What you can make
- Dialogues: at the café, at the doctor's, a job interview, with a voice per speaker.
- Vocabulary lists, each word or phrase its own file, ready for flashcard apps.
- Listening comprehension tests with questions and pauses for answers.
- Minimal pairs for pronunciation practice: words that differ by one sound.
- Graded readers read aloud, chapter by chapter.
How to set it up
- A project per course or book, and a script per unit, created in order so the download's folders follow the units.
- Cast native-sounding voices: record samples from native speakers who agree, with the accent you're teaching, and set each voice's Language. A voice keeps the accent of its sample, so a native sample matters.
- Give every speaker a name, and paste dialogues as
Ana: ¿Qué vas a tomar?. - Make two speeds: one cast with the Calm style for beginners, and one with Balanced for natural speed. Calm, short lines and extra full stops slow the delivery.
- Generate, check with a native speaker, save, and download.
An example dialogue
Ana: Buenos días. ¿Qué vas a tomar?
Luis: Un café con leche, por favor.
Ana: ¿Algo más?
Luis: Sí, una tostada. ¿Cuánto es?
Ana: Son tres euros cincuenta.
Note tres euros cincuenta rather than 3,50 €: writing numbers as words makes sure they're
read the way you're teaching them.
Matching files to exercises
Each line becomes a numbered WAV file, and Lines.csv lists each line with its speaker and text, which
makes a ready-made transcript for the answer key. For vocabulary, one word per line gives one file per flashcard.
Pauses for answers
For "listen and repeat" or test questions, the gaps are added when you join the files, so you choose them exactly: two seconds to repeat a word, ten to answer a question. See editing in Audacity.
Quality checks
- Have a native speaker listen to everything before learners do. AI voices can misplace a stress or read a rare word oddly.
- Each language has quirks worth knowing: written accents in Spanish, tones in Chinese, kanji readings in Japanese. See the language pages.
- Tell learners the audio is AI-generated, and credit the speakers whose voices you used.