In MS-Word, I use the MS-recommended method of keeping images and captions together by creating a text box and inserting the image and the caption within the text box. Text can flow around the image/caption and most everything works in Word AND in export to .pdf.
On import to Caliber v 3.29, WIndows 10, from .DOCX file and conversion to EPUB, every image/caption is duplicated. On investigation, any textbox content is duplicated. Other than the duplication which I can manually delete, the content is handled properly. This is in a book manuscript so the .docx file is over 100MB so I will provide only if needed.
Sample (textbox content starts at block_58):
planted into your brain and which would take the place of any individual neuron and have an identical function.</span></p>
<p class="block_59"> </p>
<p class="block_58">Consider replacing neurons with artificial equivalents built from electronic components such as transistors and diodes—perhaps with a microprocessor. Neuron Image by Quasar Jarosz [CC BY-SA 3.0</p>
<p class="block_59"> </p>
<p class="block_58">Consider replacing neurons with artificial equivalents built from electronic components such as transistors and diodes—perhaps with a microprocessor. Neuron Image by Quasar Jarosz [CC BY-SA 3.0</p>
<p class="block_36">Through our hypothetical, completely painless microsurgical techniques, we will remove a single neuron from your brain, measure its characteristics, and replace it with one of our artificial neurons which has been adjusted to fit perfectly. As we
Attach a DOCX file (create a small extract from the full file) demonstrating/
status incomplete