Experiments with Synthesis of Swedish Dialects

We describe ongoing work on synthesizing Swedish dialects with an HMM synthesizer. A prototype synthesizer has been trained on a large database for standard Swedish read by a professional male voice talent. We have se-lected a few untrained speakers from each of the following dialectal region: Norrland, Dala, Göta, Gotland and South of Sweden. The plan is to train a multi-dialect average voice, and then use 20-30 minutes of dialectal speech from one speaker to adapt either the standard Swe-dish voice or the average voice to the dialect of that speaker