Ace Advanced Options

This is a niche within a niche: Sage Utils ships an audio branch for Ace Step 1.5, the music/sound model that turns text (and audio references) into conditioning for audio generation. Ace Advanced Options is the node that collects the sampling parameters for that process and bundles them into a single ADV_AUDIO_INFO output you plug into the actual Ace Step 1.5 Audio Encode node. Think of it as the "advanced settings drawer" broken out so the main encode node stays clean.

You'll almost certainly never touch most of these. The mechanism is simple: each input is stored into a small dict and passed along as one Advanced Audio Info output. Here's what each one does, because the tooltips say it better than I can paraphrase:

• language - the language for encoding, a combo listing languages (en, ja, zh, es, de, fr, and a long tail). Set it to match your tags/lyrics.
• cfg_scale (default 2) - how strongly the audio encoding follows the conditioning. This is the one you'll actually fiddle with; too low and the output drifts from your tags, too high and it gets stiff.
• temperature (default 0.85) - sampling randomness for the audio code generation. Up = more variety, down = more conservative.
• top_p (default 0.9) - nucleus sampling threshold.
• top_k (default 0) - Top-K limit; 0 means disabled.
• min_p (default 0) - minimum probability floor for sampling.

The output Advanced Audio Info feeds the adv_audio_info input on the Ace Step 1.5 encode node. It's optional there - skip this node entirely and the encode uses its defaults, which are sane for most uses. Where this earns its keep is when you're doing music-gen-style work and want reproducible control: the same settings bundled into one wire are easy to save across workflows.

Fair warning on the ecosystem side: audio-in-ComfyUI remains a fairly small corner of the community, and Ace Step specifically is an enthusiast pick. If you landed here from a workflow you downloaded that uses Sage Utils audio nodes, this is the fine-tuning knob - treat the defaults as correct until you have a concrete reason to change them. The one I'd start with is cfg_scale (2 is a good baseline; try 1.5 if output feels overcooked), then temperature if you want more variety. The rest are sampler-geek territory.

Install

Same pack as everything else: ComfyUI Manager → search Sage Utils → install → restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/arcum42/ComfyUI_SageUtils.git
cd ComfyUI_SageUtils && pip install -r requirements.txt

Restart ComfyUI. No extra model downloads from the pack itself - the Ace Step model weights are yours to source separately, and the only Python dependency is dynamicprompts. If you don't do audio generation, this node (and the whole audio branch) is safe to ignore; it costs you nothing until you wire it in.