Search results

• LVBERT - Latvian BERT

  4 resources

  LVBERT is the first publicly available monolingual BERT language model pre-trained for Latvian. For training we used the original implementation of BERT on TensorFlow with the whole-word masking and the next sentence prediction objectives. We used BERT-BASE configuration with 12 layers, 768 hidden units, 12 heads, 128 sequence length, 128 mini-batch size and 32,000 token vocabulary.

  Use "LVBERT - Latvian BERT"

• Ilvars - Latvian Male VITS Text-to-Speech Model (vers. 2022)

  1 resources

  A neural model for text-to-speech synthesis in Latvian. Trained using VITS on a 20-hour speech corpus of audiobooks read in a male voice. Currently released for research purposes only.

  Use "Ilvars - Latvian Male VITS Text-to-Speech Model (vers. 2022)"

• Ilvars - Latvian Male VITS Text-to-Speech Model (vers. 2023)

  1 resources

  A neural model for text-to-speech (TTS) synthesis in Latvian. Trained using VITS on a 25-hour speech corpus of audiobooks read in a male voice. Available for academic and non-commercial purposes via an API. To get access to the API, please, send a request to info@ailab.lv.

  Use "Ilvars - Latvian Male VITS Text-to-Speech Model (vers. 2023)"