The audio engine, on iOS, requires a user input first to be initialized.
You should try to play a sound once your player has touched the screen (for example a sprite button) and then you likely could use speech synthesis.
The fact that the example are buggy on your device seems to indicate your device's support as the issue.
In doubt, please post a bug report to the GitHub issue tracker here:
github.com/Scirra/Construct-3-bugs
You must follow the bug report guidelines or your issue will be closed without investigation.