| --- |
| license: apache-2.0 |
| language: |
| - yo |
| - ha |
| - ig |
| - pcm |
| - en |
| tags: |
| - tts |
| - text-to-speech |
| - nigerian |
| - yoruba |
| - hausa |
| - igbo |
| - pidgin |
| - fish-speech |
| base_model: fishaudio/fish-speech-1.5 |
| datasets: |
| - UbuntuFarms/nigerian-tts-dataset |
| --- |
| |
| # Nigerian Fish Speech TTS |
|
|
| Voice cloning configuration for Nigerian languages using Fish Speech. |
|
|
| ## Speakers |
|
|
| This configuration includes 145 speaker references for: |
| - Nigerian English |
| - Nigerian Pidgin |
| - (Expandable to Yoruba, Hausa, Igbo) |
|
|
| ## Dataset |
|
|
| - **Source**: [UbuntuFarms/nigerian-tts-dataset](https://huggingface.co/datasets/UbuntuFarms/nigerian-tts-dataset) |
| - **Total Samples**: 145 |
| - **Speakers**: 145 |
|
|
| ## Usage with Fish Speech |
|
|
| ```python |
| from fish_speech import FishSpeech |
| |
| # Load model |
| tts = FishSpeech(model_path="fishaudio/fish-speech-1.5") |
| |
| # Use Nigerian speaker reference |
| audio = tts.synthesize( |
| text="Hello, how are you today?", |
| reference_audio="path/to/nigerian_speaker.wav", |
| reference_text="Reference transcription" |
| ) |
| ``` |
|
|
| ## Speaker References |
|
|
| The `speakers.json` file contains reference audio paths for each speaker. |
|
|
| ## Base Model |
|
|
| - [fishaudio/fish-speech-1.5](https://huggingface.co/fishaudio/fish-speech-1.5) |
|
|
| ## Training |
|
|
| This repository contains speaker reference configurations for zero-shot voice cloning. |
| Fish Speech uses these references to clone voices without fine-tuning. |
|
|
