Mureka AI Describe Song (Official Edition)

Song-to-text is the quiet workhorse of the Mureka node set. You hand it an audio file and it hands you back a description of the song - genre, mood, structure, instrumentation. That sounds like a party trick until you realize what it unlocks: describing a reference track is exactly the step you need before recreating its vibe with a text-to-music node. This is the reverse-engineering node for people who think in workflows rather than prompts.

What it is

An "Official Edition" Mureka API node inside BizyAirPlus. It takes a song URL and returns a text analysis. It's the analysis half of the music workflow, paired with the T2A (text-to-audio) nodes that do the generation half.

How it works

You feed it a hosted audio URL - the tooltip is explicit: the URL of the song to process, MP3 or M4A. The Mureka service runs its recognition/analysis model on the clip and returns a structured description as text. The node is an output node (you don't need anything wired after it, though you can), so the result is mostly for you to read and repurpose.

The inputs and outputs

• audios - required. The song URL to analyze. This is an AUDIO-type input; in the BizyAir model world that means a hosted file, so if your audio only exists on disk you'll need it somewhere reachable by URL first.
• outputs - result (the description as STRING) and request_id.

api_config and skip_error round out the standard pack options.

Installing

Via ComfyUI Manager (search "BizyAirPlus") or:

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

Restart, toggle the BizyAirPlus button ON, enter your API key from bizyair.ai.

Where people get burned

The URL requirement is the #1 stumble - people try to drop a local file in and nothing happens. Get the audio hosted (or served from a reachable endpoint) and the AUDIO input does its thing. Second, the description quality varies with the source: a clean studio MP3 describes far better than a low-bitrate rip. And while this node is cheap compared to generation, it's still a metered API call - batch your analysis instead of re-running it on the same track.