Swap faces and characters in images and videos via the RunComfy CLI.
npx skills add https://github.com/genmedia-labs/skills --skill face-swap-genmedia-labs
Please help me install this Agent Skill. Skill: face-swap Source: https://github.com/genmedia-labs/skills/tree/main/face-swap Command: npx skills add https://github.com/genmedia-labs/skills --skill face-swap-genmedia-labs
runcomfy CLI with the right JSON payload. ## Core Features & Use Cases - Intent-based model routing: Picks among Wan 2-2 Animate (video character swap with audio), Kling 2-6 Motion Control Pro (motion transfer), GPT Image 2 Edit (multi-reference still swap), Nano Banana 2 Edit (batch identity-preserving swap), and Flux Kontext Pro (single-image prose-described face edit). - Ready-to-run CLI commands: Each route includes the exact runcomfy run <model_id> --input '{...}' invocation with schema notes and prompting tips. - Consent and safety guidance: Explicit dual-use disclosure requirements and refusal criteria for non-consensual or harmful synthetic media. - Use Case: A marketer wants to cast a brand spokesperson into existing footage — the Skill selects Wan 2-2 Animate, builds the input JSON with the spokesperson's portrait and the original audio track, and downloads the rendered video. ## Quick Start Ask the agent to swap the face from one image into a target photo or video using RunComfy, and it will pick the right model route and run the CLI command for you.High-intent search queries and answers about installing and using this skill.
Use the Wan 2-2 Animate endpoint via `runcomfy run community/wan-2-2-animate/api` with a reference image of the new identity and an audio track. The model produces a video where the new character drives the performance.
Use GPT Image 2 Edit for precise multi-reference swaps on a single hero still, Nano Banana 2 Edit for swapping one identity consistently across up to 20 frames, or Flux Kontext Pro when you have no reference face image and describe the new face in prose.
Wan 2-2 Animate is audio-driven character animation from a single reference image. Kling 2-6 Motion Control Pro transfers the exact motion and blocking of a reference performance video onto a target character image, with face swap as a byproduct.
Yes. Nano Banana 2 Edit accepts 1-20 input images per call. Place the identity reference as the first image URL, list the scene images after it, and lock aspect_ratio and resolution for consistent batch output.
No. The Skill requires rights to both the target face and the source asset, and instructs agents to refuse requests targeting real people without consent or aiming at defamatory, sexually explicit, or otherwise harmful synthetic media.
Reference image, audio, and video URLs are untrusted inputs and a known injection target. Ingest only URLs the user explicitly provided, and suspect the reference asset when output diverges from the prompt.
Pixazo-AIReplace a video's audio track with a different audio file via the Pixazo API.
ai-service-incubatorGenerate 1080x1920 MP4 short-form videos from scripts with subtitles, images, and music.
Growth4U-systemsGenerate talking-head videos and lip-synced avatars from scripts.
ahostbrTranslate ComfyUI workflow JSON node graphs into LiteImage REST API call sequences.