Become a voice

We are always looking for new speakers who want to become a new voice for the Hungarian RHVoice.

If you feel yourself comfortable with reading over 3000 sentences with perfect pronunciation and you have an access to high-quality recording hardware, you can become a new voice for RHVoice.

Until we stabilize and normalize texts which speakers need to read, we cannot publish them here. So, at the time you can contact our team directly using one of the methods described on the contacts page.

Also we will be glad if you share with us some open source non-commercial datasets you found on the internet.

Rules to keep in mind when recording

Recordings should be fully aligned with the text. One line of text in the given txt file should correspond to exactly one wav file;
Recordings should not contain echo or strong room reflections;
Recordings should be clear of background noises;
The speaker should avoid long pauses (longer than 0.7 seconds);
The speaker’s pronunciation should be clear and articulated;
The speaker’s delivery should be as monotonic as possible, without strong exclamation or question intonations;
The speaker should avoid sharp drop in their voice (what is called “vocal fry” in American English);
Even when reading dialogs, the speaker should not try to play with their voice or change it;
Recordings should be clear of any sound effects;
Recordings should not be preprocessed with any software that uses sound spectrum restoration and/or resynthetizing.

How to organize recordings

Each audio file should correspond to exactly one line in text file;
Audio files should be in wav format;
Audio files should be in maximum possible quality;
Minimum quality threshold for audio files is 44kHz sample rate with 16 bits per sample;
Recordings should be named with ascending 4-digit numerals starting from 0000 or 0001;
Recordings should be put in a single folder without any extra folder structure.