[FLUX] T2I with LLM prompt版本v2.0 (ID: 1021975)

# FLUX Intelligent Prompt Processing & Image Generation Pipeline

A low VRAM hardware-friendly workflow designed for FLUX that combines intelligent prompt processing with comprehensive image generation and refinement. This pipeline processes prompts through a series of LLM nodes for optimal results while maintaining reasonable VRAM requirements.

## Prompt Processing Chain

1. Tag Cleaning Node

- Takes raw inputs including style preferences, LoRA details, and reference materials

- Cleans and organizes tags for thematic consistency

- Outputs structured format: Main Idea || Subject Description || Tags

2. Prompt Generation Node

- Processes cleaned tags and descriptions

- Creates cohesive, optimized prompts for FLUX

- Combines natural language descriptions with technical tags

## Image Generation Pipeline

1. Initial Sample Generation

- Creates base image from optimized prompt (initialization, prompt, and 1st pass sample)

- Balances quality with performance

2. Multi-Stage Refinement

- Highres Fix for improved detail

- Tiled Upscaling for larger outputs

- Face Detail Enhancement

- Focused Eye/(any BBOX detailer) Detail Pass

- Final Post-Processing

## User Feedback

- Posting images generated with this workflow are appreciated

- Comments and feedback welcome for future improvements

- This is Version 1.0 - feedback is welcome

## Getting Help

Feel free to:

- Share generated images

- Report issues or difficulties

- Request assistance through comments

- Suggest improvements

描述:

Updated to remove setnode and getnode. (I learned that these nodes don't work for some). This workflow still uses the CR Use Everywhere nodes to reduce the node pipe spaghetti.

训练词语: