CSV Writer (Soze)

The write-side complement to CSV Reader - appends a row to a CSV file instead of reading one. No image or model output here; this is a pure logging node, useful for keeping a record of exactly what produced what across a big batch run.

Pair it with CSV Reader and you've got a genuinely useful loop: read a row of source prompts, generate, log the result - seed, prompt, output path - back out to a second CSV as you go. Come back later and you have a real audit trail of a hundred-image batch instead of a folder of numbered PNGs and a fading memory of what settings produced which one.

How it works

Every time it runs, it appends one row - up to ten values - to the CSV file at the path you give it. It's a sink: is_output_node is true and it has no outputs of its own, so its entire job is the side effect of writing that line to disk.

The inputs and outputs that matter

• csv_filename_path (required) - the file to append to.
• value1, value2 (required, STRING) - the first two columns of the row. You need at least these two.
• value3 through value10 (optional, STRING) - up to eight more columns if you need them.
• No outputs.

The node doesn't write a header row for you - decide what order you're putting things in (seed, prompt, filename, model, whatever matters to you) and keep track of it yourself, since nothing in the schema labels the columns beyond their generic valueN name.

How to install it

Via ComfyUI Manager: search "Quality of Life Nodes for ComfyUI", install, restart. Manually:

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

No dependencies beyond the base pack - plain file writing.

Common issues & troubleshooting

You want a running log of a big batch - seed, prompt, output filename - for later reference. This is exactly the use case. Wire your seed value, your prompt string, and your Save Image's output path (or Download URL's saved_path, if that's how you're saving) into three of the value slots, and you've got a spreadsheet of your whole run afterward instead of trying to reconstruct it from filenames.

Rows appear in the wrong column when you open the CSV later. Since there's no header and no per-slot label beyond valueN, this is almost always a discipline problem, not a bug - write down, outside the node, what you decided value1..value10 mean for this particular graph.

The file doesn't exist yet. Check whether the node creates a new file automatically or expects one to already exist at that path - if you get an error on a brand-new path, create an empty file there first, or point at a path in a directory you know exists.

The CSV keeps growing across restarts and you didn't want that. Since this is an append, not an overwrite, running the same graph again on a different day adds to the same file rather than replacing it - which is usually what you want for a log, but worth remembering if you were expecting a fresh file each session. Point at a new filename per run if you need a clean slate every time.