README.md

---
configs:
- config_name: default
  data_files:
  - split: train
    path: train/**/*.parquet
  - split: validation
    path: validation/**/*.parquet
  - split: test
    path: test/**/*.parquet
---

# Emilia S2S Mimi Q8 Named Speakers

TinyAya codec-tokenized TTS dataset built from `gs://gp-s2s-data/s2s-data/emilia-s2s/chunk_00036_200k/parquet`.

Rows are Mimi codec-only examples. The training `text` field already includes the
user-facing speaker prefix:

```text
Ira: transcript...
```

Speaker mapping:

- `SP_SP010` -> `Ira:`
- `SP_SP061` -> `Aisha:`
- `SP_SP073` -> `Siya:`
- `SP_SP079` -> `Zoya:`

Reference conditioning is disabled for this artifact. No `ref_*` fields are emitted.