ScheduleToModel

Prompt Control's LoRA scheduling - turning <lora:name:0.8>-style tags in your text into LoRAs that actually fade in and out across the generation - runs through ComfyUI's own hook system under the hood. The README is specific about it: the pack "constructs a graph of LoraLoaders and CreateHookLoras as necessary." ScheduleToModel is the node that applies exactly that, but taking an already-built PROMPT_SCHEDULE instead of raw text.

How it fits into the pack

Most people reach for PCLazyLoraLoader ("PC: Schedule LoRas") - the README's headline node for this - which takes your text directly and does the parse-then-apply in one step. ScheduleToModel is what you use once you're building a pipeline out of the primitive schedule nodes instead: PromptToSchedule → maybe FilterSchedule or PCApplySettings → ScheduleToModel. The payoff is composability - your LoRA scheduling reads from the exact same filtered, settings-applied schedule as the rest of your pipeline, rather than starting over from a separate text field.

The inputs and output

Both inputs required, no optional knobs:

• model - the MODEL to patch, straight from your checkpoint loader (or wherever it's coming from upstream).
• prompt_schedule - a PROMPT_SCHEDULE containing whatever LoRA syntax you scheduled in your prompt text.

Output is a MODEL with the scheduled LoRAs applied as hooks, ready to wire into your sampler the same way you'd wire any patched model.

How to install it

• ComfyUI Manager - search "ComfyUI Prompt Control", install, restart.
• Manual - cd ComfyUI/custom_nodes && git clone https://github.com/asagi4/comfyui-prompt-control, restart ComfyUI.

Requires ComfyUI v0.8.0+.

Common issues & troubleshooting

Your model has no LoRA scheduling in it at all. Check that the PROMPT_SCHEDULE you're feeding this actually came from text that had LoRA tags in it - if PromptToSchedule parsed a prompt with none, ScheduleToModel has nothing to schedule and just passes the model through unchanged.

Weird interactions when you stack this with other hook-based patches. Because scheduling here rides on ComfyUI's native hook system, stacking multiple independent hook sources onto the same model - another scheduler, a hook-based caching node, anything else patching via hooks - can behave unpredictably. Keep hook-based patching to a single source per model where you can, and if you're debugging odd LoRA behavior, that's the first thing to rule out.

You're not sure whether to use this or PCLazyLoraLoader. If you're writing one prompt for one pass and just want LoRA scheduling, use PCLazyLoraLoader directly - it's simpler and it's what the README leads with. Reach for ScheduleToModel specifically when a PROMPT_SCHEDULE you've already built elsewhere in the graph (filtered, settings-applied, mask-attached) needs to drive LoRA loading too, and you don't want to duplicate that work from raw text a second time.