Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -20,4 +20,88 @@ configs:
|
|
20 |
data_files:
|
21 |
- split: train
|
22 |
path: data/train-*
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
23 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
data_files:
|
21 |
- split: train
|
22 |
path: data/train-*
|
23 |
+
license: cc
|
24 |
+
task_categories:
|
25 |
+
- automatic-speech-recognition
|
26 |
+
- text-to-speech
|
27 |
+
language:
|
28 |
+
- en
|
29 |
+
pretty_name: 'AnimeVox: Character TTS Corpus'
|
30 |
+
size_categories:
|
31 |
+
- 10K<n<100K
|
32 |
---
|
33 |
+
|
34 |
+
# AnimeVox: Character TTS Corpus
|
35 |
+
|
36 |
+
## 🗣️ Dataset Overview
|
37 |
+
|
38 |
+
AnimeVox is an English Text-to-Speech (TTS) dataset featuring 11,020 audio clips from 19 distinct anime characters across popular series. Each clip includes a high-quality transcription, character name, and anime title, making it ideal for voice cloning, custom TTS model fine-tuning, and character voice synthesis research.
|
39 |
+
|
40 |
+
The dataset was created and processed using **[TTSizer](https://github.com/taresh18/TTSizer)**, an open-source tool that automates creating high-quality TTS datasets from raw media.
|
41 |
+
|
42 |
+
**Watch the Demo Video:**
|
43 |
+
[](https://youtu.be/POwMVTwsZDQ?si=rxNy7grLyROhdIEd)
|
44 |
+
|
45 |
+
## 📊 Dataset Statistics
|
46 |
+
|
47 |
+
- **Total samples:** 11,020
|
48 |
+
- **Characters:** 19
|
49 |
+
- **Anime series:** 15
|
50 |
+
- **Audio format:** 44.1kHz mono WAV
|
51 |
+
- **Storage size:** ~3.5GB
|
52 |
+
|
53 |
+
## 🎧 Dataset Structure
|
54 |
+
|
55 |
+
* **Instances:** Each sample is a dictionary with the following structure:
|
56 |
+
```python
|
57 |
+
{
|
58 |
+
"audio": {"path": "...", "array": ..., "sampling_rate": 44100},
|
59 |
+
"transcription": "English text spoken by the character.",
|
60 |
+
"character_name": "Character Name",
|
61 |
+
"anime": "Anime Series Title"
|
62 |
+
}
|
63 |
+
* **Fields:**
|
64 |
+
* `audio`: Audio object (44.1kHz).
|
65 |
+
* `transcription`: (str) English transcription.
|
66 |
+
* `character_name`: (str) Name of the speaking character.
|
67 |
+
* `anime`: (str) Anime series title.
|
68 |
+
* **Splits:** A single train split with all 11,020 samples from 19 characters.
|
69 |
+
|
70 |
+
## 🛠️ Dataset Creation
|
71 |
+
|
72 |
+
### Source
|
73 |
+
Audio clips were sourced from official English-dubbed versions of popular anime series. The clips were selected to capture diverse emotional tones and vocal characteristics unique to each character.
|
74 |
+
|
75 |
+
### Processing with TTSizer
|
76 |
+
This dataset was generated using **[TTSizer](https://github.com/taresh18/TTSizer)**, which offers an end-to-end automated pipeline for creating TTS-ready datasets. Key features utilized include:
|
77 |
+
|
78 |
+
* **Advanced Multi-Speaker Diarization:** To accurately identify and segment speech for each of the characters, even in complex audio environments.
|
79 |
+
* **State-of-the-Art Model Integration:** Leveraging models such as MelBandRoformer (for vocal separation), Gemini (for diarization), CTC-Aligner (for precise audio-text alignment), and WeSpeaker (for speaker embedding/verification).
|
80 |
+
* **Quality Control:** Implementing automatic outlier detection to flag and help refine potentially problematic audio-text pairs, ensuring higher dataset quality.
|
81 |
+
|
82 |
+
The tool's configurable nature allowed for fine-tuning the entire process to suit the specific needs of this anime voice dataset.
|
83 |
+
|
84 |
+
## 📜 Licensing & Usage
|
85 |
+
|
86 |
+
* **License:** Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0).
|
87 |
+
|
88 |
+
## 🚀 How to Use
|
89 |
+
|
90 |
+
```python
|
91 |
+
from datasets import load_dataset
|
92 |
+
|
93 |
+
# Load the dataset
|
94 |
+
dataset = load_dataset("taresh18/AnimeVox")
|
95 |
+
|
96 |
+
# Access the training split
|
97 |
+
train_data = dataset["train"]
|
98 |
+
|
99 |
+
# Print dataset information
|
100 |
+
print(f"Dataset contains {len(train_data)} samples")
|
101 |
+
|
102 |
+
# Access a specific sample
|
103 |
+
sample = train_data[0]
|
104 |
+
print(f"Character: {sample['character_name']}")
|
105 |
+
print(f"From anime: {sample['anime']}")
|
106 |
+
print(f"Transcription: {sample['transcription']}")
|
107 |
+
```
|