Unlock Magical Pet Photography with This One-Click Workflow

CN
ComfyUI.org
2025-04-23 10:09:50

1. Workflow Overview

m9trvbaifg6o37rzm233bf8a7a305ad013fe8f4f04277223b05c20d1494cd2fa55b8d4cd9cd87add49.gif

This workflow specializes in generating commercial-style pet photography with magical themes (e.g., cats/dogs in Hogwarts costumes). Built on Anything v6.8 and HIRO Photography LoRA, it's optimized for one-click operation.

2. Core Models

  • Base Model: Anything v6.8 F.1 (FP8 optimized)

  • LoRA: HIRO Pets in Magic World v1.0

  • CLIP Encoders: Dual t5xxl_fp8 + clip_l

3. Key Nodes

Node

Function

Installation

CLIPTextEncodeFlux

Natural language prompt parsing

Requires Flux extension

RandomNoise

Deterministic noise generation

Noise Nodes add-on

Dependencies:

  • Custom ae.sft VAE model

  • HIRO LoRA series (must include c5507930ae4e49c6...)

4. Workflow Groups

Group A: Preset Config (Don't Modify)

  • Handles VAE, Euler sampler, latent decoding

Group B: FP8 Optimization

  • UNET/CLIP loaders with fp8_e4m3fn precision

Group C: Style Injection

  • Applies HIRO LoRA at 1.0 strength

Group D: Prompt Engineering

  • Uses Flux-enhanced text encoder

  • Template: Must start with HIRO,

5. I/O Specification

  • Inputs:

    • Magic-themed prompts (e.g., "cat in Ravenclaw uniform")

    • Fixed seed 969501611665423

  • Output:

    • 768x1024 images (saved to IP Brand Design folder)

6. Critical Notes

⚠️ Requirements:

🔧 Pro Tips:

  • Keep BasicScheduler in simple mode

  • Add , after HIRO for better style adherence