SongBloom Audio Saver

Every SongBloom workflow ends at the same wall: SongBloomGenerator returns audio but deliberately doesn't save it - its saved_path output literally reads "Not saved". That's this node's job. Give it the generated AUDIO tensor and it writes a file to disk, with sensible naming, a timestamp so you never overwrite a take, and a bunch of validation that most savers skip.

What makes it worth using over rolling your own save: it's genuinely defensive about garbage input. It checks the audio dict has the right keys, that the waveform is a tensor, that the sample rate is sane, and that you're not about to save silence or a clipped mess. It warns on silent audio (RMS below a threshold), on clipping (peak near 1.0), on very short output, and on very long output. When a generation has quietly gone wrong, this is the node that tells you before you've committed a bad file.

The inputs that matter

• audio (required) - the AUDIO from SongBloomGenerator (or any node producing one).
• filename (required) - base name, default songbloom_output. It gets a timestamp appended (songbloom_output_1699999999.flac) so repeat runs don't collide.
• format (required) - flac (default), wav, or mp3. FLAC is the safe choice for a lossless master; use MP3 only for quick exports.

The optional pair:

• output_dir - leave blank and it writes to ComfyUI's standard output folder. Set it to any absolute path if you want your music somewhere specific.
• quality - high / medium / low, default high. Mostly relevant for MP3 export.

Outputs are audio (pass-through AUDIO, handy if you want to save and preview in the same run) and file_path (STRING, the absolute path of what got written - wire it into a Show Text or an export node).

Install

Pack install, nothing special:

cd ComfyUI/custom_nodes
git clone https://github.com/xuchenxu168/SongBloom_ComfyUI
cd SongBloom_ComfyUI
pip install -r requirements.txt

ComfyUI Manager: search "SongBloom_ComfyUI". Restart and it's under SongBloom/Audio.

One honest gotcha: MP3 export depends on ffmpeg being available, and it's marked as an optional dependency in the pack's requirements - if format is set to mp3 and saving fails, that's the reason. WAV and FLAC don't need it, which is another argument for leaving the format on flac by default. And if you're running a SongBloomBatchProcessor, note it saves its own .flac files to the output folder directly - this saver is for single generations.