Style-Controlled VALL-E for Few-Shot Emotional German TTS
Abstract
We present an expressive German Text-to-Speech (TTS) system built on a modified VALL-E neural codec language model. Our approach introduces a style-conditioning mechanism to enable emotional and prosodic control during speech synthesis. Using the Thorsten German Emotional TTS dataset, we preprocess and augment 2,400 utterances across 8 emotions. Our model integrates emotion tokens and style embeddings to guide expressive generation without explicit supervision. Preliminary results suggest that this method can produce prosodically varied and natural-sounding German speech, demonstrating its potential in low-resource emotional TTS settings.
Full text
Style-Controlled VALL-E for Few-Shot Emotional German TTS 1st Rami Kammoun Department of Telecommunications and Artificial Intelligence Budapest University of Technology and Economics Budapest, Hungary [email protected] 2nd Mohammed Salah Al-Radhi Department of Telecommunications and Artificial Intelligence Budapest University of Technology and Economics Budapest, Hungary [email protected] Abstract—We present an expressive German Text-to-Speech (TTS) system built on a modified VALL-E neural codec language model. Our approach introduces a style-conditioning mechanism to enable emotional and prosodic control during speech synthesis. Using the Thorsten German Emotional TTS dataset, we preprocess and augment 2,400 utterances across 8 emotions. Our model integrates emotion tokens and style embeddings to guide expressive generation without explicit supervision. Preliminary results suggest that this method can produce prosodically varied and natural-sounding German speech, demonstrating its potential in low-resource emotional TTS settings. Index Terms—Speech Synthesis, Emotional TTS, VALL-E, Neural Codec, Low-Resource Languages, German TTS. I. INTRODUCTION