You can deploy this model and start earning money today!

yusasif/
nigerian-accent-english

Name: nigerian-accent-english
Availability: InStock

Views

No views yet

Nigerian Accent English TTS — XTTS v2

A fine-tuned XTTS v2 model specialized for Nigerian-accented English text-to-speech synthesis with zero-shot voice cloning.

Given a short reference audio clip of any English speaker with a Nigerian accent, the model synthesizes any English text in that voice — preserving the natural Nigerian accent, rhythm, and prosody.

Model Details


Base model	XTTS v2 (coqui/XTTS-v2)
Language	English (Nigerian accent)
Architecture	GPT + VITS decoder
Task	Text-to-speech, zero-shot voice cloning
Sample rate	24 kHz
License	Coqui Public Model License (CPML)

Capabilities

Synthesize natural English speech with an authentic Nigerian accent
Clone any Nigerian-accented English voice from a short reference clip (~6 seconds)
Preserve the speaker's unique tone, rhythm, and accent characteristics
Suitable for Nigerian media, education, accessibility, and voice applications

Installation

pip install TTS

Usage

Simple voice cloning

python

from TTS.api import TTS

tts = TTS("yusasif/nigerian-accent-english", gpu=True)

tts.tts_to_file(
    text="Welcome to Nigeria. How can I assist you today?",
    file_path="output.wav",
    speaker_wav="reference_speaker.wav",  # ~6 seconds of clean Nigerian-accented English
    language="en",
)

Low-level API for more control

python

from TTS.tts.configs.xtts_config import XttsConfig
from TTS.tts.models.xtts import Xtts
import soundfile as sf

config = XttsConfig()
config.load_json("/path/to/model/config.json")
model = Xtts.init_from_config(config)
model.load_checkpoint(config, checkpoint_dir="/path/to/model/", eval=True)
model.cuda()

# Extract voice characteristics from reference audio
gpt_cond_latent, speaker_embedding = model.get_conditioning_latents(
    audio_path="reference_speaker.wav"
)

# Synthesize
outputs = model.inference(
    text="Welcome to Nigeria. How can I assist you today?",
    language="en",
    gpt_cond_latent=gpt_cond_latent,
    speaker_embedding=speaker_embedding,
    temperature=0.7,
    speed=1.0,
)

sf.write("output.wav", outputs["wav"], 24000)

Tips for Best Results

Use 6–12 seconds of clean, noise-free reference audio for best voice cloning
The reference audio should be recorded in a quiet environment
The model captures Nigerian accent most accurately when the reference speaker also has a Nigerian accent
Avoid reference clips with background music or heavy noise

Reference Voice Guidelines

Condition	Result
Clean studio recording	Best quality
Phone recording, quiet room	Good quality
Noisy background	Reduced quality
Less than 3 seconds	May not clone correctly
6–12 seconds	Optimal