zhenzhen-flowmusic-lowprice（9合1）

If you've ever wanted a music studio hiding inside a single ComfyUI node, this is it. Comfly_flowmusic_lowprice is a 9-in-1 wrapper for Flow Music (the "9合1" in the display name is literal), covering the whole lifecycle of a track: generate it, write lyrics for it, extend it, replace a section, cover it in a new style, split it into stems, download it, or cut a video clip to it. It bills through the domestic Seedance2 low-price API, so each operation is a metered call rather than a subscription.

The headline update in the pack's changelog added a lyria-3.5 version toggle - the Lyria 3.5 music-model series - alongside the default Flow Music models, so you get two flavors of the same pipeline.

The operation switch is the whole node

The required operation dropdown decides which inputs actually matter:

• flowmusic-generation - the main event. Describe the sound in sound_prompt ("风格或声音描述" per the tooltip), drop in lyrics if you have them, set title, bpm, and length (up to 240 seconds).
• flowmusic-lyrics - writes lyrics from prompt (max 3000 chars) instead of generating audio.
• flowmusic-upload-audio - takes local audio or a public audio_url (not both - the tooltip says so) and gives you back a clip_id.
• flowmusic-extend - pick a source with clip_id, then extend from extend_from_s by extend_s seconds (up to 164).
• flowmusic-replace - replace a section between start_s and end_s with strength controlling how hard the replacement lands.
• flowmusic-cover - redo a track in a new style, driven by sound_prompt.
• flowmusic-stems - split into separate parts; preset (simple/modern/player) picks the stem layout.
• flowmusic-download-audio - pull audio back out in mp3 or wav.
• flowmusic-video-clip - cut a video to the track.

Every operation is chained through clip_id - the output of one feeds the next, which is how you build "generate → extend → cover" without leaving the graph.

Outputs

This node is generous with outputs, and they're type-tagged so they land where you expect: audio1 and audio2 are AUDIO tensors, video is a VIDEO tensor, text carries the lyrics when you ran a lyrics op, clip_id is your chain token, then primary_url, result_urls, primary_path, result_paths, task_id, and response cover the hosted artifacts, local paths, and raw JSON. For a single-node workflow you'll mostly grab audio1 and clip_id.

Setup and the usual caveats

cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/Comfyui-zhenzhen

Restart. No models to download - this is pure API. You need a key from the domestic site (api.seedance.nz) wired into the Seedance2 low-price api_config, and note that music jobs are the ones where "oops, I left the default settings" turns into several billable calls - generation plus a couple of extensions adds up. Upstream 500s are usually transient, so a re-run beats a panic. And remember: any audio you upload leaves your machine and lands on a server you don't control - fine for a demo track, worth a second thought for anything proprietary.