github.com/Viprasol-Tech/claude-video-editing-skills/tree/main/plugins/video-studio/skills/video-essay
claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install viprasol-tech-claude-video-editing-skills@llmmart
git clone https://github.com/Viprasol-Tech/claude-video-editing-skills.git
The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole viprasol-tech/claude-video-editing-skills collection as a plugin from our marketplace. Git is the plain clone.
Skill manifest
Video Essay
An argument told in pictures: VO leads, every visual is evidence, chapters keep the viewer oriented.
- Best for: analysis, history, film/media essays, science and culture deep dives
- Inputs: the recorded VO (or on-camera narration), archival clips/stills and B-roll in a folder, sources list, optional music
- Default skin:
essay-mode· alternativesdoc-real,paper-cut,swiss-poster(use--no-mattewithout a presenter) - Formats: 16:9 1920x1080 master, chapters text,
captions.srt; optional 9:16 teaser - Duration target: 10-40 min; genre preset
video-essay
Before you start
- Needs video-studio-core in the same skills folder.
CORE=video-studio-core,SKILLS= the folder holding both. - First time:
python CORE/scripts/check_env.py. Preset:python CORE/scripts/editor.py genre video-essay. - The
--genrepreset sets autocut's pause cut (0.6 s) and run.py respects it; pass--max-gaponly to override. - Fair use: quote clips briefly and comment on them; credit sources on screen; never present someone's claim out of context. Keep a
sources.md.
How a pro cuts this
| beat | % | what happens |
|---|---|---|
| cold open | 0-5 | the most intriguing image or question, before any title |
| thesis | 5-10 | what this video argues, in one sentence |
| chapters | 10-85 | 3-6 chapters, each: claim -> evidence (clip/quote/chart) -> implication |
| turn / counter-argument | 85-92 | the strongest objection, answered |
| conclusion | 92-100 | thesis restated as an insight; end on an image, not "thanks for watching" |
- Pacing: ASL 4-8 s; a visual change at least every 6 s. Let a key archival moment play 8-12 s with its own sound. Slow down for the emotional chapter, quicken in montage sections.
- Cut rules: cut the VO first (radio edit), then picture. Visuals land 0.1 s before the words that need them. Pauses between chapters are allowed (0.6-1 s).
- J/L cuts: archival sound up under the end of a VO line (J) so the clip "speaks"; L-cut the VO over the next visual.
- Transitions: cut; dip to black or a chapter card between chapters; dissolves for time passing.
- Music/SFX: score bed 24-28 LU under VO, changes mood per chapter; drop it for the revelation. SFX almost none (page-turn on cards).
- Colour: unify archive and new footage:
documentary-natural0.5 orfuji-eterna0.6; archive can stay as is withsepiafor flashbacks. - Captions: sidecar SRT; on-screen text only for quotes, names, dates and sources.
- Client judges: clarity of the argument, research visuals, accuracy, chapters that make sense.
- Common mistakes: visuals that decorate instead of prove, unattributed quotes, music wall-to-wall, a 2-minute intro, uncredited clips.
Make the edit
With a narration video (on-camera or a VO muxed over a still - see faceless-explainer for the TTS route):
python CORE/scripts/run.py narration.mp4 --style SKILLS/essay-mode --genre video-essay \
--smart-cuts --retakes --model small --prompt "names, places" --work essay --stop-after plan
python CORE/scripts/story.py outline --work essay --genre video-essay # paper edit: topic shifts = chapter candidates
python CORE/scripts/broll.py --work essay --library archive/ --frequency high --merge essay/plan.json
python CORE/scripts/editor.py review essay/plan.json --genre video-essay --fix
python CORE/scripts/render.py essay/plan.json --frames 2,60,180,400
python CORE/scripts/render.py essay/plan.json
python CORE/scripts/finish.py essay/plan.json --platform youtube
broll.py matches archive file names to spoken words; name files for what they show (berlin-wall-1989.mp4, newspaper-headline.jpg) or add NAME.txt keyword files. Place the important evidence shots by hand in cutaways ({"src": "../archive/speech.mp4", "start": 84.2, "end": 95.0, "srcStart": 12}).
Refine the plan
- Chapter cards:
{"type": "titlecard", "kicker": "Chapter 2", "title": "The Money", "start": 312.0, "end": 315.0}; also top-level"chapters": [{"start": 0, "title": "..."}, ...]. - Quotes:
{"type": "quote", "text": "exact words", "author": "Name", "role": "Source, year"}held for reading time (>= 1 s per 4 words). - Evidence:
linechart,bars,barracefor data (cite the source in achipunderneath);mapwithpinsfor places (stylised);stepsfor timelines. - Highlighting a document:
textwith"fx": {"name": "highlight", "target": "key phrase"}. - Attribution: a
chip"Footage: Archive name" in a corner for every borrowed clip. - Sound:
"audio": {"voice": {"eq": "broadcast", "deess": true}},"music": {"src": "score.wav", "under": 26, "duck": "speech-aware"}. - Look:
"color": {"look": "fuji-eterna", "intensity": 0.6},"finish": {"grain": 7, "vignette": 0.3, "letterbox": 0}.
Deliver
python CORE/scripts/deliver.py essay/plan.json --to youtube --out-dir essay/deliver
python CORE/scripts/qc.py essay/deliver/final_youtube.mp4 --plan essay/plan.json --platform youtube
- Paste
chapters.txtandsources.mdinto the description; uploadcaptions.srt. - Teaser: cut a 45 s vertical from the cold open with
retention-reel.
Check before delivering
- Every claim on screen has a source; every quote is verbatim and in context.
- Chapters valid (first at 0:00, >= 3, each >= 10 s) and match the cards.
- VO intelligible over music everywhere; archive audio levels matched.
qc.pypasses; no black frames between cutaways.- Tell the user: runtime, chapter list, sources used, files.
Example requests
- "Edit my video essay narration with this archive folder, chapter cards and a timeline graphic."
- "Documentary-style YouTube video about the 2008 crash from my VO and these charts."
- "Make it feel like a film essay - filmic grade, quotes on screen, chapters."
Files (claude-video-editing-skills)
-
SKILL.md 6.3 KB
--- name: video-essay description: "Video-essay and documentary-style YouTube editor: script-first, voice-over-led edits with archival footage, B-roll, quote cards, maps, timelines, charts and chapter cards, cut at the pace of an argument (4-8 s shots), filmic grade, sourced on-screen attribution and YouTube chapters. Use when the user says: video essay, documentary-style YouTube video, explain this topic in depth, narrated essay, history/analysis/film-analysis video, edit my VO with archive footage, chapter cards. Default skin essay-mode; any Video Studio skin works." license: MIT --- # Video Essay An argument told in pictures: VO leads, every visual is evidence, chapters keep the viewer oriented. - **Best for:** analysis, history, film/media essays, science and culture deep dives - **Inputs:** the recorded VO (or on-camera narration), archival clips/stills and B-roll in a folder, sources list, optional music - **Default skin:** `essay-mode` · alternatives `doc-real`, `paper-cut`, `swiss-poster` (use `--no-matte` without a presenter) - **Formats:** 16:9 1920x1080 master, chapters text, `captions.srt`; optional 9:16 teaser - **Duration target:** 10-40 min; genre preset `video-essay` ## Before you start - Needs **video-studio-core** in the same skills folder. `CORE` = `video-studio-core`, `SKILLS` = the folder holding both. - First time: `python CORE/scripts/check_env.py`. Preset: `python CORE/scripts/editor.py genre video-essay`. - The `--genre` preset sets autocut's pause cut (0.6 s) and run.py respects it; pass `--max-gap` only to override. - **Fair use:** quote clips briefly and comment on them; credit sources on screen; never present someone's claim out of context. Keep a `sources.md`. ## How a pro cuts this | beat | % | what happens | | --- | --- | --- | | cold open | 0-5 | the most intriguing image or question, before any title | | thesis | 5-10 | what this video argues, in one sentence | | chapters | 10-85 | 3-6 chapters, each: claim -> evidence (clip/quote/chart) -> implication | | turn / counter-argument | 85-92 | the strongest objection, answered | | conclusion | 92-100 | thesis restated as an insight; end on an image, not "thanks for watching" | - **Pacing:** ASL 4-8 s; a visual change at least every 6 s. Let a key archival moment play 8-12 s with its own sound. Slow