FisherKing-ZiT-I2I-Guided-ReferenceWorkflow-v1.0 (Low VRAM)

Intent Use a reference image to guide composition, framing, and approximate pose while allowing the prompt and LoRAs to redefine the subject, clothing, environment, and artistic direction. Version 1.0 Created as a simple, reference-guided Image-to-Image workflow for users getting started with ZiT. Workflow Scope This workflow is a simple, reference-guided Image-to-Image implementation. The recommended model, text encoder, VAE, LoRA stack, and sampling settings were selected and validated together. Changing these components may produce different results or reduce reference consistency. Advanced workflows for: ✗ Exact identity preservation ✗ Precise pose transfer ✗ Background transfer ✗ Object replacement ✗ Region-specific editing typically require additional techniques such as ControlNet, IPAdapter, masking, inpainting, or other specialized nodes . These capabilities are intentionally outside the scope of this reference workflow. Reference Image Usage The input image is used as visual guidance . Expected to preserve: ✓ Overall composition ✓ Approximate pose ✓ Subject placement ✓ General framing Not intended to preserve: ✗ Exact facial identity ✗ Clothing ✗ Background ✗ Pixel-perfect pose matching The reference image provides visual guidance, while the prompt defines the final scene, subject, and style. Purpose ✓ Educational ✓ Minimal ✓ Easy to extend Verified ✓ Image-to-Image Verified Settings CFG : 1.0 Shift : 4.0 Denoise : 0.75 Sampler : Euler Ancestral Scheduler : Beta These settings were used to validate this workflow and are recommended as the baseline configuration. Recommended Resolution ✓ 960 × 1440 (Image scaler included. Use with caution, as resizing may alter the final output.) ✓ 1120 × 1440 Hardware Tested ✓ NVIDIA RTX 2080 (8 GB) ✓ 64 GB System RAM ✓ ComfyUI 0.27.0