Dictation audio generator for teachers
A dictation needs a voice that is slow, clear and patient: each phrase read once to listen, once to write, with a pause long enough for the slowest pen in the room. Reading it aloud live works, but it cannot be repeated for a pupil who was absent, given as homework or prepared in five languages.
SuperClonica turns the dictation text into audio with natural voices, free to test online and offline on your own computer. This guide shows how to split the text, how to slow the voice down and how to build a track with pauses for writing.
What a dictation track needs
A track can be played again for a pupil who missed the lesson, sent as homework or reused next year.
- A slower pace than normal speech, without sounding unnatural.
- Phrases of a few words, each read twice.
- Pauses long enough to write, growing with the length of the phrase.
- The whole text read once at the start or the end for checking.
Hear a phrase free online
On the page of any voice, paste a phrase of 30 to 100 characters, choose the language and press Voice it. In a few seconds you can hear whether the voice is clear enough for your class and download the phrase, with no account.
Prepare the text
For spelling dictations the voice reads exactly what is written, so double-check the text: a typo will be read too.
- Split the text into short phrases, one per line or per file, along the natural breaks of the sentence.
- Write punctuation the way you want it heard: a comma for a short breath, a full stop for a clear end.
- Write numbers as words, as pupils will write them.
- Check unusual words and names by ear before the lesson.
Build the track offline
Install the free SuperClonica engine: python -m pip install superclonica-cpu. Voice each phrase slowly, for example superclonica --text-file phrase_03.txt --voice F2 --lang en --speed 0.8 --seed 5 --output phrase_03.wav; --speed accepts values from 0.7 to 2.0, the default is 1.05, and 0.8 to 0.9 is a calm dictation pace. Then join the pieces: in Python, numpy.concatenate([audio, numpy.zeros(rate * 8), audio]) gives a phrase, eight seconds of silence and the phrase again, and tts.save writes the result. An audio editor such as Audacity does the same by hand. It all runs offline on an ordinary processor.
In the classroom
Play the track from a classroom speaker or share it with pupils. The ten stock voices of the model, M1 to M5 and F1 to F5, are free and neutral; for a language class, choose a voice recorded in the language you teach, since most catalog voices were recorded in Russian and keep a slight accent elsewhere. Every voice speaks the 31 languages of Supertonic 3; Chinese is not supported yet. A synthetic voice is a helper, not a native speaker, so listen to the track once yourself before class.