Skill Runway Video Gen

Generate short product videos from images using Runway Gen4 Turbo. Use for TikTok ads, UGC-style product demos, Reels, and YouTube Shorts.

Security Scan

The skill largely does what it says (submits an image to Runway and downloads a video), but there are mismatches between its declared requirements and what the code actually uses, and the script targets a 'dev' Runway API host — these inconsistencies warrant caution before installing or running.

Purpose & Capability

The script implements image→video via Runway Gen4 Turbo and the SKILL.md usage matches the code. However, registry metadata lists no required env vars or config paths while the script expects a RUNWAY_API_KEY or ~/tiktok-api.json → runway.apiKey. Declared binary 'uv' is reasonable for running the script.

Instruction Scope

SKILL.md instructs running the included Python script and documents the API key lookup (env var or ~/tiktok-api.json). The script reads an image, base64-encodes it, sends it to https://api.dev.runwayml.com, polls for results, and downloads the produced MP4 — behavior consistent with purpose. Concern: it reads a user home config file (~/tiktok-api.json) which was not declared in registry metadata and could contain other secrets; SKILL.md does document this file but the top-level skill metadata omitted it.

Install Mechanism

No install spec (instruction-only plus a script). The Python shebang declares dependencies (requests) but there is no automated install step to ensure them. This is low risk but operationally brittle; the skill does not download or execute code from unknown URLs.

Credentials

The script requires a Runway API key (RUNWAY_API_KEY) or a key in ~/tiktok-api.json, but the skill's registry metadata lists no required environment variables or config paths. That mismatch is a proportion/visibility problem. Also the code posts your image data to api.dev.runwayml.com (a 'dev' subdomain) rather than an obvious production API host — that increases the risk that data could be handled by non-production infrastructure.

Persistence & Privilege

The skill does not request persistent autonomy (always is false), does not alter other skills or system configuration, and writes only the requested output file. No elevated privileges requested.

What to consider before installing

Before installing or running this skill: 1) Verify and provide a dedicated RUNWAY_API_KEY (or a safe ~/tiktok-api.json entry) — do not reuse high-privilege or unrelated keys. The registry metadata does not declare this requirement even though the script needs it. 2) Confirm the API host: the script uses https://api.dev.runwayml.com (a dev/staging-looking endpoint). Confirm with Runway documentation or the skill author that this is intentional; otherwise you may be sending images to a non-production endpoint. 3) Review ~/tiktok-api.json contents if you plan to rely on it — it may contain other secrets; prefer an env var with a least-privilege key. 4) Ensure Python (>=3.10) and the 'requests' dependency are available; the skill has no install step. 5) Remember your images and generated video are uploaded to an external service — consider privacy and IP implications. If anything above is unclear or you cannot verify the Runway API host, treat this skill as untrusted until clarified.

Wraps Runway Gen4 Turbo image-to-video API. Point at a product image, describe the motion, get an MP4. No browser needed.

Usage

uv run scripts/generate_video.py \
  --image product.jpg \
  --prompt "water droplets falling, soft bokeh, slow motion" \
  --output output.mp4 \
  --duration 10 \
  --ratio 720:1280

Args

Arg	Default	Description
--image	required	Path to source product image
--prompt	""	Motion description (optional but recommended)
--output	required	Output MP4 path
--duration	10	Video length: 5 or 10 seconds
--ratio	720:1280	Aspect ratio (vertical for TikTok)

Config

API key is read from (in order):

1. RUNWAY_API_KEY environment variable
2. ~/tiktok-api.json → runway.apiKey

Cost

Duration	Cost
5s	$0.25
10s	$0.50

Always use 10s — you get more content to work with.

Known quirks

Last 1–2s often freezes. Fix: stretch to 12s at 0.83x speed using MoviePy:

from moviepy import VideoFileClip
clip = VideoFileClip("output.mp4")
slow = clip.with_effects([vfx.MultiplySpeed(0.83)])
slow.write_videofile("output_12s.mp4", fps=30)

Rate limits

No per-minute quota (unlike Veo). Runway charges per second of output.

Pipeline integration

This skill feeds into skill-tiktok-video-pipeline as the video generation step. The pipeline handles slowmo stretch and caption overlay automatically.