Load Image From Bucket

Remember how Input Image Bucket makes you type a UUID key? Fine for API workflows, awful for when you're sitting in the UI and just want to look at what's in your bucket. Load Image From Bucket is the same loader with the human interface: the key comes from a dropdown of every image asset in the bucket.

The outputs are identical to the Input variant - image (IMAGE), mask (MASK), and manifest (BUCKET_MANIFEST) - because under the hood it literally shares the loader. The only difference is how the key string gets into the node.

How the dropdown works

When the pack can enumerate the bucket (S3_ALLOW_KEY_LISTING=true, the default), the key input becomes a COMBO populated live from the /in_n_out/image/keys route, with a refresh button. It collapses all the objects of an asset into one entry - a batch saved as <key>_0.png, <key>_1.png, ... shows up as a single <key> - and it derives keys from raw objects, so even assets saved before manifests existed appear. There's deliberately no "select after generate" behavior, so a generation never yanks your chosen key out from under you.

If you set S3_ALLOW_KEY_LISTING=false (privacy or performance reasons), the node degrades to a plain text key input - identical to Input Image Bucket. Same node, two faces.

Two things that make it pleasant in practice: select a key and the image previews immediately, and there's a "choose file to upload" button that PUTs a local file into the bucket, sets the key, and adds it to the dropdown.

Why you'd reach for it

Interactive workflow building, mostly. If you're assembling a pipeline that consumes images already sitting in the bucket - say you batch-processed a folder of renders and now want to run a second pass on specific ones - this is how you grab them without a script. And since it resolves a whole batch under one key, choosing a key can feed an entire batch of frames into your pipeline as a single tensor.

Install

Install the pack via ComfyUI Manager (search In-N-Out Bucket) or clone + requirements:

cd ComfyUI/custom_nodes
git clone https://github.com/littleowl/ComfyUI_In-N-Out_Bucket.git
pip install -r ComfyUI_In-N-Out_Bucket/requirements.txt
cp ComfyUI_In-N-Out_Bucket/.env.example ComfyUI_In-N-Out_Bucket/.env

Restart, then look under buckets/image. Deps are boto3 + python-dotenv only.

Troubleshooting

An empty dropdown is the classic complaint, and it usually means one of two things: S3_ALLOW_KEY_LISTING is set to false (then you'd see a text box, not a dropdown), or the bucket couldn't be reached when the node loaded - the node registers with empty options rather than crashing, so a dead endpoint looks like "no images." Fix the endpoint/credentials, hit refresh. Also note the dropdown reflects the bucket at load time; after uploading with the button, the new key is added automatically, but assets uploaded by another machine may need the refresh button before they appear.