⚡ SmartSave for ComfyUI: Say Goodbye to Folder Chaos

🚀 NEW in v1.1.0: Full Video & Audio Pipeline Support!

SmartSave now handles video generations alongside images. It automatically encodes and splits renders into high-bitrate Master MP4s (Raw Video/<Subject>/) and optimized WebMs (Webm/<Subject>/), embeds workflows into companion First-Frame PNGs, and natively muxes optional AUDIO streams into both files.

🛑 The Problem:

You batch 300 generations overnight. You open your ComfyUI output folder the next morning... and it's an unreadable swamp of thousands of random images. Finding that one character or variation takes 20 minutes of manual scrolling and tagging.

💡 The Solution:

SmartSave brings zero-configuration, dynamic organization directly to your canvas. Powered locally by Ollama, it automatically reads your prompt, figures out the core subject, and cleanly routes your generations into isolated, numbered subject folders. No cloud keys. No data leaks. Zero manual sorting.

---

📸 What It Looks Like in Action

```text

📁 ComfyUI/output/

├── 📁 Batman/

│ ├── Batman_0001.png

│ └── Batman_0002.png

├── 📁 Cyberpunk Street/

│ └── Cyberpunk Street_0001.png

└── 📁 Portrait Study/

├── Portrait Study_0001.jpg

└── Portrait Study_0002.jpg

🌟 Core Highlights

• 🧠 Local LLM Intelligence: Integrates locally with Ollama (defaults to the featherweight, lightning-fast llama3.2:3b). It strips out filler words, lighting tags, and style noise to extract pure 1–3 word subject names.

• 📂 Dynamic Auto-Routing: Folders create themselves on the fly. Change the character in your prompt, and your output paths adapt instantly.

• 🔢 Bulletproof Numbering: Never lose work. It checks existing folder contents and increments indices (_0001, _0002) without overwriting older renders.

• 🖼️ Dual-Engine Export:

  • Lossless PNG: Full prompt/sampler metadata intact for drag-and-drop workflow retrieval.

  • Optimized JPG: Clean web-ready previews with adjustable quality compression.

• ⚡ Zero-Lag Session Caching: Re-running or iterating on identical prompts skips LLM evaluation entirely using local memory caching.

⚡ Quick Start & Installation

1️⃣ Install Custom Node

• Open ComfyUI Manager ➔ Install via Git URL

• Paste:

  Plaintext

  [https://github.com/phibbyrizz/ComfyUI-SmartSave](https://github.com/phibbyrizz/ComfyUI-SmartSave)
  

• Restart ComfyUI.

2️⃣ Run Ollama (Local)

SmartSave runs 100% locally. Ensure Ollama is installed, then open your terminal and pull the lightweight default model:

Bash

ollama pull llama3.2:3b

3️⃣ Load the Included Workflow

Download and drag the attached basic_workflow.json directly onto your ComfyUI window. Hit Queue Prompt, and watch your clean folders generate in real time!

⚙️ Node Parameters Cheat Sheet

• images: Connect directly from VAE Decode (IMAGE).

• positive_prompt: Connect string from your prompt primitive or text encoder.

• format: Choose between metadata-rich PNG or lightweight JPG.

• subfolder_prefix: Leave empty for output/[Subject]/, or add custom tags like Raw or Upscaled.

• quality: Slider (1–100) for JPG compression quality.

• ollama_model: Default is llama3.2:3b. Target any local model installed in your Ollama library.

🧹 Auto-Sort & Library Cleanup Tool (Included)

Tired of a cluttered output directory or messy multi-character subfolders? The repository includes a standalone post-processor (run auto sort.bat) that keeps your library clean and organized:

• Automated Subject Sorting: Uses a lightweight local LLM (llama3.2:3b via Ollama) to inspect image metadata and route unorganized images into clean, sequentially numbered subject folders.

• Smart Combo Merger: Automatically detects multi-character folders (e.g., Ariel and Selena or Taylor Swift & Selena Gomez), merges the files into the primary subject's folder without file collisions, and removes the leftover empty directories.

• Plug-and-Play: Double-click run auto sort.bat—it automatically detects standard ComfyUI embedded Python environments and your default output folder.

🤝 Open Source & Support

• 🐙 Source Code & Issues: GitHub Repository

• ☕ Found this useful? Help fuel updates and new node tools by buying me a coffee!

• ⭐ Don't forget to leave a review or drop a like on this post if it cleaned up your workflow!