ComfyUI SmartSave

Automatically organize your ComfyUI generations by subject.

SmartSave is a ComfyUI custom node that identifies the primary subject in your prompt using a local Ollama model, then automatically saves images and videos into clean subject-based folders.

No manual renaming. No giant output folder full of unrelated generations. Classification stays local.

Automatic Image Organization

SmartSave can create both an archival master and a compressed sharing copy from the same generation.

Raw PNG
Raw/ExampleCharacter/ExampleCharacter_00001.png

Preserves ComfyUI workflow and prompt metadata.

Compressed JPEG
Compressed/ExampleCharacter/ExampleCharacter_00001.jpg

Creates a smaller sharing copy while storing recoverable prompt information in EXIF metadata.

Video Support

SmartSave Video provides the same automatic organization for video generations.

MP4 master
Raw Video/ExampleCharacter/ExampleCharacter_00001.mp4

Compressed WebM
Webm/ExampleCharacter/ExampleCharacter_00001.webm

A workflow PNG is saved alongside the master video, and audio is preserved when supplied.

Local AI Classification

SmartSave uses Ollama locally on your machine to identify subjects such as:

• People and characters

• Creators

• Usernames and handles

• LoRA trigger names

• Single-word identities

• Multi-word identities

The subject can appear anywhere in a long prompt.

The default model is llama3.2:3b.

SmartSave also includes conservative local fallback logic for cases where Ollama misses an identity. If there still isn't enough information to determine the subject safely, the generation goes to Unsorted rather than forcing a guess.

Auto-Sort Your Existing Library

SmartSave includes a full-library Auto-Sort utility for existing ComfyUI outputs as well as files SmartSave couldn't classify correctly during generation.

It can:

• Recover supported files from Unsorted and Misc

• Organize an existing large ComfyUI output library

• Repair recoverable SmartSave classification misses

• Preserve related generation numbering

• Match related workflow/media files

• Protect existing files from filename collisions

• Safely maintain an already-organized SmartSave library

Auto-Sort is preview-first.

Run:

auto_sort_preview.bat

Nothing is moved. Review the proposed changes first.

When you're satisfied with the plan, run:

auto_sort_apply.bat

Existing canonical SmartSave folders are trusted by default, preventing old or noisy metadata from unexpectedly reorganizing an already-clean library.

Designed to Fail Safely

SmartSave intentionally prefers Unsorted over a confident-looking wrong answer.

If a file doesn't contain enough evidence to determine its subject, it stays unresolved for later recovery instead of being forced into the wrong folder.

Auto-Sort also protects against overwriting an existing destination file.

[H2] Tested on a Real Library

The v1.0.0 Auto-Sort build was validated against a real 1,539-file image/video library.

After recovery:

• 1,539 files kept in place

• 0 additional moves proposed

• 0 files skipped

• 5 legacy WebM files intentionally remained Unsorted because they contained no usable recovery evidence

The included regression suite contains 15 automated tests, with 15/15 passing for v1.0.0.

Installation

Place the repository in:

ComfyUI/custom_nodes/ComfyUI-SmartSave

Install the dependencies from:

requirements.txt

Install Ollama and pull the default model:

ollama pull llama3.2:3b

Restart ComfyUI. The SmartSave nodes will then be available in your node list.

Source Code & Releases

Full documentation, source code, regression tests, example workflow, and current releases are available on GitHub:

https://github.com/phibbyrizz/ComfyUI-SmartSave

Current release:

https://github.com/phibbyrizz/ComfyUI-SmartSave/releases/tag/v1.0.0

Support SmartSave

SmartSave is free and open source. If it saves you time and you'd like to support continued development, you can support the project on Ko-fi:

Support SmartSave on Ko-fi