Suno 音乐生成与处理（34 合 1）

Suno doesn't have open weights - you will never download it, so the only way to get it into a ComfyUI graph is to call it. This node is that call, and then some: Suno_Music is a single node with an operation dropdown covering 31 documented Suno actions, from generation and lyrics to uploads, extend, cover, mashup, stems, sounds, voice creation, export, and analysis. The node's own test log claims all 31 reached their final success state, so the packaging is the real feature here: one node, one dropdown, a whole music API.

A note on what it actually is before you get excited. This is not Suno's official API wrapped by a ComfyUI partner node. Suno_Music lives in T8mars/ComfyUI_Seedance, a pack that talks to the author's own API aggregator (api.seedance.nz) - a Chinese reseller storefront that also proxies video, image, and audio models. That means you pay per call through that reseller, your audio and prompts leave your machine, and upstream moderation applies. Suno is closed and censored at the source; nothing the node does bypasses that. If your goal is local, controllable music, the community's actual answer is ACE-Step, which runs on a potato GPU and is where people go for instrumentals. Reach for this node when you specifically want Suno-quality vocals and lyrics inside a workflow.

How it works

Pick an operation and the node redraws itself to show only the fields that action uses. Generation (suno-generation) posts to /v1/music/generations; the rest hit /v1/music/generations/{action} and then poll the task until it finishes. Local AUDIO references get uploaded automatically - no image hosting or public URL needed on your end, though each audio_url slot accepts a public URL instead if you have one. The really useful bit is task chaining: the task_id output plugs straight into the next Suno node, so extend/cover/stems/export workflows are just nodes wired end to end.

Inputs and outputs that matter

• operation - the whole node. suno-generation, suno-lyrics, suno-upload, suno-extend, suno-cover-song, suno-inspo, suno-mashup, suno-upsample-tags, suno-sounds, suno-create-voice, suno-stems, and 19 more.
• prompt - the prompt, lyrics, or edit instruction for the current action.
• version - v3.5 through v5.5; only sent when the action supports it.
• task_id / task_id_2 - chain from a previous Suno node; the second one is for mashup.
• audio1–audio4 / audio_url1–audio_url4 - reference audio for upload, create-voice, or inspo. Import sources need at least 6 seconds, and don't fill a local audio and a URL in the same slot.
• custom (custom lyrics), instrumental, title, style, vocal_gender, tags, name, continue_at, start_s/end_s, duration_s, speed - all action-specific; the node hides the ones that don't apply.

Outputs: audio1/audio2 (decoded AUDIO), video (for the mp4-type actions), text (lyrics, tags, or analysis), primary_url/result_urls and primary_path/result_paths for every result, task_id for chaining, and response with the full JSON.

Install and setup

Same as every node in this pack - install once and you have all of them:

# ComfyUI Manager: search "ComfyUI Seedance" or "seedance", or:
comfy node install seedance

Or manually:

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

The only real dependency is requests. Then grab an API key from api.seedance.nz/console and wire it in - add a Seedance API Config node and connect its api_config output, or set SEEDANCE_API_KEY before launching ComfyUI.

Where people get burned

The two classic ones are forgetting the 6-second minimum on uploaded source audio, and assuming cover/mashup/sample/add-* actions run without a prompt - they don't. Mashup specifically needs two task_ids. And remember the economics: every run is a metered API call, so a 31-action music pipeline is a credit-burner by design. Keep skip_error handy for batch runs - on failure it hands back placeholder audio and an error response instead of killing the whole queue.