Capybara for Image Editing

Capybara’s image‑to‑image mode lets you edit an existing image by instruction, keeping structure and identity while changing style, content, or lighting.

What it is

• A unified visual model where ti2i / image‑edit mode takes an input image plus a text instruction and outputs an edited version (local or global changes).

• Exposed in ComfyUI via the “Capybara: Image Edit” template and 4‑in‑1 Capybara workflows.

Key features

• Instruction‑based edits: prompts like “replace the background with a sunset beach” or “make it cyberpunk at night” without masks for many cases.

• Handles both local edits (object replacement, expression tweaks) and global edits (time of day, color grading, style change).

• Supports multi‑turn editing—you can apply several sequential instructions while preserving identity and layout.

• Works in the same pipeline as text‑to‑image and image‑to‑video, so style and look stay consistent across assets.

Best use cases

• Background and atmosphere swaps on portraits or product images (day ↔ night, indoor ↔ outdoor, realistic ↔ stylized).

• Object / outfit / prop changes while keeping the same person, pose, and framing.

• Look‑dev iterations: quickly trying different styles or moods on a base keyframe before animating it with image‑to‑video.