⭐ Star Show Everything
Star Show Everything — stop guessing what that red wire is carrying
- anything
- anything
- info
- name
Every ComfyUI user eventually hits the same wall: a node refuses a connection, or a workflow silently does something weird, and you have no idea what's actually flowing down that wire. Is that MODEL the checkpoint you think it is? Is that latent 512x512 or 512x768? Which LoRA did the loader actually grab? Star Show Everything is the debug node for all of it.
The pitch is in the name - connect anything to it and it tells you what it is. MODEL, IMAGE, LATENT, CONDITIONING, STRING, INT, FLOAT, MASK, whatever. It's the closest thing to a print() for your node graph, and it's the one helper from the StarNodes pack I'd argue is worth installing the whole pack for.
How it works
It accepts any input through a wildcard socket (a proxy type that matches everything), then does three things: builds a human-readable summary, previews image tensors inline, and passes the value through untouched. For models it digs into the object and pulls out a name - it checks a long list of attributes like model_name, clip_name, vae_name, lora_name, ckpt_name, repo_id - and walks nested objects like the patcher, CLIP, and VAE. For tensors you get shape, dtype, and device. The summary is capped so a giant model dump doesn't freeze your browser.
Inputs and outputs
anything- plug in any single value.anything(output) - the same value, unchanged. This is the killer feature: you can splice the node into an existing connection without breaking the wire. Whatever came in goes out, so downstream nodes keep working while you inspect.info(STRING) - the human-readable summary.name(STRING) - best-effort label extracted from the value (a model filename, for instance).
Because the input is passed through, you can leave a Show Everything permanently parked mid-graph as a live status display - handy when you're iterating on a complex workflow and want to see what a node is really producing at that point.
Installing
It ships in StarNodes, so:
cd ComfyUI/custom_nodes
git clone https://github.com/Starnodes2024/ComfyUI_StarNodes
cd ComfyUI_StarNodes
pip install -r requirements.txt
Restart, double-click the canvas, search star, and it's under ⭐StarNodes/Helpers And Tools. ComfyUI Manager users can search "Starnodes" instead of cloning.
Where people get burned
Two things. First, it's an inspection node, not a pure passthrough - it forces that branch of the graph to recompute, so don't leave a wall of them in a production run you're timing. Second, the inline previews land in ComfyUI's temp folder; on long sessions with lots of image inputs it can quietly fill up temp space. And if the info string shows something like "Unknown Model" for a loaded LoRA, that usually means the object doesn't expose a name attribute the node recognizes - the value is still fine, you just didn't get a label. The summary text is still right there to read.
Inputs (1)
| Name | Type | Default | Description |
|---|---|---|---|
| anything | * | Connect any data type — MODEL, IMAGE, LATENT, CONDITIONING, STRING, INT, FLOAT, etc. The value is previewed and passed through unchanged. |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| anything | * | The connected value, passed through unchanged. |
| info | STRING | Human-readable summary of the value (type, shape, dtype, model name, etc.). |
| name | STRING | Best-effort name/label extracted from the value (e.g. model filename). |