Images and PDFs

How to attach

• the Attach a file button in the input field, then Image or PDF document;
• drag a file into the window;
• paste an image from the clipboard (Ctrl+V / Cmd+V).

Limits per message:

Images

An image is sent to the model with the message, so the model must be able to read images. If the selected model can't, MindDuck warns you: The selected model cannot read images. Choose another model above the input field.

The first time for each provider MindDuck asks for permission: the images will go to that provider for analysis. You can refuse: the image is kept in the chat locally, but isn't sent to the model.

When answering a message with an image, the model also writes a short description of it and recognizes the text on it. This description is stored encrypted, and later you can find the image by its content. There is no separate paid request for this. Details: Image descriptions.

Click an image in the chat to open it full screen. The saved description and recognized text are shown there too.

PDFs

MindDuck extracts the text from a PDF and sends the text to the model, not the file itself. The original is kept encrypted on this device.

Before sending, the input field shows how many pages and characters could be extracted.

• There is a text layer: all good.
• Only some pages have text: MindDuck asks you to confirm: Use the partial text from this PDF.
• There is no text layer (a scan, photos of pages): such a PDF can't be sent: No readable text layer.

An attached PDF has a Create text document button: it turns it into a MindDuck document you can edit and search. See Documents.

Attachments on other devices

Image previews and PDF text move to your other devices, while the originals stay where they were added.