Ollama has also started making more cloud models available. I have to say I'm finding Kimi-K2.5 to be really good at image descriptions. I've made the choice to subscribe to Ollama Pro for $20 a month to use this model for descriptions. Since cloud models are not gated by your local hardware, I'm getting excellent descriptions back in an average of 15 seconds per image on a wide range of equipment.
I recently generated descriptions for all my photos dating back to 2000 and it was something around 10,000 images and that used just 20% of the weekly limit from Ollama.
I know there are countless apps to get the occasional description. I use a bunch of them myself. I'm probably not 100% objective but in hunting around to solve my problem of describing hundreds and thousands of pictures, I never found a tool to do what I wanted, which is what lead me to start creating IDT.
IDT supports both command line interaction and a graphical app. I also added a fairly extensive user guide as a part of this 4.2 beta release.
Once you have things installed, getting descriptions to start with is really as quick as either:
1. entering idt workflow <image directory> in a windows or Mac terminal or
2. Loading the ImageDescriber app, choosing a directory of images and videos and then choosing process all from the processing menu.
From there you can customize pretty much everything about the image description process. Pick different models, AI providers choose from multiple predefined prompts or create some of your own and more.
This tool has been great for me to browse descriptions of so many photos and relive memories from over the years.
I'm sure there are some rough spots and I continue to improve the project.