image-outpainting
๐ฏSkillfrom runcomfy-com/skills
|Extend an image beyond its original canvas on RunComfy โ uncrop, change aspect ratio, or fill in what the camera missed. Routes across Nano Banana 2 Edit, GPT Image 2 Edit, FLUX Kontext Pro, and brand edit endpoints based on whether the outpaint is prose-driven, reference-driven, or brand-locked.
Overview
Image Outpainting is a Claude Code skill that extends still images beyond their original canvas boundaries via the RunComfy CLI. It handles use cases like uncropping photos, changing aspect ratios (e.g., square to 16:9 or portrait to landscape), and filling in content the camera did not capture. The skill automatically routes to the best model for the task: Nano Banana 2 Edit for prose-driven outpainting, GPT Image 2 Edit for reference-style matching, FLUX Kontext Pro for maximum content preservation, or brand-specific edit endpoints for style consistency.
Key Features
- Smart model routing: Automatically selects from Nano Banana 2 Edit (default, spatial-language driven), GPT Image 2 Edit (multi-reference style matching), FLUX Kontext Pro (single-shot maximum preservation), and brand edit endpoints based on the outpainting intent
- Aspect ratio conversion: Change image dimensions for platform-specific needs, such as converting a 1:1 square to 9:16 for Reels or 21:9 for cinematic banners, while preserving the original content
- Flexible prompting patterns: Lead with the canvas change direction, describe what should extend into the new area, and set explicit preservation goals to keep original content intact
- Multi-step chaining: Supports progressive outpainting by feeding one pass's output into the next, extending roughly 30-50% per pass to maintain boundary quality
Who is this for?
- Social media managers who need to reformat images across platforms with different aspect ratio requirements
- Product photographers preparing catalog images that need wider framing or additional environmental context
- Designers creating banner and hero images from portrait or square source photos
- Content creators restoring or uncropping historical and tightly-framed photographs
Same repository
runcomfy-com/skills(30 items)
Installation
npx vibeindex add runcomfy-com/skills --skill image-outpaintingnpx skills add runcomfy-com/skills --skill image-outpainting~/.claude/skills/image-outpainting/SKILL.mdSKILL.md
More from this repository10
Generate and edit images on RunComfy via a smart router across 11+ AI models including FLUX 2, Nano Banana 2, GPT Image 2, Seedream 5, and Qwen Image. Covers both text-to-image and image-to-image endpoints, automatically selecting the best model for the user's intent.
A Claude Code skill that animates still images into video on RunComfy, routing to HappyHorse 1.0 for general animations, Wan 2.7 for audio-driven lip-sync, or Seedance 2.0 Pro for multi-modal composition based on user intent.
A Claude Code skill for generating and editing images with OpenAI GPT Image 2 (ChatGPT Images 2.0) via the RunComfy API, excelling at embedded text, multilingual typography, logos, and directive precision for layout-critical imagery.
A Claude Code skill that generates custom OpenAI Codex Pets on RunComfy, turning a single reference image into a Codex-compatible spritesheet and pet.json using GPT Image 2 edit plus ImageMagick transforms, requiring only a RunComfy token.
A Claude Code skill for editing existing videos on RunComfy, routing to Wan 2.7 Edit-Video for general restyle and background swap, Kling 2.6 Pro Motion Control for precise motion transfer, or Lucy Edit Restyle for lightweight identity-stable outfit swaps.
A Claude Code skill that generates full songs and instrumental tracks with ElevenLabs Music on RunComfy, producing studio-quality 44.1 kHz stereo audio from 5 seconds to 5 minutes with section-level control for intros, verses, choruses, and bridges.
A Claude Code skill for region-based video editing on RunComfy, enabling object removal, watermark cleanup, and motion-matched region replacement across video frames using Wan 2-7 Edit-Video and other endpoints.
Generate AI videos on RunComfy through a smart router across the full video-model catalog including HappyHorse 1.0, Wan 2-7, Seedance v2, Kling 3.0, Veo 3-1, and Hailuo 2-3. Supports text-to-video, image-to-video, and video-extend modes with automatic model selection based on user intent.
A Claude Code skill for pose-conditioned image and video generation on RunComfy, routing across Kling Motion Control Pro for motion transfer, Wan 2-2 Animate for audio-driven character animation, and Z-Image Turbo for pose-conditioned image generation.
A Claude Code skill for extending or continuing existing video clips using Google Veo 3-1 extend-video endpoints via the RunComfy CLI, enabling shot-by-shot narrative chaining with consistent motion, lighting, and subject identity.