Lazy Global Selector

The Lazy Global Selector is the control panel for the whole Vsaan212 video stack, and it's the node that makes the "lazy" in this pack's name make sense. It's one dropdown - T2V, I2V, FL2V, R2V - and a single string output that fans out to every other node that needs to know which mode you're in. Change the dropdown, and your image loader, your prompt engineer, your MiniMax conditioner and your model switcher all reconfigure at once.

If you've built a MiniMax H3 or LTX video workflow by hand, you know the pain this solves. Four modes means four different ways to wire a graph: text-to-video needs no reference images, image-to-video needs a first frame, first+last-frame video needs two, and reference-to-video needs a whole different model and audio encoder. Without a shared mode string, you set the mode in five separate nodes and hope you didn't miss one. This node makes the mode a value that lives in one place.

How it works

The mechanism is almost insultingly simple: the node takes your dropdown pick and outputs it as a plain STRING on global_Selector_output. That's it - there's no logic in here. The cleverness is downstream, where the pack's nodes all speak the same dialect:

• Lazy Image Loader takes the mode via its workflow_role + global_selector_input and only emits the image when the mode actually needs it (a first-frame loader goes silent in T2V mode).
• Lazy-subject-and-scene-automation routes the whole subject/scene automation through it.
• Lazy MiniMax All-in-One uses it to decide whether to run the T2V/I2V path or the R2V path.
• Lazy Model Switcher reads it to pick between the two MiniMax UNETs.
• LazyPrompt's Prompt Engineer takes it as selector_Out so the LLM knows what it's writing for.

The string tolerates a lot of spellings - the pack normalizes aliases like t2va, fl2va, ref2va and fflf down to the canonical modes, and it can also eat the tagged [Workflow] blob that the automation node emits. You can drive this node's consumers with something other than this node if you want; the point is you usually don't have to.

The "lazy" part is what makes the fan-out cheap. Downstream nodes declare their unused branches as lazy inputs, so when you pick T2V, ComfyUI doesn't bother executing the branches for R2V at all - the graph is fast to reconfigure because the dead paths are skipped, not just ignored.

The input and output that matter

• workflow_type - the dropdown. T2V (text-to-video), I2V (image-to-video), FL2V (first + last frame), R2V (reference-to-video). Defaults to I2V.
• global_Selector_output - the mode string. Wire it to any node with a selector / selector_in / global_selector_input input in this pack.

There's a lightweight read on this node: it's just a string holder with extra steps, and you could absolutely type "R2V" into each consumer manually. But that's exactly the repetition this family of nodes exists to kill. One authoritative source for the mode is the difference between a workflow you can hand to someone else and a workflow only you understand.

Installing it

This node ships in vsaan212/Vsaan212-workflow-utilities (MIT). Install via ComfyUI Manager - search Vsaan212-workflow-utilities - or clone:

cd ComfyUI/custom_nodes
git clone https://github.com/vsaan212/Vsaan212-workflow-utilities

Restart ComfyUI, then find Lazy Global Selector under the vsaan212/automation menu. No models to download, no API keys, no external services. The heavy Python deps in this pack belong to LazyPrompt's local LLM stuff, not to this node.

The honest caveat

It only earns its keep if you're invested in this pack's ecosystem. If your graph is a single MiniMax sampler, this is overkill - just set the mode on the conditioner. The payoff comes when you're running the full stack (image loader + automation + prompt engineer + MiniMax), where one dropdown genuinely replaces five.