MD: Advanced Media Save

The default Save Image node has been doing the same job for years and never got a nickname. This is the version you switch to when you stop saving PNGs one at a time and start wanting a single node that handles stills and animations, embeds your workflow, and names files like a person instead of ComfyUI_00042_.

AdvancedMediaSave takes your IMAGE batch and writes it out as PNG, JPEG, or WEBP - or, if you pick one of the animation formats, compiles the whole batch into a single GIF, MP4 (H.264), or WEBM (VP9). One node, six formats. That's the whole pitch, and for batch-heavy work it genuinely removes a node and a half from your graph.

How it works

The filename side is where the polish shows. filename_prefix accepts Python strftime codes, so AMS_%Y-%m-%d becomes a dated folder prefix, and the node automatically appends a timestamp so you never overwrite a previous run. Keep the default or prepend a project name - the tooltip's suggestion of projectX_%Y-%m-%d is a good habit.

Formatting is straightforward. jpeg_quality (default 95) and webp_quality (90) are there when you want smaller files; webp_lossless defaults to true, which means quality-perfect WEBPs that are noticeably fatter. For animations, framerate defaults to 8fps - the classic GIF look - and video_quality (1-10, default 8) controls the H.264/VP9 encode. If you're saving a batch as MP4 for actual playback you'll probably want to nudge the framerate up past 8.

The metadata behavior is the ComfyUI-native part. save_metadata defaults to true, embedding the full workflow and prompt into the file - the "drag the image back in and the whole graph reconstructs" trick the ecosystem runs on. One gotcha the tooltip is honest about: very large workflows bloat file size, and past 256KB the node switches to a sidecar JSON instead. If you share images publicly and care about privacy, flip it off. custom_notes lets you add personal text to that metadata, but only embeds when metadata saving is on.

You get a file_path STRING out, which is handy for wiring into a text display or a downstream node that needs to know where things landed.

Installing it

Everything ships in MD Nodes (MDMAchine/ComfyUI_MD_Nodes), so this arrives alongside 80+ siblings. Easiest route is ComfyUI Manager - search for MD_Nodes and hit Install. Manual route:

cd path/to/ComfyUI/custom_nodes
git clone https://github.com/MDMAchine/ComfyUI_MD_Nodes.git
cd ComfyUI_MD_Nodes
pip install -r requirements.txt

Restart ComfyUI after. Python 3.10+, PyTorch 2.0+ with CUDA, and the pack is GPL v3 pure-Python source - no hidden binaries. Note the requirements file is heavy (audio DSP, LLM glue, and more), so a bare pip install -r requirements.txt is the safe move even though most of it is irrelevant to saving images.

Common issues

The one that bites: animation saving needs imageio, and the source warns it will silently disable GIF/MP4/WEBM if that import is missing. If your static saves work but animations fail, pip install imageio (and imageio-ffmpeg for MP4/WEBM) and restart. Everything else - the timestamp suffix, the sidecar metadata switch, the format list - behaves exactly as labeled.