runwayml / runway-studio-skills

The single skill routes to five uv-run scripts: generate_video.py covers text-to-video, image-to-video, video-to-video, and character performance; generate_image.py covers text- and image-to-image across gen4_image, gen4_image_turbo, and gemini_2.5_flash; generate_audio.py covers sound effects, TTS, speech-to-speech, voice dubbing, and voice isolation; get_task.py polls, retrieves, or deletes a running task; and runway_helpers.py is the shared API client, polling, and download logic the other four call into. Each script declares its own dependencies inline, so uv run needs no separate install step beyond uv itself, and the repository ships a COOKBOOK.md of end-to-end recipes alongside the skill's own SKILL.md.

This is for someone who already has, or is willing to buy, credit on Runway's own platform and wants an agent that calls generate_video.py directly instead of hand-writing the API calls — not a way to use Runway for free. Our dependency check labels the whole package account_key: nothing in it runs without RUNWAYML_API_SECRET set, and usage is billed on Runway's side on top of whatever running the skill itself costs.

READ THE FULL ANALYSIS＋

Where this overlaps with runwayml/skills. Runway also publishes a second, larger package, runwayml/skills, with seventeen skills covering the same video, image, and audio generation plus account inspection, file uploads, real-time avatar characters, and framework-specific server integration code. Three of that package's skills — rw-generate-video, rw-generate-image, rw-generate-audio — cover close to the same ground as this one's generation scripts under different names; the difference here is the product-ad framing (COOKBOOK.md, the seedance2 examples) rather than new capability. If you already have runwayml/skills installed, check there first.