agentleFS
LordVayronLordVayron/video-factory/.claude/skills/cast/SKILL.md

cast

Skill67 starsChanged 18 days agoMITView on GitHub

Creates and maintains the reusable bible of characters, sets and styles (bible/), including character sheets, voice choice and voice samples, and the "clone me" / AI-influencer intake from the user's own photos and voice. Use when a story needs a character or location that doesn't exist yet, when a character drifts between shots, or when the user says "clone me", "make an AI influencer", "new character", "crea un personaje".

10 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use mend.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • frankieramirezSkill

    mend

    Resolve an in-progress git merge, rebase, cherry-pick, or revert that has conflict markers or unmerged paths, or merge a pull request's base into the current branch and resolve what conflicts. Use when asked to mend, resolve merge conflicts, fix rebase conflicts, finish a conflicted rebase or cherry-pick, resolve the conflicts on this PR, my PR has conflicts, merge main into this branch, bring this branch up to date, /mend, or /resolve-merge-conflicts.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 1 is distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 10 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Look at the result yourself.

    line 16
  2. Very specific:

    Add one running gag or catchphrase to traits.

    line 22
  3. Very specific:

    Use realistic scale words ("normal-sized house cat, small next to the furniture") and natural animal posture.

    line 30
  4. Very specific:

    Ask the user for 6–12 photos of themselves (front, ¾, profile, two outfits, good light) dropped in bible/characters/<id>/photos/, plus 1–2 …

    line 45

“Common” is measured: kept by at least 6 in 10 of the 10 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Cast and bible

Everything here is reused across projects, so it is worth getting right once.

Character (bible/characters/<id>/)

character.json fields: id, name, species, look (one dense sentence: body, face, hair/fur, wardrobe, one signature accessory, and every distinctive mark by name: freckles, moles, scars, glasses, earrings. The prompt check refuses any video prompt that drops one, and the identity eval checks each of them on every take), traits, flaw, role (straight man / eccentric / …), voiceDesc, voice, sheets.

  1. Sheet: one 3:2, 2K image showing front, ¾ and full body on a plain backdrop. node engine/gen.mjs still "<sheet prompt>" bible/characters/<id>/sheet.png --ar 3:2 --size 2K Template: "Character reference sheet on a plain warm cream (#F2E8D5) studio backdrop, photoreal, 35mm film look, soft even key light. The same character shown three times side by side: (1) front view, (2) three-quarter view, (3) full body. Character: <look>. Identical wardrobe and markings in all three views. No text, no labels." Look at the result yourself. Reroll for wrong anatomy, mismatched views, text. Crop with ffmpeg rather than paying for a reroll when one view is bad.
  2. Voice:
    • voice.id: an ElevenLabs preset that Kie serves (list in engine/voices.txt, preview at https://static.aiquickdraw.com/elevenlabs/voice/<id>.mp3). Optional per-language overrides voice.en/voice.es.
    • voice.eleven: a voice on the user's own ElevenLabs account, used when Kie's endpoint fails. Free plans can only use premade voices: onwK4e9ZLuTAKqWW03F9 (Daniel, British), TX3LPaxmHKxFdv7VOQHJ (Liam, young), JBFqnCBsd6RMkjVDRZzb (George, warm British), EXAVITQu4vr4xnSDxMaL (Sarah), pNInz6obpgDQGcFmaJgB (Adam).
    • voice.sampleText.{en,es}: about 8s of in-character speech.
    • node engine/gen.mjs sample <id> --lang en writes voice-en.mp3. Seedance uses it as @audioN so the voice stays the same in every shot.
  3. Add one running gag or catchphrase to traits.
  4. Approval (flow v5 cast gate): show the creator the sheet, the look and a voice sample; with their yes, node engine/studio.mjs approve-character <id> --said "<their words>" (or cast.mjs approve for in-style sheets). Lock the face anchor from their approved take with continuity.mjs <slug> --identity <take> --lock-anchor.

Realistic styles: native references (the rule that fixed the "sticker" look)

Image models composite a studio reference photo into the scene, keeping its pose and its lighting, so the character looks pasted in. For any realistic style:

  1. Make the empty set plate first.
  2. Generate each character inside the set from text only (only the set plate as a ref), with Nano Banana Pro at 2K: "a frame grab ... a real, normal-sized <animal/person> ... lit by the same light as the room, soft shadow on the floor, same video texture ... not composited". Clothing must be described as real garments that fit imperfectly ("loose, bunches at the shoulders").
  3. Crop the character out of that frame → ref-native.png, set "refs": ["ref-native.png"]. The studio sheet.png stays only as a design reference; never feed it to shots.
  4. Use realistic scale words ("normal-sized house cat, small next to the furniture") and natural animal posture. A realistic character works best with a native reference (bible/characters/<id>/ref-native.png) in addition to the in-style sheet.

Set (bible/sets/<id>/set.json)

look (one dense sentence), plates (a wide 9:16 empty plate at minimum), geography (where things are, and who stands on which side of frame). node engine/gen.mjs still "<style string> wide establishing shot, empty of characters. <look>. No text, no people." bible/sets/<id>/plate-wide.png --size 2K

In-style character sheets (phase 3, the current method)

node engine/cast.mjs sheet <char> --style <style> --set <set> makes a 2x2 sheet in the style (realistic styles: shot inside the set plate), splits it into front, threeQuarter, profile and full views under bible/characters/<char>/<style>/, and records them in character.json → looks.<style>. Review it and score: cast.mjs approve <char> --style <style> consistency=N anatomy=N design=N style=N layout=N. Rejected sheets: regenerate with --note "<fix>". Shots then get only the views their framing needs.

Styles

Style sheets live in bible/styles/<id>/style.json and are managed with engine/style.mjs (see the studio skill and studio/references/styles.md).

Clone me / AI influencer

The user's own likeness, a fictional person, or a real person the user says authorized it. Record whose likeness it is with node engine/cast.mjs likeness <id> --kind self|fictional|real-person; a real person needs --said "<the user's words that they were authorized>", and the cast gate refuses the character without it. Ask the user before building a real person's sheet.

  1. Ask the user for 6–12 photos of themselves (front, ¾, profile, two outfits, good light) dropped in bible/characters/<id>/photos/, plus 1–2 minutes of clean speech in voice-source.mp3.
  2. Build the sheet with Nano Banana from the photos as refs (--ref each photo, up to 10): "Character reference sheet of this exact person…".
  3. Voice: instant voice cloning needs an ElevenLabs paid plan (Starter). Clone in the ElevenLabs UI, then put the new voice id in voice.eleven with "provider": "elevenlabs". Until then, pick the closest preset.
  4. For talking-head content, node engine/gen.mjs talk <photo-or-keyframe> <line.mp3> out.mp4 --model infinitetalk ($0.015/s at 480p). For scenes, use the character in a normal shot list.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.