As far as possible, all the recordings are made by real people, who are all native speakers of the language.
We do not use people who have learned the languages as second languages.
For example, 100% of the Thai words, phrases and dialogues (currently 12274 entries) are spoken by real people, all of whom are native Thai speakers.
We have started to use text-to-speech software for some of the languages where we do not have recordings by real people available. We believe that recordings by real people are still the ideal. Nonetheless, we also believe that the quality of audio generated by text-to-speech software is nowadays so good that it's better to have software generated audio than no audio at all.
Some examples of computer-generated audio are:
At the bottom of each entry, we clearly mark if the audio is computer generated:
You can then follow the link to see a list of all the entries which are computer-generated, e.g.
You can check the full list of who made each audio for each language at, e.g.