CLI utility for applying cinematic color grading and film looks to visual media.

Install

mkdir -p .claude/skills/hance && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/18069" && unzip -o skill.zip -d .claude/skills/hance && rm skill.zip

Installs to .claude/skills/hance

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Apply cinematic film looks to images/video via the Hance CLI. Use when asked to grade, color, or "make it look like" film.
122 chars✓ has a “when” trigger
Beginner

Key capabilities

  • Install and set up Hance CLI
  • Apply cinematic presets to images or videos
  • Match the look of a reference image or video
  • Refine an existing color grade with adjustments
  • Batch process multiple images or videos
  • Open the Hance UI editor

How it works

The skill routes user requests to one of six subcommands (`setup`, `run`, `try`, `refine`, `batch`, `ui`) based on intent, executing the Hance CLI.

Inputs & outputs

You give it
User request to grade, color, or apply a film look to images/video
You get back
Command-line output from Hance CLI or UI launch

When to use hance

  • Apply cinematic color grade
  • Match film reference
  • Batch process images
  • Refine color look

About this skill

/hance

A single entry point that routes to one of six subcommands: setup, run, try, refine, batch, ui.

Routing

Pick the subcommand from the user's request. When in doubt, ask.

User intentSubcommandDoc
"install hance" / "set up hance"setupsubcommands/setup.md
"apply <preset> to <file>" / "grade with <preset>"runsubcommands/run.md
"make this look like X" / "match this reference" / "show me some looks for this"trysubcommands/try.md
"make this one look better" / "tune/refine this grade" / "more grain, warmer, etc."refinesubcommands/refine.md
"apply X to all of these" / batch a folderbatchsubcommands/batch.md
"open the editor" / "open the UI"uisubcommands/ui.md

Runner selection (every subcommand)

Pick one runner per invocation, in order:

  1. If command -v bun succeeds → use bunx @orva-studio/hance (preferred — fast cold-start, cached after first fetch).
  2. Else if command -v node succeeds → use npx @orva-studio/hance.
  3. Else → inform the user that a compiled binary is available for download from GitHub releases, but do not install it for them.

Runner priority: Bun (bunx), then Node (npx), then compiled binary as a last resort. Never install the binary for the user — just let them know it exists if they have neither Bun nor Node. Every subcommand must work on a fresh machine; setup only verifies runtime + ffmpeg and shows examples.

The compiled binary expects a hance-gpu sidecar alongside it; a bare ./hance checkout without it fails with ENOENT … hance-gpu. This is why bunx/npx are preferred — they always work. When developing in this repo, run the source directly: bun run packages/cli/src/cli.ts ….

Hard rules

  • Never invent flags. Only use names listed in hance --help or hance preset --help.
  • Never run a full video render to explore looks — use hance preview (still frame) for try.
  • Windows is unsupported. If uname -s reports Windows, stop and tell the user.

References

  • references/preset-index.mdpresets/index.json schema and rebuild contract.
  • references/compare-page.md/compare route query params and Edit hand-off.
  • references/grading.md — how to judge and dial in a film look: the render→read→adjust loop, what makes a grade read as film, and common artifacts. Read before refine.

When not to use it

  • When running on a Windows operating system
  • When needing to invent flags not listed in `hance --help`
  • When running a full video render to explore looks

Limitations

  • Windows is an unsupported operating system
  • Cannot invent flags not documented in `hance --help`
  • Cannot run full video renders to explore looks

How it compares

This skill provides a structured entry point for various cinematic color grading tasks, automating the selection of the correct Hance subcommand and runner.

Compared to similar skills

hance side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
hance (this skill)01moNo flagsBeginner
jimeng-mcp-skill194moCautionIntermediate
ai-media02moNo flagsIntermediate
generate04moNo flagsIntermediate

Try saying

Example prompts that trigger this skill in your AI assistant.

Search skills

Search the agent skills registry