img2prompt
img2prompt is a workflow for ComfyUI that allows users to generate detailed image captions and prompts from input images using state-of-the-art models. The generated prompts are saved to a file for later review and reuse.
This workflow integrates the following models:
WD14
MiaoshouAI
JoyCaption 2
Florence
Features
Converts input images into detailed and accurate captions or prompts.
Utilizes multiple popular image captioning models for diverse output styles.
Saves all generated prompts into a file for easy access and future use.
How to Use
Load an input image in ComfyUI.
Run the
img2promptworkflow.The workflow will process the image using the integrated models and generate prompts/captions.
All generated prompts are automatically saved to a file for later use.
Output
The generated prompts are saved in a structured text file. This allows you to:
Review and edit the prompts.
Use the prompts in other workflows or tools.
Models Overview
WD14: A powerful model for generating descriptive tags and keywords.
MiaoshouAI: Known for generating natural and context-aware captions.
JoyCaption 2: Provides highly descriptive and aesthetic captions.
Florence: A versatile captioning model that excels in creating detailed scene descriptions.
Acknowledgments
Special thanks to the creators of ComfyUI and the developers of the integrated models:
Screenshots
Enjoy generating amazing image prompts with img2prompt!
Description
First release
FAQ
Comments (16)
Never succeeded to use JoyCaption 2
I got it working, and also spend a day rewriting the InternVL2 comfyui one, but was not as good as I hoped so I removed it.
for what its worth, the JoyCaption 2 is amazing :)
@kiko9 I will let You know after testing. Thanks, for response
@AIStudio80 have you been able to get it working? I have to wait to get home to download this but I hate spending my limited time making things work instead of making pretty pictures.
@2018cfh Not yet. always gives error. i changed worflow to florence
@2018cfh Not yet. always gives error. i changed worflow to florence
@AIStudio80 ya I have one that uses Florence fine but Florence is no good at even remotely NSFW. That's why I been searching for a joy caption one
@2018cfh yes right. same here.after download 14+gb of models Joycaption still gives error
@2018cfhi searched a lot for solutions but still not successful
@AIStudio80 well thx for all the effort. I was using a joycaption online node..works great for about 10 or so images then just stops doing anything
Would probably help if the HF page of his want in Japanese or whatever it is
@2018cfh ok. I never checked online .
@AIStudio80 I'm at work now. I can't remember what the node was called...something NY/NJ joy caption online I think
I believe this is what I've been looking for. I have a captio ing flow I built but joy caption never seems to work. Does this contain any instructions on how and where to install whatever is needed for the captioning to work?