What is the fastest way to generate images using this skill?

Using the Flux 2 Klein 9B (Distilled) or Flux.1 Dev + Turbo LoRA patterns allows for high-quality image generation in just 4 steps.

How does it handle different text encoders?

The skill intelligently switches between DualCLIPLoader (T5XXL + CLIP-L) for Flux.1 Dev and single CLIPLoader (Qwen3-8B) for Flux 2 Klein variants.

Why is BF16 recommended over FP8 for SRPO models?

Flux.1 Dev SRPO requires BF16 because FP8 quantization is known to produce broken results or visual artifacts with this specific fine-tune variant.

Does Flux support traditional negative prompts?

No, Flux is guidance-distilled and typically uses a CFG of 1.0. This skill implements the 'ConditioningZeroOut' pattern or empty CLIP encoding to handle negative conditioning effectively.

Flux Txt2Img Workflows

Name: Flux Txt2Img Workflows
Author: artokun

byartokun

•

数据科学与机器学习

Generates and configures high-performance Flux.1 and Flux 2 image generation pipelines for ComfyUI.

This skill empowers Claude to construct sophisticated Flux text-to-image workflows, covering the full spectrum of Flux variants including Flux.1 Dev (SRPO), Flux 2 Klein 9B, and Turbo LoRA implementations. It automates the complex node wiring required for dual CLIP loaders, Flux-specific guidance, and distilled sampling patterns. By providing domain-specific expertise on sampler settings like ipndm/beta and hardware-specific optimizations like BF16 requirements for SRPO, it ensures users can generate high-fidelity 1MP images with optimal VRAM efficiency directly through Claude Code.

主要功能

01Hardware-aware VRAM optimization and resolution scaling for native Flux output.

02Support for Flux 2 Klein 9B distilled 4-step generation patterns.

0311 GitHub stars

04Automated configuration for Flux.1 Dev SRPO with BF16 precision matching.

05Advanced conditioning logic for CLIPTextEncodeFlux and FluxGuidance nodes.

06Integrated Turbo LoRA workflows for high-speed image synthesis.

使用场景

01Converting natural language image descriptions into production-ready ComfyUI JSON workflows.

02Rapidly prototyping different Flux model variants (Klein vs Dev) to compare generation speed and quality.

03Optimizing Flux pipelines for specific hardware constraints like 20GB or 24GB VRAM limits.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add artokun/comfyui-mcp flux-txt2img

For use in Claude.ai and ChatGPT

Download Skill