BASE TTS: Lessons from building a billion-parameter Text-to-Speech model on 100K hours of data.
Mateusz LajszczakGuillermo CámbaraYang LiFatih BeyhanArent van KorlaarFan YangArnaud JolyÁlvaro Martín-CortinasAmmar AbbasAdam MichalskiAlexis MoinetSri KarlapatiEwa MuszynskaHaohan GuoBartosz PutryczSoledad López GambinoKayeon YooElena SokolovaThomas DrugmanPublished in: CoRR (2024)
Keyphrases
- text to speech
- input data
- experimental data
- computational model
- data sets
- simulation data
- prior knowledge
- mathematical model
- training data
- statistical methods
- raw data
- database
- linear model
- data points
- empirical data
- data processing
- statistical analysis
- parameter values
- synthetic data
- measured data
- probability distribution
- predictive model
- learning models
- parameter space
- data analysis
- data collection
- end users
- xml documents
- network structure
- prior information
- neural network
- neural network model
- data quality
- parameter estimation
- em algorithm
- autoregressive
- similarity measure
- object oriented
- knowledge discovery
- parameter estimates