KSampler Node

Sample latent image using model and prompts.

Function

Core sampling node that generates images from noise.

Inputs

Input	Type	Description
model	model	Model from CheckpointLoader
positive	conditioning	Positive prompt encoding
negative	conditioning	Negative prompt encoding
latent	latent	Latent image to sample

Outputs

Output	Type	Description
LATENT	latent	Sampled latent image

Parameters

Parameter	Type	Description
seed	int	Random seed
steps	int	Sampling steps (20-50)
cfg	float	CFG scale (7-10)
sampler_name	dropdown	Sampler algorithm
scheduler	dropdown	Scheduler type
positive	conditioning	Positive conditioning
negative	conditioning	Negative conditioning

Sampler Types

Sampler	Speed	Quality	Notes
euler	Fast	Good	Recommended
euler_ancestral	Fast	Good	More variation
dpm++ 2M	Medium	Better	Smooth
dpm++ 2M Karras	Medium	Best	Consistent
ddim	Slow	Good	Traditional
UniPC	Fast	Good	New algorithm

Schedulers

Scheduler	Effect
normal	Standard
karras	Tighter at end
exponential	Smoother
simple	Basic

Parameters Explained

Steps

Value	Use Case
10-15	Quick preview
20-30	Standard quality
35-50	High quality

CFG Scale

Value	Effect
3-5	Creative, less strict
6-8	Balanced
9-12	Strict prompt following

Usage Example

CheckpointLoader
    ↓ MODEL → KSampler
    ↓ CLIP
CLIPTextEncode (positive) → KSampler
    ↓
CLIPTextEncode (negative) → KSampler
    ↓
EmptyLatentImage → KSampler
    ↓ LATENT
VAEDecode → SaveImage

Tips

1. Start with defaults: 20 steps, CFG 8
2. Increase for detail: More steps = more detail
3. Adjust CFG: Higher for strict following

Troubleshooting

Issue	Solution
Black image	Check model/VAE
Garbled output	Lower CFG
Too similar	Increase steps

Next Steps

• VAEDecode - Decode latent to image
• EmptyLatentImage - Create noise