ACE-Step 1.5 Denoising Trajectory Aggregator (Experimental) ⚡🅡🅞🅣🅘

This pack's Generation Steering node exposes six raw latent-channel-group gains - precise, but you need to already know the mapping to use them meaningfully. This node is the higher-level version of the same idea: four named musical knobs - brightness, articulation, energy, density - each one a pre-built "recipe" combining specific channel groups from that same steering system. The author is upfront that these are "hypothesized musical abstractions," and the node's own tooltip states plainly: "The labels are working names, not empirically validated descriptions." Treat the names as a starting intuition to test against your own ears, not a guaranteed mapping to what "brightness" or "energy" mean in a DAW.

How it works

Each control bundles multiple channel groups from the pack's underlying six-band system: brightness combines the two inverted groups (G1 and G7) plus keystone channel 13; articulation combines G3 with keystone channel 29, and inherits G3's own documented risk - boosting articulation past roughly 1.5 can cause pitch shifting, the same caveat that applies to G3 directly on the Generation Steering node; energy combines G0+G4, G5+G6, and keystone channel 56; and density combines G5+G6 (moving down) with G7 (moving up) - the tooltip flags this one specifically as subtle, meaning don't expect a dramatic swing even at the extremes.

Mechanically, this node works the same way as Generation Steering underneath: it runs the model twice per diffusion step and steers along the difference, so the same "changes what's generated, not just how it sounds" framing and the same extra-forward-pass cost apply here. guidance_scale is the master strength control and, like on Generation Steering, accepts a list of floats for temporal scheduling across the generation. effect_start_pct/effect_end_pct window where in the denoising trajectory the effect is active.

Notably, this node doesn't accept a keystone_config input the way Generation Steering and Latent Channel EQ do - the keystone channels it uses (13, 29, 56) are already baked into the four recipes rather than left open for you to tune individually. If you want that level of control, that's exactly the gap Generation Steering plus its own Keystone Config node is for.

Inputs and outputs

• model (required, MODEL) - the ACE-Step 1.5 model being steered.
• brightness, articulation, energy, density - the four macro controls, 0 to 2, neutral at 1.0.
• guidance_scale - overall steering strength and direction.
• Output - a single MODEL, wired into your sampler setup like any other patched ACE-Step model.

Installing it

Install through ComfyUI Manager (search RyanOnTheInside) or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/ryanontheinside/ComfyUI_RyanOnTheInside

Then pip install -r requirements.txt and restart ComfyUI. Needs an ACE-Step 1.5 checkpoint already set up - this node patches an existing model rather than shipping one.

Common issues

Expect slower generation than a plain guider setup - like Generation Steering, this runs an extra model forward pass per step, and that cost doesn't show up until you're mid-run wondering why things are taking longer than a normal ACE-Step generation.

The pack's own general guidance elsewhere in this family - stay in roughly the 0.75-1.25 range for most controls, treat 1.5+ as "risks artifacts" - is the sane starting discipline here too, especially for articulation given its documented pitch-shift risk at higher values. And given the author's own honesty that these labels are working names rather than validated descriptions, don't be surprised if "brightness" doesn't sound like what you'd expect from an audio EQ context - this is closer to an experimental hypothesis you're testing by ear than a settled control.