CheckpointLoader Node
Load Stable Diffusion checkpoint models.
Function
Loads checkpoint files (.safetensors, .ckpt) containing model weights.
Inputs
| Input | Type | Description |
|---|---|---|
| model | (from node) | Model file selection |
Outputs
| Output | Type | Description |
|---|---|---|
| MODEL | model | Model for KSampler |
| CLIP | clip | CLIP model for text encoding |
| VAE | vae | VAE model for encoding/decoding |
Parameters
| Parameter | Type | Description |
|---|---|---|
| checkpoint_name | dropdown | Available model files |
Usage
Basic Setup
CheckpointLoader
↓ MODEL → KSampler
↓ CLIP → CLIPTextEncode → KSampler
↓ VAE → VAEDecodeWith SDXL
SDXL checkpoints have different structure:
CheckpointLoader (SDXL)
↓ MODEL → KSampler
↓ CLIP1 → CLIPTextEncode (positive)
↓ CLIP2 → CLIPTextEncode (negative) / combined
↓ VAE → VAEDecodeModel Files
Location
ComfyUI/models/checkpoints/Supported Formats
| Format | Extension | Recommended |
|---|---|---|
| SafeTensors | .safetensors | ✅ |
| PyTorch | .ckpt, .pth | Use if needed |
Tips
- Model selection: Choose based on your needs
- SDXL models: Use SDXL-specific checkpoints
- Refresh: Click refresh if model not showing
Troubleshooting
| Issue | Solution |
|---|---|
| Model not found | Check file location |
| Out of memory | Use fp16 or smaller model |
| Slow loading | Normal for large models |
Next Steps
- CLIPTextEncode - Encode prompts
- KSampler - Sample image
- VAE operations - Encode/decode