custom-tts-model

Multi-Speaker TTS Model

Trained Text-to-Speech model with the following specs:

Model Details

• Speakers: 1
• Mel Channels: 80
• Sample Rate: 22050
• Vocab Size: 50000

Training Stats

• Final Mel Loss: 5.3586
• Validation Mel Loss: 5.4784
• Training Epochs: 105

Mel Normalization

• Mean: -4.8906
• Std: 4.9261

Usage

import torch
from transformers import AutoTokenizer

# Load model
checkpoint = torch.load('checkpoint_best.pt')
model = MultiSpeakerTTS(checkpoint['config'])
model.load_state_dict(checkpoint['model_state_dict'])
model.eval()

# Generate
tokenizer = AutoTokenizer.from_pretrained("miscovery/tokenizer_v3")
text = "Your text here"
tokens = tokenizer.encode(text)
# ... inference code ...

Speaker Mapping

{ "mahmoud": 0 }