Analyze Images for LM Studio Plugin for multimodal image analysis in LM Studio.
It allows multimodal models to pass local image files to themselves for analysis, including images found online and downloaded by other tools.
List Local Images - lists image files in the working directory.
Call this first to get valid filenames before calling Analyze.Analyze Local Image - analyzes a single local image using the currently
loaded vision model.Analyze Multiple Images - analyzes 2-8 images in a single model call.
Use for comparison, sequences, or multi-image questions.
Configuration
Max Response Tokens - hard cap on tokens per analysis call (default: 2048).Target Response Tokens - soft token target stated in the prompt (default: 512).Response Language - force analysis responses to a specific language
(e.g. "Russian"). Leave blank for model default.
Improvements (v0.2.0)
Added List Local Images tool - LLM can now discover available images
without guessing filenames.
Added Analyze Multiple Images tool - compare or summarize 2-8 images
in a single vision call.
Added plugin config: adjustable token limits and response language.
Improved error messages: model name shown on vision capability error;
no-model-loaded case handled gracefully.
preferFullImageIfThumb now picks the largest available full-res file
instead of the first extension match.Extended supported image formats: HEIC/HEIF, JXL, SVG, ICO.
Usage and Installation Just click Run in LM Studio .
The model can list local images, select valid filenames, and analyze one image or multiple images in a single call.
If no model is loaded, or if a non-vision model is loaded, the tool returns an explicit error explaining what to load next.
Works great with Duck-Duck-Go Reworked and improved Visit Website plugin LLM can use DuckDuckGo Reworked to find relevant URLs, optionally Visit Website to extract structured page content and images. Then it can list available local images and analyze them via Analyze Images.
Example