ACE-Step 1.5 Denoising Trajectory EQ (Experimental) ⚡🅡🅞🅣🅘
ACEStep15GenerationSteering
[Experimental] Steers audio generation by running the model twice per diffusion step — once normally, once with channel-scaled input — and steering along the difference. This is NOT post-processing EQ. It changes what the model generates: timbral character, spectral balance, rhythmic activity, and harmonic content. The 64 channels are grouped into 6 semantically meaningful bands. All sliders range 0-2 with 1.0 as neutral. Internal sensitivity scaling normalizes each band so that equal slider movement produces roughly equal perceptual impact. BANDS: - foundation (ch 0-7, 32-39): Low-frequency energy. Bass and sub-bass body. - body (ch 8-15, 40-47): Broadband energy and fullness. Guidance acts as a brightness tilt. - texture (ch 16-23): Mid/presence definition. Highest timbral impact of any band. - balance (ch 24-31): Spectral balance. Inverse: attenuating boosts mids. - weight (ch 48-55): Sub-bass dominance and low-end mass. - air (ch 56-63): Brilliance/shimmer. Inverse: attenuating adds air. Also affects rhythm. OPTIONAL: Connect a Keystone Config node to fine-tune the 6 most impactful individual channels. guidance_scale controls strength. Positive steers toward the emphasis, negative steers away. Each step runs an extra model forward pass.
Pack: RyanOnTheInside
custom_nodes.ComfyUI_RyanOnTheInside
Inputs (12)
| Name | Type | Required |
|---|---|---|
| model | MODEL | required |
| guidance_scale | FLOAT | required |
| foundation_gain | FLOAT | required |
| body_gain | FLOAT | required |
| texture_gain | FLOAT | required |
| balance_gain | FLOAT | required |
| weight_gain | FLOAT | required |
| air_gain | FLOAT | required |
| effect_start_pct | FLOAT | required |
| effect_end_pct | FLOAT | required |
| temporal_mask | MASK | optional |
| keystone_config | KEYSTONE_CONFIG | optional |
Outputs (1)
| Name | Type |
|---|---|
| MODEL | MODEL |