Edit images via RunComfy CLI with identity-preserving background and object swaps.
npx skills add https://github.com/runcomfy-com/skills --skill nano-banana-edit
Please help me install this Agent Skill. Skill: nano-banana-edit Source: https://github.com/runcomfy-com/skills/tree/main/nano-banana-edit Command: npx skills add https://github.com/runcomfy-com/skills --skill nano-banana-edit
It solves the problem of performing high-quality image-to-image edits (like background swaps or localized object changes) while preserving the subject’s identity across one or many reference images.
Use the nano-banana-edit skill to swap the background of a portrait while keeping the subject identity unchanged by prompting: "Keep the subject identity, pose, and clothing unchanged; convert the background into a rainy neon cyberpunk street."
High-intent search queries and answers about installing and using this skill.
To perform identity-preserving image edits, swap or modify specified visual elements using concrete spatial prompts while the system maintains the subject's identity. It processes up to 20 input images per call with controllable aspect ratio and resolution.
Yes, you can apply changes to specific regions using concrete spatial language in your prompts, such as left object only or upper-right corner. This targets localized object edits without altering the rest of the image.
You can batch edit up to 20 input images per call with locked framing via consistent aspect ratio and resolution. This ensures multi-image batch consistency across all processed files.
Yes, performing these identity-preserving image-to-image edits requires invoking the RunComfy CLI. The CLI calls the google/nano-banana-2/edit endpoint using an input JSON schema containing your prompt and image_urls.
For text-in-image edits or pure generation tasks, prefer sibling skills like GPT Image 2 edit or Nano Banana 2 t2i instead. Use this tool specifically for Nano Banana 2 edit scenarios requiring identity preservation.
For single-reference precise local edits, prefer the Flux Kontext sibling skill. This tool is optimized for multi-image batch consistency and broad identity-preserving edits rather than single-ref localized changes.
runcomfy-comRoute still images to RunComfy image-to-video models for animated video output.
rshankrasGenerate App Store screenshots via UI tests with device framing and localized captions.
sixtycat2000-ctrlGenerate AI images via Volcengine Seedream with text-to-image and editing.
agentspace-soEdit single images with Flux 1 Kontext Pro via RunComfy CLI.
