3. Script Language > AI - Artificial Intelligence Commands > SD - Stable Diffusion AI > SDL. - Stable Diffusion Local System > ! SDL

Possible Parameters

Previous Top Next

MiniRobotLanguage (MRL)

SDL. - Image to Image

Use an provided Image and change it using AI.

Intention

Here are all supported Parameters for Img2Img, using SDL. and the Automatic1111-API.

'***********************************

' SDL.-Sample

'***********************************

' Sample Script that shows how to use these Parameters.

$$PRO=hero,propaganda (0.6),epic,russia(0.4),sexy

$$EXT=

$$EXT+"prompt": "$$PRO",

$$EXT+"denoising_strength": 0.75,

$$EXT+"image_cfg_scale": 4,

$$EXT+"batch_size": 2,

$$EXT+"n_iter": 3,

$$EXT+"steps": 70,

$$EXT+"cfg_scale": 7,

$$EXT+"width": 512,

$$EXT+"height": 512,

$$EXT+"restore_faces": true,

$$EXT+"tiling": false,

SDL.Set Extra Parameter|$$EXT

$$IMG=?exeloc\TP01.jpg

$$TAR=?exeloc\TP02.png

SDL.Img2Img|$$IMG|$$TAR

SDL.Show Error

'MBT.No File generated.

ENR.

TP01 --> TP02_6

Details and Parameter description

Parameter	Default Value	Value Range	Explanation
init_images	["string"]	Array of strings	Initial image(s) for the model Base64 encoded
resize_mode	0	Integer	Mode for resizing images, •0: Just Resize •1: Crop and Resize •2: Resize and Fill
denoising_strength	0.75	0.0 - 1.0	Strength of denoising filter, higher means more AI Influence in the result
image_cfg_scale	0	Integer	Scale configuration for image
mask	"string"	String	Mask applied to the image Base64 encoded
mask_blur	0	Integer	Amount of blur applied to mask
mask_blur_x	4	Integer	X-axis blur for mask
mask_blur_y	4	Integer	Y-axis blur for mask
inpainting_fill	0	Integer	Fill value for inpainting
inpaint_full_res	true	Boolean	Whether inpainting is done at full resolution
inpaint_full_res_padding	0	Integer	Padding for inpainting at full resolution
inpainting_mask_invert	0	Integer	Inversion of inpainting mask
initial_noise_multiplier	0	Integer	Multiplier for initial noise
prompt	""	String	Text prompt for the model
styles	["string"]	Array of strings	Styles applied to the model
seed	-1	Integer	Random seed
subseed	-1	Integer	Subseed for randomization
subseed_strength	0	Integer	Strength of subseed effect
seed_resize_from_h	-1	Integer	Seed resize height
seed_resize_from_w	-1	Integer	Seed resize width
sampler_name	"string"	String	Name of the sampling algorithm
batch_size	1	Integer	Number of images processed in a batch
n_iter	1	Integer	Number of iterations
steps	50	Integer	Number of steps for the algorithm
cfg_scale	7	Integer	Configuration scale
width	512	Integer	Width of the output image
height	512	Integer	Height of the output image
restore_faces	false	Boolean	Whether to restore faces in the image
tiling	false	Boolean	Whether tiling is enabled
do_not_save_samples	false	Boolean	Flag to prevent saving of samples
do_not_save_grid	false	Boolean	Flag to prevent saving of grid
negative_prompt	"string"	String	Negative text prompt for the model
eta	0	Integer	Learning rate parameter
s_min_uncond	0	Integer	Minimum unconditional step size
s_churn	0	Integer	Churn rate
s_tmax	0	Integer	Maximum temperature
s_tmin	0	Integer	Minimum temperature
s_noise	1	Integer	Noise level
override_settings	{}	JSON Object	Settings to override
override_settings_restore_afterwards	true	Boolean	Whether to restore settings afterwards
script_args	[]	Array	Additional script arguments
sampler_index	"Euler"	String	Index of the sampler
include_init_images	false	Boolean	Whether to include initial images

Parameter Explanation

The denoising_strength parameter is used in the img2img/inpaint module.

It is a float that represents the strength of denoising applied to an image. According to a discussion on GitHub, the following values are recommended for the denoising_strength parameter2:

•If you are confident about the model you are using and want fewer modifications on the image, lower your denoising_strength to ~0.4. Switch your hires. fix upscaler to non-Latent upscalers (e.g. LANCZOS, ESRGAN). You will get some nice images.

•If you want more modifications and more potential, apply a higher denoising_strength (~0.55 to ~0.7). Pick a favorite Latent upscaler. You will get just what you want.

•For a general solution, don’t use hires. fix. Use extra->upscale to upscale your image (make your image sharper if possible), send the upscaled image to img2img. If you want more potential, using Inpaint and Inpaint Sketch is better. Process and finetune. You will get some awesome images (just wasting too much time compared with other alternative solutions).

The image_cfg_scale parameter is used in the img2img/inpaint module.

It is a float that represents the scale of the image configuration.

According to a discussion on GitHub, the image_cfg_scale parameter is added as an option to the XYZ Plot script on the img2img tab when an InstructPix2Pix model is loaded.
However, there is another discussion on GitHub where a user reported that the image_cfg_scale slider for InstructPix2Pix models provided in the img2img tab seems to not do anything.

The mask parameter is used in the img2img/inpaint module.

It is a binary image that represents the region of the image to be inpainted.

According to a discussion on GitHub, the mask parameter can be used in conjunction with the mask_blur option to apply a Gaussian blur on the section of the image that has the mask.

This can help to produce a smoother blend with the original image and force the regions inside to match up more smoothly with the parts at the edges which aren’t changing as much.

Example

'***********************************

' IRS.-Sample

'***********************************

Remarks

Limitations: