Developing TTS
This pages gives an overview of production of text-to-speech (TTS) programs for minority languages. Programs built so far can be found at the Borealium resource page.
Text processing overview
The development and build status of the text processing (aka normalisation) modules for all TTS languages can be seen on the overview page.
Phases in the production process
- Collecting and preparing text to be used (several months)
- Recording the texts (one week of studio time)
- Text postprocessing (months)
- Splitting the recordings and text transcripts to sentence-long individual files, eventually splitting exceptionally long strings
- Procedures for normalisation of input text
- Building a voice
- Combining parts and distribution