Page MenuHomePhabricator

Recording short texts for modelling
Closed, InvalidPublic

Description

An alternative to training your own voice from scratch is to use a shorter text and adapt an existing speech synthesis voice based on this. The interface works as above and will rely on the same code but also provides the possibility to listen to the adapted voice.
Existing: No
To do: This is not part of MVP and any development happens outside of the scope of this project. Develop the web interface for recording using HTML5 audio and a backend with initial analysis and storage of audio files.

Identified as a component during the pilot study.

Event Timeline

Sebastian_Berlin-WMSE renamed this task from [Task] Recording short texts for modelling (Wikispeech) to Recording short texts for modelling.Oct 28 2019, 1:59 PM

This task was initially created for Wikispeech-Text-to-Speech, but should be reusable for Wikispeech-Speech-Data-Collector. It might need some rewording.