ComfyUI.org

Create Breathtaking Visuals: "miluo丨Blue Sky & White Clouds" Social Media Post Workflow Revealed

CN
ComfyUI.org
2025-06-11 09:06:31

1. Workflow Overview

mbrq5gsszt9lxkmo22图片压缩83a7cd2a518ad57b891c54179fbe632b667c56909add80b265655cc42c069175.png

This workflow, named "miluo丨Blue Sky & White Clouds_Social Media Post", generates aesthetic social media images featuring clear skies, artistic portraits/scenes, ideal for posters or online content.

  • Core Models:

    • Stable Diffusion: 高级质感人像-FLUX_v1.0 (optimized for portrait quality).

    • LoRA: miluo丨蓝天白云去旅行_文艺文案风_v1.8 (enhances sky/cloud and artistic style).

  • Function:

    • Generates high-quality images from prompts.

    • Supports online and local execution modes (grouped in the workflow).

2. Key Nodes

Node Name

Description

Installation

CheckpointLoaderSimple

Loads the main model (高级质感人像-FLUX_v1.0).

Manually download to models/checkpoints.

LoraLoader

Loads LoRA (miluo丨蓝天白云去旅行_文艺文案风_v1.8).

Download .safetensors to models/loras.

CLIPTextEncode

Encodes prompts (e.g., default Asian woman description).

Built-in.

KSampler

Controls generation (steps: 30, CFG: 1, sampler: euler).

Built-in.

VAEDecode

Decodes latent images to pixels.

Built-in.

ttN text

Allows dynamic prompt input.

Install ttN nodes custom node.

SamplerCustomAdvanced

Advanced sampler (for local execution).

Install ComfyUI-Custom-Sampler.

3. Workflow Structure

Two main modules:

  • Online Execution (Default):

    • Group 1: Model Loading

      • Nodes: CheckpointLoaderSimple, LoraLoader.

      • Output: Loaded models + CLIP encoder.

    • Group 2: Generation

      • Nodes: CLIPTextEncode, KSampler, VAEDecode.

      • Input: Prompts, resolution (768x1248), sampling params.

      • Output: Generated image.

  • Local Execution (Manual):

    • Uses SamplerCustomAdvanced for high-performance local GPUs.

4. Inputs & Outputs

  • Inputs:

    • Resolution: 768x1248 (online) or 768x1024 (local).

    • Prompts: Default includes "blue sky/white clouds" and portrait descriptions (customizable).

    • Sampling: Steps 30, CFG 1, euler sampler.

  • Output:

    • Final image: Artistic HD render (PNG).

5. Notes

  • Model Dependencies:

  • Hardware:

    • Online: ≥6GB VRAM.

    • Local: ≥8GB VRAM (e.g., RTX 3060+).

  • Troubleshooting:

    • Poor results: Adjust prompts or LoRA weight (default: 0.4).

    • Resolution issues: Match output size to model training dimensions.

previous
Unlock the Secrets of Traditional Chinese Art: A One-Click Landscape Painting Workflow
next
Mastering AI Portrait Editing: A Step-by-Step Workflow for Stunning Results

Recommend