Mureka BGM 背景音乐（2 合 1）

Mureka is a music-generation service, and this node is its background-music face: pick the v8 or v9 model, describe the track (or hand over an existing instrumental ID), ask for up to three results, and you get them back as an ordered list of AUDIO outputs. It's a deliberately tiny node - four inputs, five outputs - which makes it the easiest win in the T8mars/ComfyUI_Seedance pack's audio lineup. If your workflow needs a score behind a generated video and you don't want to fight a 9-operation music node, this is the one.

API wrapper, as always with this pack: key from api.seedance.nz, no local compute, requests only.

How it works

Simple async audio pipeline: build the request, submit, poll, download, decode to ComfyUI AUDIO. The one structural detail worth knowing is that results are ordered and kept - the node downloads every track the API returns in sequence order and hands you the full list, so n=3 means three usable tracks, not a single combined file. The audios output is a list type, which is how a single node output can carry multiple audio clips.

The inputs that matter

• model - Mureka v8 or v9 BGM.
• prompt - the music description. Or:
• instrumental_id - an existing Mureka instrumental ID. Here's the catch: exactly one of prompt or instrumental_id must be filled. Both, or neither, and strict validation rejects it. The tooltip and validation both spell this out.
• n - how many ordered results, 1 to 3.

Outputs: audios (the list of AUDIO), audio_urls (JSON array of URLs), audio_paths (local paths), task_id, response.

Installing it

ComfyUI Manager → search "ComfyUI Seedance", or:

cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/ComfyUI_Seedance.git
cd ../..
python -m pip install -r custom_nodes/ComfyUI_Seedance/requirements.txt

Restart, key from api.seedance.nz/console, Seedance API Config into api_config.

Common issues

• "provide exactly one of prompt or instrumental_id" - you left both empty or filled both. Pick one source.
• audios won't feed a single-audio node - it's a list output; if a downstream node expects one AUDIO, you may need a list-to-single step or to grab audio_paths and load from file. The n count drives how many you get.
• BGM doesn't match the brief - describe the vibe, tempo and instrumentation in prompt; short "sad piano" prompts give generic results, same as any music model.
• skip_error gives you silence - with skip_error on, a failed job returns 1 second of silence as a placeholder so your video render chain keeps going. If you suddenly have a 1-second silent track, a job failed - check response.

For a "I just need underscore music behind this clip" need, this is genuinely the least-effort audio node in the pack. If you need lyrics, covers, stems or a full song structure, the pack's Flow Music node is the 9-operation heavyweight for that.