Skip to content

brotto/kieai-skill

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

kieai — Claude Skill for Kie.ai

A Claude Skill that lets the Claude in your session generate images, videos, music, and speech through Kie.ai — 300+ models including Sora 2, Veo 3.1, Kling, Suno, ElevenLabs, Seedream, Flux, Imagen, Ideogram, Topaz, and more.

You just ask. Claude picks the best model by cost × quality, runs the generation, and gives you the URL.

Install

# 1. Clone into your Claude skills directory
git clone https://github.com/brotto/kieai-skill ~/.claude/skills/kieai

# 2. Save your Kie.ai API key (get one at https://kie.ai/api-key)
mkdir -p ~/.config/kieai
chmod 700 ~/.config/kieai
printf '%s' '<your key here>' > ~/.config/kieai/api-key
chmod 600 ~/.config/kieai/api-key

That's it. Open any Claude Code session and try:

"Gera uma imagem de um gato samurai estilo ukiyo-e."

"Make a 10-second cinematic video of a wave breaking on a black-sand beach."

"Crie uma música upbeat sobre café da manhã, instrumental, 30 segundos."

"Convert this paragraph to speech in PT-BR with a calm female voice: ..."

Claude will read the catalog, pick the right model (e.g. sora-2-t2v at $0.15), confirm with you if the cost is over $0.15, then generate.

How it works

~/.claude/skills/kieai/
├── SKILL.md          # the trigger — Claude reads this when you ask to generate media
├── models.json       # curated catalog of ~27 Kie.ai models with cost + capabilities
├── scripts/
│   ├── generate.py   # POST to Kie.ai, polls until done, prints result URL
│   └── credits.py    # check your remaining Kie.ai credits
└── README.md

The skill has no Anthropic dependency — your Claude session is the orchestrator. No extra API costs beyond Kie.ai.

Manual usage (without Claude)

# Generate an image
python3 ~/.claude/skills/kieai/scripts/generate.py \
  --model-key seedream-4-5-t2i \
  --prompt "A samurai cat in ukiyo-e style"

# Check credits
python3 ~/.claude/skills/kieai/scripts/credits.py

# List all available models
python3 -c "import json; [print(m['key'], '—', m['displayName']) for m in json.load(open('$HOME/.claude/skills/kieai/models.json'))['models']]"

API key configuration

The scripts look for the key in this order:

  1. $KIEAI_API_KEY environment variable
  2. ~/.config/kieai/api-key file

Option 2 is recommended — persistent and out of your shell history.

Catalog

models.json covers ~27 curated models across 4 categories. The full Kie.ai catalog has 309 models; this skill picks the ones with the best cost/quality balance per use case. If you want a model not in the curated list, edit models.json locally — the schema is documented inline.

Refresh the snapshot date by replacing snapshotDate in models.json when pricing changes.

Why a skill instead of an app

A web app would need its own LLM (Anthropic SDK), server, and UI. A skill puts the orchestration logic right inside whatever Claude session you're already using:

  • Zero extra LLM cost — your Claude session is the brain.
  • Zero infrastructure — no server, no deploy.
  • Works anywhere Claude runs — Claude Code, claude.ai, IDE extensions.
  • Easy to extend — just edit SKILL.md and models.json.

License

MIT — see LICENSE.

Status

MVP. Known limitations:

  • No automatic file upload (image-to-video needs a public image URL).
  • 5-minute polling timeout — long Sora Pro / Veo Quality jobs may exceed this. Re-run with the returned taskId or check at https://kie.ai/logs.
  • 27 of Kie.ai's 309 models are curated. Add yours to models.json to extend.

Pull requests welcome.

About

Claude Skill that lets Claude generate images, video, music and speech via Kie.ai (300+ AI models). Zero-dep Python, install in 5s.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages