metadata

language:
  - ar
base_model:
  - coqui/XTTS-v2
pipeline_tag: text-to-speech

Tunisian_TTS_System

Finetuning XTTS V2 on Tunisian custom dataset

Model Details

Model Description

This model is a fine-tuned version of coqui ai/xtts v2 on the tunisian_custom dataset (2h and 30mn). It achieves the following results of loss on the evaluation sets: First evaluation set: 0.0274 Second evaluation set: 0.0946

Training Hyperparameters

batch_size=2 num_of_epochs=6 grad_accumulation_steps=1 Max permitted audio size in seconds=11