🎨 Your First ComfyUI Generation

From Zero to AI Art in 5 Minutes

Don't worry about that "Missing Models" dialog - we'll use the models you already have installed!

⚠️ Saw "Missing Models" Dialog? That's okay! ComfyUI opened an example workflow that needs models you don't have. We'll clear it and start fresh with what you DO have.
1

Clear the Workspace

First, let's start with a clean slate:

  • Click the X to close the "Missing Models" dialog
  • Press Ctrl+A (or Cmd+A on Mac) to select everything on the canvas
  • Press Delete to clear the workspace
  • You should now have a blank canvas!
💡 Quick Tip: ComfyUI works with "nodes" - think of them like LEGO blocks you connect together to build an AI image generator!
2

Add Your First Node - Load Checkpoint

This loads the AI model that will generate images:

  • Right-click on the empty canvas
  • Navigate: loaders → Load Checkpoint
  • A node will appear! In the node, click the dropdown
  • Select: sd_xl_base_1.0.safetensors
📦 Load Checkpoint
ckpt_name: sd_xl_base_1.0.safetensors
Outputs: MODEL, CLIP, VAE →
✅ What this does: Loads SDXL - a powerful AI that can create detailed 1024x1024 images from text descriptions!
3

Add Text Prompts (What You Want to Create)

Now tell the AI what to generate:

Positive Prompt:

  • Right-click → conditioning → CLIP Text Encode (Prompt)
  • In the text box, type something like:
"a majestic dragon flying over mountains at sunset, fantasy art, highly detailed, 8k"

Negative Prompt:

  • Add another CLIP Text Encode (Prompt)
  • This one tells the AI what to avoid:
"blurry, low quality, distorted, ugly, bad anatomy, watermark"
💡 Pro Tip: The more detailed your description, the better! Include art style, quality, and specific details.
4

Add Empty Latent Image (Canvas Size)

This sets your image size:

  • Right-click → latent → Empty Latent Image
  • Set dimensions:
📐 Empty Latent Image
width: 1024
height: 1024
batch_size: 1
⚠️ Important: SDXL works best at 1024x1024. Smaller = faster, larger = slower but more detail.
5

Add KSampler (The Generator)

This is the actual AI that creates the image:

  • Right-click → sampling → KSampler
  • Use these settings for your first try:
⚙️ KSampler Settings
seed: random (leave as is)
steps: 25 (higher = better quality but slower)
cfg: 7.0 (how closely to follow prompt)
sampler_name: dpmpp_2m
scheduler: karras
denoise: 1.0
💡 These are safe defaults! You can experiment later, but these settings work great for beginners.
6

Add VAE Decode (Convert to Image)

Converts AI data into actual pixels:

  • Right-click → latent → VAE Decode
  • No settings needed - just add it!
7

Add Save Image (See Your Creation!)

Finally, save the result:

  • Right-click → image → Save Image
  • In the node, you can optionally change the filename prefix
✅ Images save to: /data/comfyui/output/
8

Connect Everything Together!

Now the magic part - wire the nodes:

From Load Checkpoint:

  • MODEL → KSampler's model input
  • CLIP → Both CLIP Text Encode nodes
  • VAE → VAE Decode's vae input

From Positive CLIP Text Encode:

  • CONDITIONING → KSampler's positive input

From Negative CLIP Text Encode:

  • CONDITIONING → KSampler's negative input

From Empty Latent Image:

  • LATENT → KSampler's latent_image input

From KSampler:

  • LATENT → VAE Decode's samples input

From VAE Decode:

  • IMAGE → Save Image's images input
💡 How to Connect: Click and drag from an output dot (right side) to an input dot (left side). Lines will appear!
9

Generate Your First Image! 🎉

You're ready!

  • Click the Queue Prompt button (top right)
  • Watch the magic happen! You'll see:
    • Progress bar showing steps completing
    • Preview updating (if enabled)
    • Final image appears when done!
⏱️ First generation takes 30-90 seconds on your P5000. Subsequent ones are faster!
🎨 Success! Your image will appear in the Save Image node and be saved to /data/comfyui/output/
10

Try Different Prompts!

Now experiment! Just change the text in your CLIP Text Encode nodes and click Queue Prompt again.

Example Prompts to Try:

Sci-Fi: "futuristic cyberpunk city at night, neon lights, flying cars, rain, highly detailed, 8k" Fantasy Character: "portrait of an elf warrior, silver armor, magical forest background, fantasy art, detailed" Landscape: "serene mountain lake at dawn, mist rising, pine trees, reflection, photorealistic" Game Asset: "medieval sword, ornate design, glowing runes, white background, game asset, isometric view"

🎓 What's Next?

You've created your first AI image! Now you can:

📚 Advanced Workflows 📖 Model Reference
🚀 Performance Tips: