toobusy MiniMax H3 Image Latent:
Creates an empty starting latent for a MiniMax H3 image workflow that uses a T=1 image VAE. It prepares a single-frame video component together with the audio component required by this latent format, giving a compatible workflow a structured starting point rather than an image that already contains visual content. This is useful when you need to initialize generation in the format expected by MiniMax H3; the node itself does not generate or decode an image.
toobusy MiniMax H3 Image Latent Input Parameters:
width
Sets the intended image width in pixels. The node uses this value to size the video portion of the latent at one-sixteenth of the requested width, so a larger width creates a larger starting latent and may require more memory in later workflow steps. The default is 1344; the minimum is 32, the maximum is 16384, and the control changes in steps of 32 pixels.
height
Sets the intended image height in pixels. The node uses this value to size the video portion of the latent at one-sixteenth of the requested height. Increasing it creates a larger starting latent and may increase the memory needed by downstream processing. The default is 768; the minimum is 32, the maximum is 16384, and the control changes in steps of 32 pixels.
toobusy MiniMax H3 Image Latent Output Parameters:
latent
A LATENT value containing an empty, zero-filled starting state for the MiniMax H3 format described by the node. Its video portion has one frame and spatial dimensions determined by the requested width and height; it also contains a fixed-size, zero-filled audio portion. Both portions are packaged together inside the latent’s samples value. This is a starting point for compatible processing, not a finished image, audible audio, or separately selectable video and audio outputs.
toobusy MiniMax H3 Image Latent Usage Tips:
- Start with the default 1344 × 768 dimensions when you want the node’s preset size, then adjust
widthandheightto match the dimensions planned for the rest of your workflow. - Use values in 32-pixel increments, as provided by the node controls. For example, 1024 × 1024 produces a one-frame video latent with a 64 × 64 spatial grid.
- If a workflow runs short of memory, try smaller dimensions first. This reduces the size of the video portion created here and can also reduce the workload of later steps.
- Connect
latentonly to a workflow step that accepts this combined video-and-audio latent format. A step expecting a conventional image-only latent may not be compatible.
toobusy MiniMax H3 Image Latent Common Errors and Solutions:
Width or height is outside the allowed range
- Explanation: The node’s controls accept values from 32 through 16384 pixels. A value outside that range may be rejected during input validation.
- Solution: Set both
widthandheightwithin that range, preferably using the controls’ 32-pixel steps.
Out of memory during latent creation or downstream processing
- Explanation: Larger dimensions create a larger video tensor. Creating it, or processing it in a later node, may exceed available memory. The exact error text depends on the device and workflow.
- Solution: Lower
widthandheight, then run the workflow again. Increase them gradually after confirming the smaller size works.
Incompatible latent or samples format in a downstream node
- Explanation: This node stores both video and audio components together in
samples. A downstream node that expects a conventional single-tensor image latent may not accept that structure. - Solution: Use a downstream workflow designed for the MiniMax H3 latent format, and check that each connected node supports the combined structure.
Device allocation or backend error
- Explanation: The node creates its zero-filled components on ComfyUI’s intermediate device. If that device is unavailable or cannot allocate the tensors, execution may stop with a device- or backend-specific error.
- Solution: Check that your ComfyUI installation and selected compute device are working, close other memory-intensive tasks, and retry with smaller dimensions.
