Whisper Large V3 Narrow Accent

🚀 Whisper-Large v3 for Narrow Accent Classification

This model is designed for narrow accent classification, offering a solution to accurately identify diverse English accents.

🚀 Quick Start

Download repo

git clone git@github.com:tiantiaf0627/vox-profile-release.git

Install the package

conda create -n vox_profile python=3.8
cd vox-profile-release
pip install -e .

Load the model

# Load libraries
import torch
import torch.nn.functional as F
from src.model.accent.whisper_accent import WhisperWrapper

# Find device
device = torch.device("cuda") if torch.cuda.is_available() else "cpu"

# Load model from Huggingface
model = WhisperWrapper.from_pretrained("tiantiaf/whisper-large-v3-narrow-accent").to(device)
model.eval()

Prediction

# Label List
english_accent_list = [
    'East Asia', 'English', 'Germanic', 'Irish', 
    'North America', 'Northern Irish', 'Oceania', 
    'Other', 'Romance', 'Scottish', 'Semitic', 'Slavic', 
    'South African', 'Southeast Asia', 'South Asia', 'Welsh'
]
    
# Load data, here just zeros as the example
# Our training data filters output audio shorter than 3 seconds (unreliable predictions) and longer than 15 seconds (computation limitation)
# So you need to prepare your audio to a maximum of 15 seconds, 16kHz and mono channel
max_audio_length = 15 * 16000
data = torch.zeros([1, 16000]).float().to(device)[:, :max_audio_length]
logits, embeddings = model(data, return_feature=True)
    
# Probability and output
accent_prob = F.softmax(logits, dim=1)
print(english_accent_list[torch.argmax(accent_prob).detach().cpu().item()])

✨ Features

• Accent Classification: This model can classify narrow English accents, including 'East Asia', 'English', 'Germanic', etc.
• Based on Whisper: Built on the openai/whisper-large-v3 base model.
• Benchmark Support: Implements the narrow accent classification described in Vox - Profile: A Speech Foundation Model Benchmark for Characterizing Diverse Speaker and Speech Traits (https://arxiv.org/pdf/2505.14648).

📦 Installation

Download repo

git clone git@github.com:tiantiaf0627/vox-profile-release.git

Install the package

conda create -n vox_profile python=3.8
cd vox-profile-release
pip install -e .

💻 Usage Examples

Basic Usage

# Load libraries
import torch
import torch.nn.functional as F
from src.model.accent.whisper_accent import WhisperWrapper

# Find device
device = torch.device("cuda") if torch.cuda.is_available() else "cpu"

# Load model from Huggingface
model = WhisperWrapper.from_pretrained("tiantiaf/whisper-large-v3-narrow-accent").to(device)
model.eval()

# Label List
english_accent_list = [
    'East Asia', 'English', 'Germanic', 'Irish', 
    'North America', 'Northern Irish', 'Oceania', 
    'Other', 'Romance', 'Scottish', 'Semitic', 'Slavic', 
    'South African', 'Southeast Asia', 'South Asia', 'Welsh'
]
    
# Load data, here just zeros as the example
# Our training data filters output audio shorter than 3 seconds (unreliable predictions) and longer than 15 seconds (computation limitation)
# So you need to prepare your audio to a maximum of 15 seconds, 16kHz and mono channel
max_audio_length = 15 * 16000
data = torch.zeros([1, 16000]).float().to(device)[:, :max_audio_length]
logits, embeddings = model(data, return_feature=True)
    
# Probability and output
accent_prob = F.softmax(logits, dim=1)
print(english_accent_list[torch.argmax(accent_prob).detach().cpu().item()])

📚 Documentation

Model Description

This model includes the implementation of narrow accent classification described in Vox - Profile: A Speech Foundation Model Benchmark for Characterizing Diverse Speaker and Speech Traits (https://arxiv.org/pdf/2505.14648).

The included English accents are:

[
  'East Asia', 'English', 'Germanic', 'Irish', 
  'North America', 'Northern Irish', 'Oceania', 
  'Other', 'Romance', 'Scottish', 'Semitic', 'Slavic', 
  'South African', 'Southeast Asia', 'South Asia', 'Welsh'
]

Some observations we have seen for this model (List to add as we observe more):

1. TTS samples have a higher tendency to be recognized as Germanic.

• Library: https://github.com/tiantiaf0627/vox-profile-release

📄 License

This model is released under the bsd - 3 - clause license.

📚 Citation

@article{feng2025vox,
  title={Vox-Profile: A Speech Foundation Model Benchmark for Characterizing Diverse Speaker and Speech Traits},
  author={Feng, Tiantian and Lee, Jihwan and Xu, Anfeng and Lee, Yoonjeong and Lertpetchpun, Thanathai and Shi, Xuan and Wang, Helin and Thebaud, Thomas and Moro-Velazquez, Laureano and Byrd, Dani and others},
  journal={arXiv preprint arXiv:2505.14648},
  year={2025}
}

📦 Model Information