ACEStep Audio Post Processor ⚡🅡🅞🅣🅘

Extend a song or repaint a section of it with ACE-Step, and you're left with two pieces of audio - the original and the freshly generated part - that need to meet somewhere. A hard cut at that join almost always sounds bad: a click, a jump in room tone, sometimes an outright pop. This node exists specifically to smooth that seam over with a crossfade.

Ships with ryanontheinside's ACE-Step extend/repaint tooling - it's the finishing step you reach for right after ACEStepExtendGuider or ACEStepHybridGuider has done its job.

How it works

It takes both audio clips plus the actual GUIDER object that generated the new section - that's the detail worth noting, since it means the crossfade isn't blind. The guider presumably knows where the boundary between old and new material sits, so this node can blend across exactly that region instead of you having to specify a timestamp by hand.

The inputs and outputs that matter

• original_audio and processed_audio - the before-and-after AUDIO clips.
• guider - the same GUIDER instance used to generate processed_audio. This has to be the actual object from that run, not a fresh one you build separately.
• crossfade_duration - 0 to 1 second, default 0.1. How long the blend window is.

Output is a single stitched AUDIO.

How to install it

Via ComfyUI Manager, search RyanOnTheInside. Manually:

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

then restart. No extra downloads - this operates on audio and a guider you've already produced elsewhere.

Common issues & troubleshooting

Feeding it the wrong guider. Since the node relies on the guider to locate the boundary, reusing a different or stale guider instance (say, from a previous run) will either error or crossfade the wrong region. Keep it downstream of the exact extend/repaint step in the same graph execution.

Crossfade too short still leaves a click. The 0.1s default is conservative. If you can still hear the seam, push crossfade_duration up toward 0.3–0.5s.

Crossfade too long smears things. On short clips or fast, percussive sections, a long crossfade can blur a transition that should be tight - if the join now sounds mushy instead of clicky, come back down.