💪TB | Tiny Upscaler (advanced)

Upscale an image using custom sampler and sigmas to improve the quality of the result.

By martin-rizzo·Created 2 years ago·Updated 12 months ago· 42

◄enabledtrue►

◄extra_noise0.6►

◄cfg4.0►

◄seed1►

◄mode▾►

◄scale_by3.0►

Category💪TinyBreaker/upscaler

Inputs (13)

Name	Type	Default	Description
image	IMAGE		The image to upscale.
model	MODEL		The diffusion model used to improve the upscaling quality.
positive	CONDITIONING		The positive conditioning to use for the upscale.
negative	CONDITIONING		The negative conditioning to use for the upscale.
sampler	SAMPLER		The sampler to use for improving the upscaled image. The `KSamplerSelect` node from ComfyUI can be connected here.
sigmas	SIGMAS		The sigmas of each step of the sampling process. Any sampler node from ComfyUI that has `sigmas` as output can be connected here.
vae	VAE		The VAE used to encode the image for the model.
enabled	BOOLEAN	true	If disabled, this node will pass the image through without alteration
extra_noise	FLOAT	0.60–1.5	The Classifier-Free Guidance scale balances creativity and prompt adherence. Higher values produce more rigid, artificial images while adhering more closely to the prompt. Lower values allow for more creative and natural results, but potentially with more errors and distortions.
cfg	FLOAT	4.00–15	The Classifier-Free Guidance scale balances creativity and prompt adherence. Higher values produce more rigid, artificial images while adhering more closely to the prompt. Lower values allow for more creative and natural results, but potentially with more errors and distortions.
seed	INT	11–18446744073709550000	The random seed used for creating the noise.
mode	COMBO		The upscaling mode (experimental).
scale_by	FLOAT	3.01.5–6	The factor by which to scale the image.

Name	Type	Description
image	IMAGE	The upscaled image.