| license | Apache-2.0 |
| name | image-generation-workflow-engine |
| description | Build image generation pipelines with Stable Diffusion, FLUX, ControlNet, LoRA, and ComfyUI workflows. Activate on: image generation pipeline, ComfyUI workflow, ControlNet, LoRA training, diffusion model. NOT for: video generation (ai-video-production-master), image classification (computer-vision-pipeline). |
| allowed-tools | Read,Write,Edit,Bash(python:*,pip:*,npm:*,npx:*) |
| category | AI & Machine Learning |
| tags | ["image-generation","stable-diffusion","flux","comfyui","lora"] |
| pairs-with | [{"skill":"ai-video-production-master","reason":"Generated images serve as keyframes and style references for video pipelines"},{"skill":"computer-vision-pipeline","reason":"CV preprocessing (segmentation, depth) feeds ControlNet conditioning"},{"skill":"multimodal-embedding-generator","reason":"CLIP/SigLIP embeddings guide generation and enable style search"}] |
Image Generation Workflow Engine
Build production image generation pipelines with FLUX, Stable Diffusion 3.5, ControlNet, LoRA, and ComfyUI for automated creative workflows.
Activation Triggers
Activate on: "image generation pipeline", "ComfyUI workflow", "ControlNet conditioning", "LoRA training", "FLUX generation", "Stable Diffusion pipeline", "batch image generation", "img2img workflow", "inpainting pipeline"
NOT for: Video generation from images (ai-video-production-master), image classification or object detection (computer-vision-pipeline), or multimodal search embeddings (multimodal-embedding-generator)
Quick Start
- Choose model — FLUX.1-dev for quality, FLUX.1-schnell for speed, SD 3.5 for ControlNet ecosystem, SDXL for LoRA abundance.
- Design workflow — Text-to-image (simplest), img2img (style transfer), ControlNet (structural guidance), inpainting (targeted edits).
- Build pipeline — ComfyUI for visual node graphs, diffusers library for code-first, or API services (Replicate, fal.ai) for managed.
- Add conditioning — ControlNet (canny, depth, pose), IP-Adapter (style transfer), LoRA (fine-tuned concepts).
- Automate — Batch generation with parameter sweeps, quality filtering, and output organization.
Core Capabilities
| Domain | Technologies | Notes |
|---|
| Models | FLUX.1-dev/schnell, SD 3.5, SDXL, Kandinsky 3 | FLUX is 2025-2026 standard for quality |
| Conditioning | ControlNet (canny, depth, pose, segmentation), IP-Adapter | Structural and style guidance |
| Fine-Tuning | LoRA, DreamBooth, textual inversion | Custom concepts in 20 min on consumer GPU |
| Workflows | ComfyUI, diffusers (Python), A1111 | ComfyUI for complex multi-step; diffusers for code |
| APIs | Replicate, fal.ai, Together AI, HF Inference | Managed GPU, pay-per-image |
| Local | qwen-image-mps (Apple Silicon), CUDA, ROCm | M4 Max: FLUX.1-schnell in 4-8 sec/image |
Architecture Patterns
Pattern 1: ComfyUI Production Pipeline
[Load Checkpoint] ──→ [CLIP Text Encode] ──→ [KSampler] ──→ [VAE Decode] ──→ [Save Image]
│ │ │
FLUX.1-dev positive + negative steps: 20-30
or SD 3.5 prompts with weights cfg: 3.5-7.5
scheduler: euler
│
[ControlNet Apply] (optional)
│
canny/depth/pose
from reference image
