Managing translations for both text and voice over: Logic dead-ends

The tool I am working on can translate some text in different languages. You can also generate a voice-over of individual paragraphs in different languages.

So translations and voice-overs have two different entry points.

Say you add an extra paragraph (for which a translation has not been generated yet), you can still generate a voice-over but the system won't find the translation to read from so it will give you an error.

How should the user be warned? Upon generating the voice-over for the new paragraph should I tell him/her, "Please update the translations before proceeding?"

I have run into a series of dead-ends for which I don't just want to place a warning as a fix. I wonder how to define the logic above that guides the user through a process, yet allowing him/her freedom of exploration.