If set to true, the safety checker will be enabled. Disabling it requires account authorization; unauthorized requests are always checked, and images flagged as unsafe are returned as black images.
default: true
nag_tau
number
The tau for NAG. Controls the normalization of the hidden state. Higher values will result in a less aggressive normalization, but may also lead to unexpected changes with respect to the original image. Not recommended to change this value.
n > 0 · default: 2.5
reference_end
number
The percentage of the total timesteps when the reference guidance is to be ended.
0 ≤ n ≤ 1 · default: 1
controller_guidance_reverse
number
The controller guidance (eta) used in the denoising process.Using values closer to 1 will result in an image closer to input.
0.01 ≤ n ≤ 3 · default: 0.75
max_shift
number
Max shift for the scheduled timesteps
0.01 ≤ n ≤ 5 · default: 1.15
nag_end
number
The proportion of steps to apply NAG. After the specified proportion of steps has been iterated, the remaining steps will use original attention processors in FLUX.
The controlnet unions to use for the image generation. Only one controlnet is supported at the moment.
default: []
nag_alpha
number
The alpha value for NAG. This value is used as a final weighting factor for steering the normalized guidance (positive and negative prompts) in the direction of the positive prompt. Higher values will result in less steering on the normalized guidance where lo…
0 < n ≤ 1 · default: 0.25
controller_guidance_forward
number
The controller guidance (gamma) used in the creation of structured noise.
The LoRAs to use for the image generation which use a control image. You can use any number of LoRAs and they will be merged together to generate the final image.
The controlnets to use for the image generation. Only one controlnet is supported at the moment.
default: []
num_inference_steps
integer
The number of inference steps to perform.
1 ≤ n ≤ 50 · default: 28
reference_strength
number
Strength of reference_only generation. Only used if a reference image is provided.
-3 ≤ n ≤ 3 · default: 0.65
base_shift
number
Base shift for the scheduled timesteps
0.01 ≤ n ≤ 5 · default: 0.5
nag_scale
number
The scale for NAG. Higher values will result in a image that is more distant to the negative prompt.
1 < n ≤ 10 · default: 3
seed
integer
The same seed and the same prompt given to the same version of the model
will output the same image every time.
nullable
guidance_scale
number
The CFG (Classifier Free Guidance) scale is a measure of how close you want the model to stick to your prompt when looking for a related image to show you.
The scale of the LoRA weight. This is used to scale the LoRA weight before merging it with the base model. Providing a dictionary as {"layer_name":layer_scale} allows per-layer lora scale settings. Layers with no scale provided will have scale 1.0.
default: 1
control_image_url*
string
URL of the image to be used as the control image.
path*
string
URL or the path to the LoRA weights.
preprocess
enum
Type of preprocessing to apply to the input image.
The percentage of the image to end applying the controlnet in terms of the total timesteps.
0 ≤ n ≤ 1 · default: 1
control_image_url*
string
URL of the image to be used as the control image.
conditioning_scale
number
The scale of the control net weight. This is used to scale the control net weight before merging it with the base model.
0 ≤ n ≤ 2 · default: 1
control_mode*
enum
Control Mode for Flux Controlnet Union. Supported values are: - canny: Uses the edges for guided generation. - tile: Uses the tiles for guided generation. - depth: Utilizes a grayscale depth map for guided generation. - blur: Adds a blur to the image. - pose: …
Control type of the image. Must be one of `spatial` or `subject`.
subject | spatial
scale
number
Scale for the control method.
0 ≤ n ≤ 2 · default: 1
control_method_url*
string
URL to safetensor weights of control method to be applied. Can also be one of `canny`, `depth`, `hedsketch`, `inpainting`, `pose`, `seg`, `subject`, `ghibli`
The scale of the LoRA weight. This is used to scale the LoRA weight before merging it with the base model. Providing a dictionary as {"layer_name":layer_scale} allows per-layer lora scale settings. Layers with no scale provided will have scale 1.0.