ComfyUI Auto Tagger
Advanced metadata extractor for AI images. Supports ComfyUI (with deep workflow tracing), A1111, and Civitai. Effortlessly organizes prompts and model info into searchable Eagle tags and notes.
ComfyUI Auto Tagger for Eagle
Advanced metadata extractor for AI images. Supports ComfyUI (with deep workflow tracing), A1111/Forge, and Civitai.
ComfyUI Auto Tagger automatically extracts generation parameters from AI-generated images and organizes them into Eagle Tags and Notes. It doesn't just read metadata; it understands it.
[Features]
Multi-Engine Support: Seamlessly parses metadata from ComfyUI, Stable Diffusion WebUI (A1111/Forge), and Civitai-generated images.
Deep Workflow Tracing: Specifically optimized for ComfyUI. It analyzes the internal node graph to extract the actual parameters used for the final image, ignoring unused branches.
Format Flexibility: Supports both PNG and WebP files.
Smart Tagging & Notes: Converts prompts, seeds, and models into searchable Eagle tags or detailed structured annotations in the Notes section.
Heuristic Error Detection (New): Detects "suspicious" nodes with missing inputs to ensure metadata accuracy even in complex or broken workflows.
Batch Processing: Efficiently handles multiple images with a visual progress bar and "Force Delete" functionality.
[Advanced Workflow Trace Logic]
Unlike simple metadata readers, this plugin dynamically analyzes the ComfyUI node graph.
Path-Based Extraction: It traces the execution path from the output nodes back to the source. This ensures that in "All-in-One" workflows, only the parameters that actually contributed to the image are recorded.
Multi-Stage Support: Captures data from every generation step, including KSamplers, FaceDetailers, and HiresFix, providing a complete "recipe" for reproduction.
Accurate Parameter Retrieval: Even with complex custom nodes like "Lora Loader Stack" or Reroute/Primitive nodes, the logic recursively resolves values to ensure accurate retrieval.
Full Audit in Notes: Every detected generation step is recorded in detail within the Eagle Notes section, providing a precise "recipe" for reproduction.
[Heuristic Node Detection (v1.3.3+)]
To handle the complexity of custom nodes and experimental workflows, the plugin now includes an intelligent detection system.
Input Validation: Automatically identifies nodes with missing required inputs that might lead to incomplete metadata.
Interactive Decisions: When a "suspicious" node is detected, you can choose to include it, exclude it, or have the plugin ask you via a dialog.
Efficiency: Use Shift + Click in the decision dialog to apply your choice to all remaining images in the batch.
[Customizable Settings]
You have full control over how your library is organized:
Suspicious Node Handling: Choose between Ask, Include, or Exclude to manage how missing metadata is handled.
Extraction Targets: Selectively import Checkpoints, LoRAs, Prompts, Seeds, Steps, CFG, and more.
Output Destination: Independently toggle Tags and Notes to match your workflow.
Process Items per Batch: Adjust the batch size for optimal performance on your machine.
[How to Use]
Select one or more AI images in Eagle.
Right-click and choose Plugins > ComfyUI Auto Tagger.
Configure your settings (First-time setup recommended).
Click Start Tagging.
🎉 New: Suspicious Node Detection: Implemented a heuristic detection system to identify nodes with missing required inputs in ComfyUI workflows.
Adds configurable handling modes: Exclude (default), Include, or Ask via a dialog.
The dialog displays affected generation steps, and Shift+Click allows applying the decision to all remaining images.
✨ Improvements:
Enhanced workflow validation to detect non-executable nodes before processing.
Improved base sampler detection for workflows using DetailerForEach nodes.
Better handling of complex multi-stage refinement workflows.
🐛 Bug Fixes:
Fixed distance calculation for samplers that process images instead of latents.
Minor UI and stability improvements.
🎉 New Features
A1111 & Forge Support: Official support for Stable Diffusion WebUI (A1111/Forge) generated images. Extracts prompts, seeds, and models from both PNG and WebP formats.
Civitai Integration: Full support for images generated directly on the Civitai platform.
Force Delete Mode: Quickly remove all tags/notes from selected items without analysis by holding Shift + Clicking the "Delete Info" button.
✨ Improvements
Advanced Workflow Tracing (ComfyUI): Enhanced logic for complex img2img and multi-stage workflows (HiresFix, FaceDetailer). It now accurately identifies the primary generation parameters by tracing VAEEncode chains.
Expanded LoRA Detection: Improved support for various LoRA loaders, including "Lora Loader Stack (rgthree)" and standard A1111 LoRA hashes.
Cleaner Annotations: Refined the Note (Annotation) format to skip empty labels or unnecessary headers when specific metadata is missing.
🔧 Internal
Robust Engine: Completely refactored metadata parser architecture for faster and more reliable extraction.
Stability: Added a comprehensive test suite with 200+ tests to ensure stable performance across different generation environments.
Major update with improved accuracy and new features.
New: Added support for writing metadata to the Notes section.
New: Added a Settings UI to filter which metadata to import (e.g., ignore Seed, keep LoRA).
Improved: Enhanced parsing logic to support complex workflows (Reroute/Primitive nodes) and API-format prompts.
Improved: Implemented batch processing with a progress bar for better performance.
Fixed: Added full support for WebP files.
Fixed: Resolved issues with dark mode and multilingual display.
