Since some wiktionaries limit the number of audios to insert in page.
Since :

• all audios are not of same audio quality
• we want regional diversity

Define best/recomended datasets to use

• Define it so it supersede existing audios.
• n>0 datasets can be defined, in a preferential order.

Take into account :

• audio quality : better microphone and/or speaker
• dataset's size : more audios the better
• gender : get both female and male voices
• regional accents : accents from countrysides
• age : child, adult, senior

See also : T275244