A media prompt combines instructions with information about a media item. For image-based tasks, AI Prompts provides separate buttons to copy the text and the image. The text preview is not a transcription of the image.
Selecting the media item
1. Open AI Prompts from the People or Media page, select Tools > AI Prompt Builder from the menu. Select a saved Media prompt, such as Caption Generator, Media Description, or Transcription & Interpretation. If needed, use Add pre-built prompts to add the starter collection.
2. Click Change. In Select media, find the item by its caption or filename. Check the file and caption shown on the right, then click OK.
3. Verify that the intended item is now named beside Change. If it still says <None selected>, choose an item before proceeding.
4. Read the text preview. Use Edit prompt to change your instructions and Data to Include. A picture's caption and description can help explain it, but they may not include the words written in the picture.
Copying the text and image
1. Click Copy prompt to clipboard to paste into AI. Switch to your chosen AI service and paste the text into its input area.
2. Return to RootsMagic and click "Copy image to clipboard" to paste into AI. Return to the service and paste the image if that service supports image input.
3. Before submitting, check that both the intended instructions and the intended image are present. Copying the image replaces the current clipboard contents, so paste the text before making the second copy.
RootsMagic's copy buttons do not submit anything to an AI service. You decide whether to share the text and image. Check the image itself for sensitive information, even if you excluded private information from the text data.
Example: requesting a census transcription
Select Transcription & Interpretation and use Change to choose the census media item. Review the preview: it may contain a caption and linked information, but not the census entries themselves. Supply the image as well as the text when you want the AI to read handwriting or printed entries.
For a more focused reusable prompt, write instructions such as: Transcribe the named household from the supplied census image. Preserve the column headings. Mark unreadable words as [illegible] and uncertain readings with a question mark. Do not fill gaps from assumptions.
Include a caption or description to identify the household or page. Make clear which words come from the document and which are your own notes. Ask the AI to copy the document's wording first, then explain it separately, so you can tell the difference.
Compare the response with the image line by line. Pay particular attention to names, ages, relationships, birthplaces, and column alignment. Keep uncertainty visible when recording your conclusions in RootsMagic.
If the AI cannot see the record
Check whether you pasted an image or only the prompt text. A caption or local filename does not transfer the document. Confirm that the selected image can be opened in RootsMagic and that the chosen AI service accepts images. For other media formats, do not assume that Copy image transfers an entire document, recording, or multi-page file.
