Treat it as one system rather than a module you drop in, because the audio content, storage, signal path, enclosure acoustics and production test all depend on each other. Start by settling finished minutes per language, trigger method, speaker and microphone needs, product form and power, since those determine the module and memory required. In a plush body the acoustics are the hard part, as filling and fabric sit between the speaker and the listener. Prove the result in the real product form rather than on a bench before committing to tooling.
Speakers and microphones
Sensitivity, placement, openings, amplifier, pickup and feedback are tested in the product form.







