Interactive picker
Which voice should your learners hear?
Five questions about your learners. You get the voice to look for in Kukarella's Voice Library and a one-minute test for your own script. Voice counts read from the catalogue on 1 October 2026.
How the picker decides
- Native voices per variety: Kukarella's voice catalogue, read on 1 October 2026 with the same list function the Voice Library uses. A native voice reads one language in the accent of its entry; voices that read many languages are not counted. The catalogue changes, so the numbers carry their date.
- Native first, for listening practice: in our own tests, a voice that reads many languages usually gives itself away within about 15 seconds, with a slip in stress or pronunciation.
- Learners need the accents they will meet. IELTS: "Different accents, including British, Australian, New Zealand and North American, are used." (ielts.org, read 1 October 2026). Every SIELE Spanish test mixes at least three varieties of Spanish (Instituto Cervantes, read 1 October 2026).
- Accent strength and familiarity affect listening comprehension, "even with quite light accents": 21,726 TOEFL test takers (Ockey and French, Applied Linguistics, 2016, abstract).
- Numbers, dates and abbreviations: in our own tests (September 2026), voices read them differently, even voices from the same maker. Test them on the voice you mean to use.
- Speed control: Kukarella's voice panel, 1 October 2026. It goes from 0.5x to 2x; it works on Microsoft and MiniMax voices, not on OpenAI or Qwen voices (our own tests and code).
- The AI assistant chooses voices from a fixed set of 22, by gender and age, not by the language of the script (Kukarella home page FAQ, 1 October 2026).
- One project, several voices: Text to Voice lets you add voices and give each paragraph its own.
- The picker stores nothing and sends nothing. It recommends where to look; you choose the voice by listening.