metadata

library_name: transformers
tags:
  - text-to-speech
license: mit
datasets:
  - sidfeels/transatlantic_accent_tts_dataset
language:
  - en
base_model:
  - parler-tts/parler-tts-mini-v1

Parler TTS Mini - Transatlantic Accent Fine-tune

Model Description

This model is a fine-tuned version of the Parler TTS Mini, specifically trained to produce a transatlantic accent. It was trained on approximately 3 hours of custom data to achieve this unique voice characteristic.

Base Model

Name: Parler TTS Mini
Version: v1
Original Source: parler-tts/parler-tts-mini-v1

Fine-tuned Model Details

Accent: Transatlantic
Training Data: ~3 hours of custom audio
Purpose: To generate text-to-speech audio with a distinct transatlantic accent

Usage

To use this model, you'll need to have the parler_tts library installed. Here's a basic example of how to generate audio with this model:

from parler_tts import ParlerTTSForConditionalGeneration
from transformers import AutoTokenizer
import torch

# Load the model and tokenizer
model = ParlerTTSForConditionalGeneration.from_pretrained("your-model-name")
tokenizer = AutoTokenizer.from_pretrained("your-model-name")

# Prepare input
text = "This is a test of the transatlantic accent."
description = "An older man's voice with a transatlantic accent, speaking calmly and deliberately."

# Generate audio
inputs = tokenizer(description, return_tensors="pt")
prompt = tokenizer(text, return_tensors="pt")

with torch.no_grad():
    output = model.generate(
        input_ids=inputs.input_ids,
        prompt_input_ids=prompt.input_ids,
        do_sample=True
    )

Colab Link for Quick Testing

You can easily experiment with this fine-tuned Parler TTS Mini model using our interactive Google Colab notebook.

To get started:

Click on the link below to open the Colab notebook:
Clone the Notebook. then run each cell.
Experiment with different text inputs and voice descriptions to explore the capabilities of the transatlantic accent model.