down for the emotional chapter, quicken in montage sections. - **Cut rules:** cut the VO first (radio edit), then picture. Visuals land 0.1 s before the words that need them. Pauses between chapters are allowed (0.6-1 s). - **J/L cuts:** archival sound up under the end of a VO line (J) so the clip "speaks"; L-cut the VO over the next visual. - **Transitions:** cut; dip to black or a chapter card between chapters; dissolves for time passing. - **Music/SFX:** score bed 24-28 LU under VO, changes mood per chapter; drop it for the revelation. SFX almost none (page-turn on cards). - **Colour:** unify archive and new footage: `documentary-natural` 0.5 or `fuji-eterna` 0.6; archive can stay as is with `sepia` for flashbacks. - **Captions:** sidecar SRT; on-screen text only for quotes, names, dates and sources. - **Client judges:** clarity of the argument, research visuals, accuracy, chapters that make sense. - **Common mistakes:** visuals that decorate instead of prove, unattributed quotes, music wall-to-wall, a 2-minute intro, uncredited clips. ## Make the edit With a narration video (on-camera or a VO muxed over a still - see `faceless-explainer` for the TTS route): ```bash python CORE/scripts/run.py narration.mp4 --style SKILLS/essay-mode --genre video-essay \ --smart-cuts --retakes --model small --prompt "names, places" --work essay --stop-after plan python CORE/scripts/story.py outline --work essay --genre video-essay # paper edit: topic shifts = chapter candidates python CORE/scripts/broll.py --work essay --library archive/ --frequency high --merge essay/plan.json python CORE/scripts/editor.py review essay/plan.json --genre video-essay --fix python CORE/scripts/render.py essay/plan.json --frames 2,60,180,400 python CORE/scripts/render.py essay/plan.json python CORE/scripts/finish.py essay/plan.json --platform youtube ``` `broll.py` matches archive file names to spoken words; name files for what they show (`berlin-wall-1989.mp4`, `newspaper-headline.jpg`) or add `NAME.txt` keyword files. Place the important evidence shots by hand in `cutaways` (`{"src": "../archive/speech.mp4", "start": 84.2, "end": 95.0, "srcStart": 12}`). ## Refine the plan - **Chapter cards:** `{"type": "titlecard", "kicker": "Chapter 2", "title": "The Money", "start": 312.0, "end": 315.0}`; also top-level `"chapters": [{"start": 0, "title": "..."}, ...]`. - **Quotes:** `{"type": "quote", "text": "exact words", "author": "Name", "role": "Source, year"}` held for reading time (>= 1 s per 4 words). - **Evidence:** `linechart`, `bars`, `barrace` for data (cite the source in a `chip` underneath); `map` with `pins` for places (stylised); `steps` for timelines. - **Highlighting a document:** `text` with `"fx": {"name": "highlight", "target": "key phrase"}`. - **Attribution:** a `chip` "Footage: Archive name" in a corner for every borrowed clip. - **Sound:** `"audio": {"voice": {"eq": "broadcast", "deess": true}}`, `"music": {"src": "score.wav", "under": 26, "duck": "speech-aware"}`. - **Look:** `"color": {"look": "fuji-eterna", "intensity": 0.6}`, `"finish": {"grain": 7, "vignette": 0.3, "letterbox": 0}`. ## Deliver ```bash python CORE/scripts/deliver.py essay/plan.json --to youtube --out-dir essay/deliver python CORE/scripts/qc.py essay/deliver/final_youtube.mp4 --plan essay/plan.json --platform youtube ``` - Paste `chapters.txt` and `sources.md` into the description; upload `captions.srt`. - Teaser: cut a 45 s vertical from the cold open with `retention-reel`. ## Check before delivering - Every claim on screen has a source; every quote is verbatim and in context. - Chapters valid (first at 0:00, >= 3, each >= 10 s) and match the cards. - VO intelligible over music everywhere; archive audio levels matched. - `qc.py` passes; no black frames between cutaways. - Tell the user: runtime, chapter list, sources used, files. ## Example requests - "Edit my video essay narration with this archive folder, chapter cards and a timeline graphic." - "Documentary-style YouTube video about the 2008 crash from my VO and these charts." - "Make it feel like a film essay - filmic grade, quotes on screen, chapters."
Comments (0)
Sign in to join the conversation.
Reviews (0)
No reviews yet.
No comments yet.