Skip to content

Images

Select a model with Vision capability before asking about an image. Text-only models cannot interpret image attachments.

  1. Choose Add images beside the message composer and select your images.
  2. Attach up to four images per message. Open a preview to inspect an image or remove an attachment before sending.
  3. Add a specific question: “What does the error in this screenshot mean?” works better than sending an image without context.
  4. Choose Send. The model on your configured server receives the image input and returns its answer in the chat.
An image attachment opened in LLMster's full-screen preview

Keep text large and readable in screenshots. Crop to the relevant area before attaching if a page contains tiny text. When using several images, identify them in your prompt, such as “Compare the first screenshot with the second.”

Model accuracy varies. Inspect the image yourself before relying on a transcription, identification, or comparison.

LLMster resizes image attachments on the device and stores chat attachments locally. Image input is sent to the configured LM Studio server for inference. It is not accurate to describe the attachment as never leaving the device.

If the model cannot process your image, confirm its vision capability, try a simpler image, and check feature troubleshooting.