| --- |
| annotations_creators: [crowdsourced] |
| language: [fr] |
| license: cc-by-4.0 |
| multilinguality: [monolingual] |
| pretty_name: French TTS Dataset |
| task_categories: [text-to-speech] |
| tags: [french, tts, audio, speech, voice] |
| --- |
| # 🇫🇷 French TTS Dataset |
|
|
| This dataset contains French speech audio paired with clean transcriptions, intended for training text-to-speech models such as [Spark-TTS](https://github.com/unslothai/spark-tts) or [Coqui TTS](https://github.com/coqui-ai/TTS). |
|
|
| ## 📁 Contents |
|
|
| - `dataset.parquet` — metadata file with audio paths, transcriptions, speaker info |
| - `Audio/` — directory of all `.wav` files used for training |
|
|
| ## 📊 Dataset Structure |
|
|
| The `dataset.parquet` file includes the following columns: |
|
|
| | Column | Description | |
| |---------------------|------------------------------------| |
| | `audio` | Path to `.wav` file | |
| | `text` | French transcription | |
| | `original_filename` | Original `.wav` file name | |
| | `clip_creation_time`| ISO-formatted timestamp (optional) | |
|
|
| ## 💡 Example |
|
|
| ```python |
| from datasets import load_dataset, Audio |
| |
| dataset = load_dataset("davidmwila/tts-french-dataset") |
| dataset = dataset.cast_column("audio", Audio()) |
| |
| print(dataset[0]) |
| ``` |
|
|
| ## ✅ License |
|
|
| CC-BY 4.0 |
|
|
| ## 🙌 Credits |
|
|
| Created by [David Mwila](https://huggingface.co/davidmwila) |
|
|
