EasyLoader (SV3D)(sv3dLoader)
EasyLoader (SV3D) is a loader node from the ComfyUI-Easy-Use package, grouped under EasyUse/Loaders.
sv3d Loader
Description
EasyLoader (SV3D) is a loader node from the ComfyUI-Easy-Use package, grouped under EasyUse/Loaders. It prepares an SV3D image-to-3D workflow by loading a checkpoint and VAE, accepting an initial image, and building the latent/batch setup used for generating a sequence of views around the subject.
The node also exposes camera trajectory controls and an optional scheduler string, making it a single entry point for feeding downstream sampling and video/3D generation nodes.
Inputs
ckpt_name
COMBO. Selects the checkpoint model to load. The available options come from the checkpoints folder.
vae_name
COMBO. Selects the VAE used to encode/decode latents. The available options come from the VAE folder.
init_image
IMAGE. The source image used as the starting view for SV3D generation. This image is encoded into the latent space and used as the visual anchor for the generated camera path.
empty_latent_width
INT. The width of the empty latent used for the generated frames. Set this to match the desired output resolution for the SV3D sequence.
empty_latent_height
INT. The height of the empty latent used for the generated frames. Set this to match the desired output resolution for the SV3D sequence.
batch_size
INT, default 21, min 1, max 4096. The number of frames/latent samples to generate. SV3D commonly uses 21 frames for its standard orbital view sequence, so the default is set to 21. Larger values increase memory usage and generation time significantly.
interp_easing
COMBO, default linear. Options: linear, ease_in, ease_out, ease_in_out. Controls the easing applied to interpolation along the camera path:
linear: constant interpolation between frames.ease_in: starts slow and accelerates.ease_out: starts fast and decelerates.ease_in_out: accelerates at the beginning and decelerates at the end.
easing_mode
COMBO, default azimuth. Options: azimuth, elevation, custom. Selects the type of camera motion used for the SV3D sequence:
azimuth: orbits around the object horizontally.elevation: moves the camera vertically.custom: uses a custom camera/scheduler path.
scheduler
STRING, optional, default "". An optional scheduler string. When empty, the loader uses its built-in/default scheduling behavior. When provided, it can be passed through to downstream nodes that consume a scheduler value.
Outputs
PIPE_LINE
An EasyUse pipeline object that bundles the loaded model, VAE, and related generation settings for convenient connection to downstream EasyUse nodes.
MODEL
The loaded SV3D diffusion model, ready to be used by a sampler or other model-consuming nodes.
STRING
The scheduler string associated with this loader setup. This can be passed to nodes that expect a scheduler value or used for debugging/custom scheduling workflows.
Usage Notes
- This node is intended for SV3D image-to-3D generation, where the initial image is used as the anchor for a sequence of generated views.
- The default
batch_sizeof 21 matches SV3D's standard frame count. If you change it, make sure your downstream model and sampling setup support that number of frames. empty_latent_widthandempty_latent_heightshould typically be set to the same value as the input image dimensions, or to the native resolution expected by the SV3D model.easing_modeandinterp_easingtogether define how the camera moves around the object over the batch. For a simple orbit, keepeasing_modeasazimuthandinterp_easingaslinear.- The optional
schedulerinput is especially useful when you want external control over the camera/scheduling behavior instead of relying on the node's defaults.
Comments
Sign in with GitHub to join the discussion.