ozguradmin / ahenk-turkish-tts

# Ahenk: Turkish text-to-speech
**English** · Türkçe
Open Turkish speech tools: Ahenk (large model, GPU), Ahenk Cep (8.6M-parameter small model, CPU), trnorm (Turkish
text normalisation) and the TR-Zor / TR-Soru test sets.
> **Synthetic voice.** Ahenk and Ahenk Cep speak in one designed voice (f022); it is not a real person's voice or
> recording. Say that it is AI-generated when you share what you make (EU AI Act Art. 50).
Try it in your browser · Ahenk Okur: listen to documents · Ahenk model card ·
Ahenk Cep model card
## Listen
GitHub does not play audio in the page: the links open the WAV files on Hugging Face. The model cards have the same
samples with players. You can try Ahenk Cep in your browser, no install: ozguradmin.github.io/ahenk-turkish-tts.
| What it shows | Text (as written) | Model | Audio |
|---|---|---|---|
| Laughing | Hahaha, bu fıkrayı daha önce hiç duymamıştım! | Ahenk | WAV |
| Cheerful | Harika haber! Sınavı geçtim, akşam hep birlikte kutlayalım! | Ahenk | WAV |
| Calm | Derin bir nefes alın, her şeyi adım adım birlikte halledeceğiz. | Ahenk | WAV |
| Plain reading | Eskişehir'den İzmir'e otobüsle gittik. | Ahenk | WAV |
| Money (trnorm: "doksan dokuz lira doksan dokuz kuruşa") | İndirimle 99,99 TL'ye düştü. | Ahenk | WAV |
| Question pair, the statement first | Cihazın garanti süresi 2 yıl. | Ahenk | WAV |
| Question pair, the same sentence with "mı?" | Cihazın garanti süresi 2 yıl mı? | Ahenk | WAV |
| Long sentence, not in training (12 s) | Karşılaştırma isteyenlere eski ve yeni paketi yan yana anlatırız; farkı müşteri kendisi görür. Akşam olduğunda … | Ahenk | WAV |
| Angry | Yeter artık! Üç gündür aynı sorun için arıyorum ve hâlâ kimse ilgilenmedi! | Ahenk | WAV |
| The same long sentence | Karşılaştırma isteyenlere eski ve yeni paketi yan yana anlatırız; … | Ahenk Cep | WAV |
How they were picked: the plain, money, laughing and emotion clips were chosen by the listener in blind listening.
The question pair and the long sentence were chosen by automatic scores (word-perfect, highest UTMOS). All are
existing evaluation clips, not rendered again.
## What is here
| Component | What it does | Size, hardware | Licence |
|---|---|---|---|
| Ahenk | Turkish LoRA for VoxCPM2: question intonation, 6 emotions and laughter, Turkish or English instructions. Recipe: `tts/voxcpm` | 2B + 36M LoRA, GPU | weights CC BY 4.0 |
| Ahenk Cep | Small student model that learned from Ahenk's voice (EMA Lightning architecture). Code: `tts/small` | 8.6M, 34 MB, RTF ~0.2 on CPU | weights CC BY 4.0 |
| Ahenk Okur | Reads PDF, Word, EPUB, plain text and photos (with OCR) aloud in the browser with Ahenk Cep. Book-length documents are made piece by piece and can be heard while they are being made; the work resumes after the tab is closed; the result downloads as MP3 or WAV. Code: ahenk-cep-pdfreader | browser, on the device | Apache-2.0 |
| `trnorm` | Turns written Turkish into its spoken form: numbers, dates, times, money, percentages, abbreviatio