Stable Diffusion 1.5 Default
Learn the core ComfyUI image pipeline with Stable Diffusion 1.5 at 512 × 512.
TESTED on NVIDIA GeForce RTX 5080 - UNCLASSIFIED
v1-5-pruned-emaonly-fp16.safetensors
Setup & downloadLOCAL AI, A CLEAR STARTING POINT
Find ComfyUI workflows for your GPU.
Explore image, video, and audio workflows with clear hardware requirements, setup instructions, and troubleshooting guidance.
50 workflows · 19 guides · JSON downloads without an account
BEST PLACE TO START
Beginner-friendly starting points
4x ESRGAN Upscale
Upscale an input image with the Real-ESRGAN x4plus model.
TESTED on NVIDIA GeForce RTX 5080 - UNCLASSIFIED
RealESRGAN_x4plus.pth
Setup & download
Flux.1 Dev — Text to Image
Generate images from text with Flux.1 Dev using separate diffusion, text encoder and VAE loaders.
TESTED on NVIDIA GeForce RTX 5080 - UNCLASSIFIED
flux1-dev.safetensors
Setup & downloadVERIFICATION STANDARDS
NeuralDrift 5-Level Validation Framework
Unlike indexes that scrape unverified graphs, NeuralDrift evaluates compatibility through a rigorous 5-level hierarchy. 37 workflows have recorded physical execution tests; the others are marked with conservative catalog estimates.
LEVEL 1JSON Schema Valid
Downloadable JSON is structurally valid with intact node bindings and no syntax corruption.
LEVEL 2Graph Connectivity Valid
Internal links, inputs, and outputs form a complete, cyclic-free execution DAG.
LEVEL 3Dependencies Identified
All models, diffusion weights, VAEs, and custom nodes inventoried with exact filenames.
LEVEL 4ComfyUI Graph Import
Loads into an active ComfyUI environment without red missing-node alerts.
LEVEL 5Execution & Semantic Output
Completes a physical hardware run, producing an inspectable output artifact.
EXECUTION-TESTED
Verified runs on physical hardware
SDXL Sketch to Image
Reinterpret a sketch using SDXL image-to-image sampling. This graph does not include ControlNet.
TESTED on NVIDIA GeForce RTX 5080 - UNCLASSIFIED
sd_xl_base_1.0.safetensors
Setup & download
SD 1.5 Object Removal (Inpaint)
Fill a masked region with an SD 1.5 inpainting model and a descriptive prompt.
TESTED on NVIDIA GeForce RTX 5080 - UNCLASSIFIED
sd-v1-5-inpainting.ckpt
Setup & download
SDXL Batch ×4 Variations
Generate a batch of four SDXL images at 1024 × 1024.
TESTED on NVIDIA GeForce RTX 5080 - UNCLASSIFIED
sd_xl_base_1.0.safetensors
Setup & downloadIMAGE GENERATION
Core text-to-image architectures
Flux.1 Schnell — 4-Step Fast
Explore a four-step Flux.1 Schnell text-to-image graph with separate model loaders.
ESTIMATED 8 GB VRAM
flux1-schnell.safetensors
Setup & download
SDXL Batch ×8 Variations
Generate eight SDXL variations at 768 × 768. Images are saved individually.
TESTED on NVIDIA GeForce RTX 5080 - UNCLASSIFIED
sd_xl_base_1.0.safetensors
Setup & download
SDXL Base + Refiner
Generate an SDXL image with a base pass followed by a refiner pass.
ESTIMATED 8 GB VRAM
sd_xl_refiner_1.0.safetensors
Setup & downloadIMAGE EDITING & CONTROL
Inpainting, ControlNet, and img2img
SDXL Image-to-Image
Transform an input image using SDXL image-to-image sampling.
TESTED on NVIDIA GeForce RTX 5080 - UNCLASSIFIED
sd_xl_base_1.0.safetensors
Setup & download
SDXL Inpainting
Use a masked input image with SDXL inpainting. Prepare the mask before running.
TESTED on NVIDIA GeForce RTX 5080 - UNCLASSIFIED
sd_xl_base_1.0.safetensors
Setup & download
ControlNet Canny — SDXL
Guide SDXL generation with a prepared Canny edge image and matching ControlNet.
TESTED on NVIDIA GeForce RTX 5080 - UNCLASSIFIED
sd_xl_base_1.0.safetensors
Setup & downloadUPSCALING & ENHANCEMENT
Detail recovery and resolution scaling
4x ESRGAN Upscale
Upscale an input image with the Real-ESRGAN x4plus model.
TESTED on NVIDIA GeForce RTX 5080 - UNCLASSIFIED
RealESRGAN_x4plus.pth
Setup & download
4x Model Upscale — Real-ESRGAN
General 4x model-based spatial upscaling using Real-ESRGAN for photographic and digital images.
TESTED on NVIDIA GeForce RTX 5080 - UNCLASSIFIED
RealESRGAN_x4plus.pth
Setup & downloadVIDEO & ANIMATION
Generative video and camera motion
AnimateDiff Walking Character
Generate a short AnimateDiff sequence with ambient motion from a text prompt.
TESTED on NVIDIA GeForce RTX 5080 - UNCLASSIFIED
realisticVisionV60B1_v60B1VAE.safetensors
Setup & download
AnimateDiff Looping Particle Preset
Explore a short AnimateDiff sequence with a loop-oriented prompt. A seamless loop is not guaranteed.
TESTED on NVIDIA GeForce RTX 5080 - UNCLASSIFIED
dreamshaper_8.safetensors
Setup & download
AnimateDiff Product Rotation
Explore a product-rotation prompt with AnimateDiff. Exact camera rotation is not controlled.
TESTED on NVIDIA GeForce RTX 5080 - UNCLASSIFIED
deliberate_v6.safetensors
Setup & downloadAUDIO & MUSIC
Local AI audio synthesis
ACE Step 1.5 — Text to Song
Generate audio from musical tags and lyrics with the ACE Step 1.5 checkpoint graph.
TESTED on NVIDIA GeForce RTX 5080 - UNCLASSIFIED
ace_step_1.5_turbo_aio.safetensors
Setup & downloadYOUR FIRST RUN
Choose. Set up. Create.
Choose a workflow
Filter by memory, output type, model and difficulty. Review the estimate and dependencies before downloading.
Explore the catalog →Install the dependencies
Follow the setup guide, get the listed models from their sources, and check any custom nodes.
Read the installation guide →Run and troubleshoot
Load the JSON in ComfyUI. Keep the first run small and use the error guides if a model or node is missing.
Find troubleshooting help →GET UNSTUCK
Keep your setup moving
How to Install ComfyUI Correctly
Complete installation guide for ComfyUI on Windows, Mac, and Linux. Portable and Desktop methods with troubleshooting.
Read guide →Where Every Model File Goes
Complete folder structure for checkpoints, LoRAs, VAEs, and ControlNet
Read guide →Understanding and Fixing Workflow Errors
Solve node not found, black images, and type mismatches
Read guide →