# Speech synthesis

Dia-1.6B-3bit is a 3-bit quantized model converted from mlx-community/Dia-1.6B, primarily used for text-to-speech tasks.

Dia-1.6B-4bit is a 4-bit quantized text-to-speech model based on the MLX format, converted from nari-labs/Dia-1.6B.

Urdu TTS model fine-tuned from microsoft/speecht5_tts with small training dataset (4,200 sentences). Commercial use requires retraining.

A text-to-speech model fine-tuned on the VoxPopuli dataset based on microsoft/speecht5_tts

A Swedish text-to-speech model fine-tuned based on Microsoft's SpeechT5 architecture, trained using the Common Voice dataset

A Transformer-based text-to-speech model built on fairseq S^2, supporting Simplified Chinese with a single female voice, trained on Common Voice v7 and CSS10 datasets.