Upload audio. A local model detects the genre. Toggle below to generate the artwork locally (tiny-sd) or remotely (LLM + Qwen-Image API).