ComfyUI-Mickmumpitz-Nodes Introduction
ComfyUI-Mickmumpitz-Nodes is a collection of custom nodes for ComfyUI created by the author, Mickmumpitz. It brings together tools for video workflows, text selection, image sizing, and preparing character datasets for LoRA training.
In ComfyUI, a node is a step you connect to other steps in a visual workflow. This collection gives you additional steps for tasks that can otherwise take several separate workflows to manage. For example, you can work with video across multiple iterations, extract frames to use as context, or organize images and captions for a character dataset.
How ComfyUI-Mickmumpitz-Nodes Works
Think of a ComfyUI workflow as a production line: each node receives something, performs a task, and passes a result to the next node. The nodes in this collection serve different parts of that line.
- For video, some nodes help manage repeated generation passes, keep track of accumulated frames, route results, and resume an iterative workflow. Others extract context images or control frames, or join video segments.
- For text, selectors and batch-processing nodes help you choose or handle story and style strings within a workflow.
- For datasets, the Consistent Character Creator nodes help move images, captions, and face-area information through preparation and review steps.
You do not need to use every category together. Choose the nodes that match the task you are building.
ComfyUI-Mickmumpitz-Nodes Features
Iterative Video
These nodes support video generation across multiple iterations. They provide frame accumulation, routing, and resume support. This is useful when your process builds a video in stages and you need to carry results from one stage into the next, or continue an interrupted workflow.
Video Context
These nodes extract context images and control frames for video workflows. A context image can give another part of your workflow visual information from a clip; a control frame can provide a frame for a control-based step. Which frames you use depends on the connections and settings in your workflow.
Video Utilities
Video utilities include concatenation tools for joining video segments. Use them when your workflow produces separate clips that you want to bring together.
String Batch
Story and style selectors, along with string batch-processing nodes, help you manage text used in a workflow. For instance, you can organize alternative story or style inputs so that the appropriate text reaches a later step.
Utilities
This category includes resolution pickers, preprocessing settings, execution gates, and other helper nodes. Resolution pickers help determine canvas dimensions; gates help control whether a branch of a workflow runs.
MiniMax H3 Resolution is designed around the aspect ratio of a reference clip. It takes that clip’s aspect ratio and scales the canvas from there. This matters because H3 reframes rather than adding letterbox bars: a wide reference clip, for example, leads to a wide canvas instead of being placed inside a differently shaped frame with empty bands.
Consistent Character Creator
These nodes support LoRA dataset preparation. The collection includes batch loaders, a model-specific captioning Prompt Studio, an Ideogram 4 tagger, bounding-box converters, an interactive Dataset Reviewer, face-area batch routing, and a branch gate.
Together, they give you ways to bring dataset items into a workflow, work on their captions, review them, and route images based on face-area information. The author describes Prompt Studio as model-specific, so choose its captioning workflow with the model you are preparing data for in mind.
ComfyUI-Mickmumpitz-Nodes Models
The collection does not list a set of bundled generation models. Its documentation does identify an optional face-detection model for the Face Area Batch Splitter:
face_yolov8m.pt: An example face model the splitter can load when itsbbox_detectorinput is left empty. This option also requires theultralyticspackage. The model belongs inComfyUI/models/ultralytics/bbox/.
You can instead connect an UltralyticsDetectorProvider node from ComfyUI Impact Subpack to the splitter’s bbox_detector input. With that connection, the face-area nodes do not need the splitter to load its own model. SEGS Filter by Relative Area consumes Impact’s SEGS output.
MiniMax H3 and Ideogram 4 are named in relation to specific nodes or workflows; the supplied documentation does not say that their models are included with this collection.
Troubleshooting ComfyUI-Mickmumpitz-Nodes
Face Area Batch Splitter cannot load a face model
- Check whether the splitter’s
bbox_detectorinput is connected. - If you intend to supply a detector through the workflow, connect an UltralyticsDetectorProvider node from ComfyUI Impact Subpack.
- If you intend to leave
bbox_detectorempty, the splitter needsultralyticsand a face model such asface_yolov8m.ptinComfyUI/models/ultralytics/bbox/.
The optional package and model are needed only for the splitter’s own model-loading path. They are not required for the node collection to load.
SEGS Filter by Relative Area has no usable input
Check that your workflow supplies the node with Impact’s SEGS output. This filter is designed to consume that output, so connecting a different kind of result will not provide the face-area data it expects.
MiniMax H3 output does not have letterbox bars
This is expected. MiniMax H3 Resolution uses the reference clip’s aspect ratio to scale the canvas because H3 reframes rather than letterboxes. Check the shape of your reference clip if the resulting canvas is not the shape you intended.
Frequently Asked Questions
Do I need a face model to use the whole collection?
No. A face model is only relevant when you leave the Face Area Batch Splitter’s bbox_detector input empty and want the splitter to load its own detector.
Can I supply a detector from another node?
Yes. Connect an UltralyticsDetectorProvider node from ComfyUI Impact Subpack to bbox_detector.
Are these nodes only for video?
No. The collection also includes text-batch, resolution, preprocessing, workflow-gating, and character-dataset tools.
Learn More about ComfyUI-Mickmumpitz-Nodes
The ComfyUI-Mickmumpitz-Nodes repository is the source for the author’s project information. For an unfamiliar node, start with its category above, then look at its inputs and outputs in your ComfyUI workflow to see what it needs and where its result can go.
