This ComfyUI workflow turns a short text idea into finished stereo audio using Stable Audio 3.0 Medium. You type a brief concept (e.g., "warm lo‑fi hip‑hop beat" or "cinematic whoosh impact"), the prompt is expanded for clarity and detail with a Qwen 3.5 2B BF16 text encoder, and the Stable Audio 3 generation node produces a high‑quality stereo clip. The audio is then written to disk by SaveAudioMP3 for easy preview and sharing.

Under the hood, a Subgraph handles prompt expansion and routing so you only need to provide a short input. The core generation is performed by the Stable Audio 3 node (type id 8b66c757-fe2f-4184-91f3-479a19deb565), which loads weights from the Comfy‑Org/stable‑audio‑3 repository and accepts key parameters like duration and seed for control and reproducibility. A MarkdownNote in the graph provides quick in‑UI guidance. The final waveform is passed to SaveAudioMP3, which saves a stereo file to your chosen path. This setup is ideal when you want fast, consistent audio ideation without spending time crafting long prompts.