Emotional-Text-to-Speech/dl-for-emo-tts

Implements Tacotron and DC-TTS architectures fine-tuned on emotional speech datasets (RAVDESS, EMOV-DB) pre-trained on LJ Speech, systematically exploring transfer learning strategies like encoder freezing and learning rate adjustments. Provides reproducible training pipelines and comparative analysis across multiple emotion corpora, with successful configurations documented for single-speaker synthesis with monotonic attention constraints.

458 stars. No commits in the last 6 months.