KJNodes/experimentalgenerated

Stable Zero123 Batch Schedule(StableZero123_BatchSchedule)

From a technical implementation perspective, this node receives an initial IMAGE and uses the CLIP Vision model to extract its visual features as conditioning guidance.

Stable Zero123 Batch Schedule

clip_vision
init_image
vae
positive
negative
latent
width
INT
height
INT
batch_size
1
interpolation
COMBO
azimuth_points_string
0:(0.0), 7:(1.0), 15:(0.0)
elevation_points_string
0:(0.0), 7:(0.0), 15:(0.0)
KJNodes

Node Function

From a technical implementation perspective, this node receives an initial image and uses a CLIP Vision model to extract its visual features as conditional guidance. Simultaneously, it utilizes a VAE to encode the initial image, generating a foundational latent representation. The most crucial part is that it allows users to define the change curves for azimuth and elevation at each step within the generation batch through two separate STRING parameters. It supports various interpolation methods to smooth these change points, enabling precise programming of the viewpoint transformation path.

Node Parameter Description - Stable Zero123 Batch Schedule

Connection Inputs (Inputs)

Parameter NameData TypeRequiredDefault ValueValue Range/OptionsDescription
clip_visionCLIP_VISIONYes--Connect the CLIP Vision model, used to encode the input initial image and extract visual features, thereby generating the conditioning information required for image generation.
init_imageIMAGEYes--Input an initial image. The node will generate subsequent image conditions for different viewpoints based on the features of this image.
vaeVAEYes--Connect the VAE model, used to encode the input initial image into the latent space, generating the starting latent representation for the subsequent diffusion process.

Control Parameters (Parameters)

Parameter NameData TypeRequiredDefault ValueValue Range/OptionsDescription
interpolationCOMBOYeslinear"linear", "ease_in", "ease_out", "ease_in_out"Specifies the interpolation method used to smooth the transition between defined key points for azimuth and elevation changes. Options include linear, ease-in (ease_in), ease-out (ease_out), and ease-in-out (ease_in_out), used to control the speed curve of the viewpoint change.
widthINTYes25616 - MAX_RESOLUTION (step: 8)Sets the width of the output latent image. The value range is from 16 to MAX_RESOLUTION, adjustable in steps of 8.
heightINTYes25616 - MAX_RESOLUTION (step: 8)Sets the height of the output latent image. The value range is from 16 to MAX_RESOLUTION, adjustable in steps of 8.
batch_sizeINTYes11 - 4096Defines the batch size for generation, i.e., the number of final output conditioning sequences and latent representations. The value range is from 1 to 4096.
azimuth_points_stringSTRINGYes0:(0.0), 7:(1.0), 15:(0.0)-Defines the key change points for azimuth (horizontal rotation angle) within the batch sequence using a specific string format. Points are separated by newlines. Format: "frame_index:(angle_value)," Example: "0:(0.0),\n7:(1.0),\n15:(0.0)\n"
elevation_points_stringSTRINGYes0:(0.0), 7:(0.0), 15:(0.0)-Defines the key change points for elevation (vertical pitch angle) within the batch sequence using a specific string format. Points are separated by newlines. Format: "frame_index:(angle_value)," Example: "0:(0.0),\n7:(0.0),\n15:(0.0)\n"

Output

Parameter NameData TypeDescription
positiveCONDITIONINGOutputs positive conditioning information, used to guide the diffusion model in generating the target image.
negativeCONDITIONINGOutputs negative conditioning information, used to guide the diffusion model to avoid generating certain content.
latentLATENTOutputs the initial latent representation encoded by the VAE, serving as the starting point for the diffusion process.

Usage Scenarios

In practical applications, this node is highly suitable for building workflows that generate 3D object rotation animations or multi-view datasets from a single image. For example, a user can connect an image loading node to provide an initial object photo, then connect this node's output to a sampler like Stable Diffusion or Zero123, ultimately generating a coherent sequence of images rotating around the object. This sequence can be used to create videos or for further 3D mesh reconstruction.

Notes

Please note that the "latent" output generated by this node is based on the encoding of the initial image. Its resolution is limited by the input parameters width and height, and the definition of the change sequence must follow a specific string format. Incorrect formats may lead to parsing errors.

The Stable Zero123 Batch Schedule node is from the ComfyUI-KJNodes node package.

Comments

Sign in with GitHub to join the discussion.

Loading comments…