Install current version the plugin from the downloads.
Obtain an API key for the model you want to use, or install Ollama instead:
Obtain Google Gemini API key (Google account is required)
Obtain ChatGPT API key (OpenAI account is required)
Configure API key in Lightroom plug-in manager. (Only if you use ChatGPT or Gemini)
Choose the model you want to use on the Lightroom plug-in manager.
Go to Library in Lightroom
Select some photos
Go to menu -> Library -> Plug-In Extras -> Analyze photo(s) with AI
If enabled a preflight dialogs pops up, where the settings for the batch run can be set/verified.
Also if enabled, a photo context dialog pops up for every photo, with a preview of the image and a text box to enter additional information for the specific image.
If enabled review before save for title, caption or Alt Text and the field is already filled with data, a review dialog will show up. There is not (yet) a dialog for reviewing keywords.
Wait it to fill your keywords, caption, title and Alt Text as selected.
If anything doesn't work as expected, send us a support request
It is possible to adapt the used prompt for the AI model to different use cases, photography genres.
For example portrait photography:
"You are a professional portrait photography analyst. You describe the model's attire, hair, eyes as detailed as possible"
The different prompts can added/edited/deleted via the Lightroom Plug-In manager.
Using Ollama Ai models is a very GPU intense process. So if you want to offload the AI processing to another machine with a faster GPU, you can do that by:
Installing and running Ollama on the "faster" machine.
Editing the Ollama base URL to point to the fast machine in the Lightroom Plug-In manager.
The Preflight dialog can be activated in the plugin settings.
It's a collection of settings that can be configured before each run. All but one of these settings can also be configured in the plugin settings.
The exception is the AI behavior value (it represents an AI parameter known in technical jargon as "temperature"). This determines how creatively the AI will approach content generation. We recommend values between 0 and 0.1 for the best and most consistent results. A low value (slider to the left) produces particularly good results when using Ollama.
There are two ways to provide the AI with some contextual information about a photo before processing. This can be helpful if there is information about the image that the AI cannot derive from the content. The two options are:
You have the set the dropdown next Metadata to "All Plug-in Metadata" to get view as on the screenshot. Afterwards there's a text input field called "Photo context" where the context can be entered in simple words.