Kızagan TTS v1.0

Kızagan TTS v1.0

Reads Turkish text aloud with a bundled reference voice, or clones a voice from a clip you upload or record. It comes with a form, a Gradio API and an MCP server.

Made by AlicanKiraz0 on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: Kızagan TTS v1.0 is a live demo that reads Turkish text aloud and can clone a voice from a short recording. It runs a Turkish fine-tune of VoxCPM2 and produces 48 kHz mono audio. Before you generate, it shows how numbers, dates, prices and abbreviations will be spoken.

Who it suits: Turkish-speaking creators, developers and teachers who want to hear written Turkish read well will find it useful. The README names no customers, so the rows below are judgment from what the demo does.

Industry | Who uses it | How they use it | Example
Media | Video and podcast makers | Turn a Turkish script into narration in the bundled voice | A voice-over made from a short written paragraph
Software | App developers | Call the Space through its Gradio API or MCP server | Sending text to the speak endpoint from Python
Education | Teachers of Turkish | Hear dates, prices and abbreviations read the way people say them | Checking how a price such as 875,40 TL is read aloud
Accessibility | Readers who prefer audio | Listen to written Turkish instead of reading it | Pasting a short passage and pressing play

You can skip it if you need speech in other languages, because the demo is built around Turkish text and a 2000 character limit.

Where to use it: Use it in a browser tab, from Python code or from an MCP client, and run a copy on Windows, macOS or Linux.

• A web browser: open the Space page on Hugging Face
• Python: call the Gradio API with gradio_client
• An MCP client: the Space also serves an MCP endpoint
• Windows, macOS or Linux: run a copy with Python 3.12

Use case | Who | What they get back
Read text aloud | Creators and editors | A streamed audio clip, then a full WAV file to download
Clone a voice | Anyone with a recording they may use | Speech in the voice of the clip they upload or record
Check Turkish normalization | Editors and developers | The text as it will be spoken, shown sentence by sentence

How to set it up: Using the Space online needs nothing. A local copy needs Python 3.12 and the listed packages.

• Python 3.12
• Gradio 6.30.0
• The packages in requirements.txt

1. Install the Hugging Face hub package.
2. Download the Space's files into a folder.
3. Move into that folder.
4. Install Gradio 6.30.0, then the listed packages.
5. Start the app with Python.

pip install -U huggingface_hub
hf download AlicanKiraz0/kizagan-tts-v1 --repo-type space --local-dir kizagan-tts-v1
cd kizagan-tts-v1
pip install gradio==6.30.0
pip install -r requirements.txt
python app.py

A normal run downloads the model. For a quick look at the interface without it, the README gives a mock mode that plays a synthetic tone instead of speech.

KZ_MOCK=1 python app.py

How to use it: Open the app and fill in one form with two voice choices. Kızagan's voice reads your text with the bundled recording, so you upload nothing. Your own voice reads it with a clip you upload or record, and typing the clip's words adds a continuation mode that skips the style instruction.

Tool | What it does
Kızagan's voice | Reads the text with the bundled reference recording
Your own voice | Reads the text with a clip you upload or record
Text preview | Shows the normalized text sentence by sentence, with expanded numbers marked
Live audio | Starts playing while it is made, then offers a full WAV file
Style | Adds a Turkish instruction in brackets before each sentence, still experimental
Advanced settings | CFG, diffusion steps, seed, normalization, sentence splitting and the pause between sentences

From code, the Space answers through the Gradio API. The README shows these calls, and the second one returns the streamed audio, the WAV path, the spoken text and the seed.

from gradio_client import Client, handle_file

client = Client("AlicanKiraz0/kizagan-tts-v1")

# Kızagan sesiyle
result = client.predict(
    text="Merhaba, bugün sana nasıl yardımcı olabilirim?",
    style="Doğal (yönerge yok)",
    api_name="/speak",
)

# Kendi sesinle
result = client.predict(
    text="Bu sesle okunacak yeni metin.",
    reference_audio=handle_file("speaker.wav"),
    transcript="",              # ya da kayıtta söylenen sözler
    api_name="/clone",
)

Example: Choose Kızagan's voice and ask: "Read this aloud: Toplam tutar 875,40 TL." The text preview spells the amount out in Turkish words. The audio starts playing while it is made, and a full WAV file is ready when it ends.

For newcomers and experts: Open the Space, keep Kızagan's voice selected, type one short Turkish sentence and generate it. Listen first, then try your own recording once the result sounds right. Use only voices you have the right to use.

The model is a Turkish fine-tune of VoxCPM2 with about 2.29B parameters in BF16. Long text is split into sentences, and each one is generated fresh from the same original reference, with a short silence between them. The upstream normalizer is switched off because it is unreliable for Turkish, so this Space expands numbers itself in kizagan_text.py. The default recipe comes from the model card, with CFG 2.0, 16 steps and seed 42. The author has not measured how well the model follows style instructions.

Text is normalized for Turkish
It is split into sentences
Each sentence uses the same reference
Audio streams to the player
A full WAV file is ready to download

Good to know:

• License: Apache-2.0, which the README also gives for the base model, the bundled clip and the inference code
• Cost to run: no API key or paid service named, and the Space runs on zero-a10g hardware at Hugging Face
• Data it reaches: the text you type and any audio you upload or record, plus a model download on a normal local run
• Not for: text beyond 2000 characters, style instructions in continuation mode, or voices you have no right to use

On Hugging Face, 13 November 2026

• Likes: 16
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/AlicanKiraz0/kizagan-tts-v1
• Files: https://huggingface.co/spaces/AlicanKiraz0/kizagan-tts-v1/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.