---
license: openrail
language:
- ru
pipeline_tag: text-to-speech
library_name: pytorch
tags:
- text-to-speech
- russian
- vits2
- expressive-tts
- emotional-tts
- speech-synthesis
datasets:
- langswap/dialogs-ru-emotional-conversations
---

# Dialogs-RU · Expressive Russian TTS (VITS2)

An expressive, multi-speaker, **emotion-conditioned Russian** text-to-speech model.
It is a **VITS2** model (multi-band iSTFT decoder) trained on the
**[Dialogs](https://huggingface.co/datasets/langswap/dialogs-ru-emotional-conversations)** corpus —
a studio-quality, expressive, *conversational* Russian speech dataset.

This is the exact model described in the paper
*"Dialogs: a studio-quality expressive conversational Russian speech corpus for dialog assistants"*
(Ilya Shigabeev & Ilya Latyshev, Langswap), trained as a proof of concept that the corpus
supports expressive, dialog-style TTS.

👉 **Try it in your browser (free CPU demo):** [frappuccino/dialogs-ru-tts](https://huggingface.co/spaces/frappuccino/dialogs-ru-tts)

## Highlights
- **3 studio voices:** Masha (F), Sveta (F), Dima (M)
- **13 emotional styles:** neutral, happy, surprise, arrogance, yawn, fear, laughing, whisper, disgust, angry, sad, tongue-twister, poem
- **22.05 kHz** output, runs in **real time on CPU** (RTF ≈ 0.2–0.4)
- Plain Russian text in; **stress is optional** (auto-placed with `ruaccent`, or marked manually with `+`)

## Conditioning

| speaker_id | voice | gender |
|---|---|---|
| 0 | Masha / Маша | female |
| 1 | Sveta / Света | female |
| 2 | Dima / Дима | male |

| emotion_id | style | | emotion_id | style |
|---|---|---|---|---|
| 0 | neutral | | 7 | whisper |
| 1 | happy | | 8 | disgust |
| 2 | surprise | | 9 | angry |
| 3 | arrogance | | 10 | sad |
| 4 | yawn | | 11 | tongue-twister |
| 5 | fear | | 12 | poem |
| 6 | laughing | | | |

## Usage

```bash
pip install -r requirements.txt
python inference.py
```

```python
from scipy.io.wavfile import write
from tts import DialogsTTS

tts = DialogsTTS()                      # downloads weights on first run
sr, audio, used = tts.synthesize(
    "В 2024 году цена выросла на 7,5%.",  # numbers/dates/etc. are spelled out
    speaker_id=0,                         # 0=Masha, 1=Sveta, 2=Dima
    emotion_id=1,                         # 1=happy
    normalize=True,                       # rutextnorm: numbers, dates, abbreviations
    auto_stress=True,                     # ruaccent: place '+' stress automatically
)
write("out.wav", sr, audio)
```

### Text front-end (normalization & stress)
`DialogsTTS.synthesize` runs a two-stage front-end before tokenization:

1. **Normalization** (`normalize=True`, default) — [`rutextnorm`](https://github.com/shigabeev/russian_tts_normalization)
   spells out numbers, dates, money, units, fractions and abbreviations
   (`7,5%` → «семь целых и пять десятых процента», `№5` → «номер пять»). It is vendored at
   `text/rutextnorm.py` (single file, MIT, zero dependencies).
2. **Stress** (`auto_stress=True`, default) — [`ruaccent`](https://github.com/Den4ikAI/ruaccent)
   places a `+` before each stressed vowel. The model was trained with these marks; you can also
   write them yourself (e.g. `з+амок` vs `зам+ок`).

The result is lowercased and mapped through the 73-symbol vocabulary (characters outside it are dropped).

## Model details
- Architecture: **VITS2** with a **multi-band iSTFT** generator (4 subbands), mel posterior encoder,
  transformer flows, deterministic duration predictor (no SDP), and a duration discriminator.
- Conditioning: speaker embedding (192-d) concatenated with emotion embedding (64-d) → 256-d global condition.
- Sample rate: 22.05 kHz · vocabulary: 73 symbols (punctuation + Latin + Cyrillic, `+` = stress).
- Training: ~615k steps on a single RTX 4090, batch size 16, following VITS2 hyperparameter defaults.
- Checkpoint: `averaged_G_615000.pth` (weight-averaged generator).
- Evaluation (paper, proof-of-concept): UTMOS ≈ **3.36**; MOS overall ≈ 2.83.

## Limitations
Per-speaker data is modest and unbalanced (Masha 9.9 h, Dima 6.2 h, Sveta 4.4 h), so this single-corpus
model is a proof of concept rather than a production system. The strongest styles are *neutral*, *happy*
and *sad*; rarer styles (*whisper*, *poem*, *tongue-twister*, *yawn*) are subtler. For production use the
authors recommend mixing Dialogs with a larger Russian corpus.

## Links
- 📚 Dataset: [langswap/dialogs-ru-emotional-conversations](https://huggingface.co/datasets/langswap/dialogs-ru-emotional-conversations)
- 🕹️ Demo Space: [frappuccino/dialogs-ru-tts](https://huggingface.co/spaces/frappuccino/dialogs-ru-tts)
- 💻 Training code: [github.com/shigabeev/vits2-emotional](https://github.com/shigabeev/vits2-emotional)
- 🔤 Text normalizer: [github.com/shigabeev/russian_tts_normalization](https://github.com/shigabeev/russian_tts_normalization) (vendored at `text/rutextnorm.py`, MIT)

## Citation
```bibtex
@inproceedings{shigabeev2025dialogs,
  title     = {Dialogs: a studio-quality expressive conversational Russian speech corpus for dialog assistants},
  author    = {Shigabeev, Ilya and Latyshev, Ilya},
  year      = {2025},
  note      = {Langswap}
}
```

## License
**OpenRAIL** — free use including commercial applications, subject to the use-based restrictions of the license.