MultiTrack Prompt Enhance To Project Apply

If you open a workflow and see a node whose category reads _easy_media/internal, you're looking at plumbing. MultiTrack Prompt Enhance To Project Apply is exactly that: the internal helper that easy multitrackPromptEnhanceToProject summons to finish its job. You almost certainly didn't place it, and you almost certainly shouldn't try to.

Here's what's actually happening. The project-level prompt enhancer walks every task segment in your timeline and asks an LLM to improve each prompt. For the cloud providers (MiniMax h3-context-ir, doubao, GLM) the answer comes back inline, one text string per segment, and the enhancer writes it straight into the segment. But the local llama.cpp path is different - it runs through a dynamically built subgraph inside ComfyUI, so the enhanced texts only materialize when that subgraph executes. That's where this Apply node enters: it collects the finished prompts, counts them against prompt_count, and applies them to the TRACKS_INFO so the project data has its updated prompts before generation ever starts.

What it actually does

Two real inputs, one odd trick:

• tracks_info - the project data whose task prompts will be replaced
• prompt_count - how many enhanced prompts to expect

The trick is that it also accepts dynamic inputs prompt_0, prompt_1, … up to prompt_count - 1 (the schema is set to accept whatever arrives). Its execute just gathers those strings, checks the count matches the number of task segments, and writes each one back into its segment's selected prompt field - respecting whether the segment is using prompt variant A or B. Outputs are the updated TRACKS_INFO and the assembled PROMPTS list, identical in spirit to what the main enhancer node returns on its own.

Do you ever need to touch it?

In practice, no. It's flagged dev-only in the source, which means it's hidden from the normal node search, and ComfyUI's graph-expansion machinery spawns it on demand when the llama.cpp path needs it. The failure modes people actually hit are around it, not in it: delete this node from a downloaded llama.cpp workflow and the workflow will error on validation until the enhancer rebuilds it, and if prompt_count disagrees with the real number of task segments you'll get a prompt-count mismatch error. If you see it sitting in your graph doing nothing wrong, that's the intended state - leave it alone and let it do its quiet job.

Installation

There's nothing separate to install. It ships inside ComfyUI-Easy-Media, so the shared pack setup applies - ComfyUI Manager ("ComfyUI-Easy-Media") or:

cd ComfyUI/custom_nodes
git clone https://github.com/yolain/ComfyUI-Easy-Media.git

restart ComfyUI, and make sure the local-LLM companion (ComfyUI-llama-cpp_vlm) is present if you plan to run the llama.cpp provider path. The node itself needs no configuration, because you're never meant to configure it.