AdoCleanCode/stage2_mixed_curriculum_v1

---
dataset_info:
features:
- name: text
dtype: string
- name: phonemes_ipa
dtype: string
- name: phoneme_tokens
dtype: string
- name: mixed_text
dtype: string
- name: xcodec2_tokens
dtype: string
- name: to_convert
dtype: string
- name: sequence
dtype: string
splits:
- name: train
num_bytes: unknown
num_examples: 1421568
download_size: unknown
dataset_size: unknown
---

# Stage 2 Mixed Text/Phoneme TTS Dataset

This dataset contains mixed text/phoneme sequences for TTS training with curriculum learning.

## Curriculum Learning

The probability of converting words to phonemes increases over the dataset:
- **Start**: p = 0.3 (more text, less phonemes)
- **End**: p = 1.0 (all phonemes)
- **Transition**: Linear over 500,000 rows

Each row uses p(i) for ALL its words/spaces, then i increments for the next row.

## Features

| Column | Description |
|--------|-------------|
| `text` | Plain text (all words) |
| `phonemes_ipa` | IPA notation of all phonemes |
| `phoneme_tokens` | All phonemes as `<\|ph_xxxx\|>` hex tokens |
| `mixed_text` | Mixed text and phoneme tokens |
| `xcodec2_tokens` | Full audio as XCodec2 tokens |
| `to_convert` | "text and phonemes" / "phonemes" / "text" |
| `sequence` | Full training sequence with markers |

## Training Format

```
Convert the {to_convert} to speech: <|start_phon_gen|>{mixed_text}<|end_phon_gen|><|SPEECH_GENERATION_START|>{xcodec2_tokens}<|SPEECH_GENERATION_END|>
```

## Processing Info

- Source: AAdonis/multilingual_audio_alignments (english)
- Index range: 0 to 499541
- Samples per audio: 3
- Total samples: 1421568
- Final global row counter: 1421568
- Final probability: 1.000