voicetut-egyptian-tashkeel

voicetut-egyptian-tashkeel

Reads Egyptian Arabic text aloud and follows the short vowel marks you write into it. A fine-tune of VoiceTut-TTS that keeps its 19 built-in voices and voice cloning, for app builders, narrators and teachers.

Made by masterofaudio2077 on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: A text to speech model for Egyptian Arabic that learned to follow short vowel marks, called tashkeel. It starts from VoiceTut-TTS, an OmniVoice 0.6B model, and was fine-tuned on 17,546 Egyptian Arabic clips. About half the words in that training text carried vowel marks. You mark the vowels you care about, and the model pronounces them as written.

Who it suits: People who need spoken Egyptian Arabic and want control over how a word is said. That includes app builders, narrators and teachers. The voice setup of the base model is kept, so you can still pick a built-in voice or clone one from a reference clip.

Industry | Who uses it | How they use it | Example
Media | Narrators | Turn scripts into spoken Egyptian Arabic | A short story read in a chosen voice
Education | Arabic teachers | Show how a word sounds with its vowels marked | A lesson that plays the same sentence with and without marks
Software | App developers | Add spoken Egyptian Arabic to an app | A reading feature in a language-learning app
Accessibility | Content teams | Read text aloud with the exact pronunciation they want | Names and unusual words spelled with vowel marks

You can skip it if you need a dialect other than Egyptian Arabic, or if you want a hosted voice service with no files to download.

Where to use it:

• Your own scripts and apps that run the model files on a computer
• Windows, macOS or Linux with Python installed
• Any workflow that already uses the base VoiceTut-TTS model

Use case | Who | What they get back
Narration | Creators | Spoken audio from Egyptian Arabic text
Pronunciation control | Teachers | Words read with the vowels you wrote
Voice choice | App builders | One of 19 built-in voices, or a voice cloned from a reference clip

How to set it up:

• Python, to install the Hugging Face client
• About 2.5 GB of disk for the F32 weights, as Hugging Face lists them
• The model card, for the steps that run it

1. Install the Hugging Face client.
2. Download the model files to your computer.
3. Open the model card and follow its steps to run the model.

pip install -U huggingface_hub
hf download masterofaudio2077/voicetut-egyptian-tashkeel

How to use it: Write your Egyptian Arabic text and add short vowel marks to the words you want pronounced a certain way. Choose one of the 19 built-in voices, or give a reference clip to copy. Text without vowel marks still works as it did before. The visible card text shows no run command, so use the steps in the model card. Clone only voices you have the right to use.

Tool | What it does
Vowel marks | Short vowel signs written into the text so each word is read as marked
Built-in voices | The 19 voices that come with the base model
Reference clip | A recording of a voice that the model can copy
Sample audio | Clips on the card that compare the base model with this one

Example: You write a short Egyptian Arabic sentence with vowel marks on every word and choose the voice Ahmed, which the card uses in its samples. The model speaks the sentence with those vowels. Remove the marks and it reads the plain text as the base model would.

For newcomers and experts: If you are new, try one sentence twice, once with vowel marks and once without, and compare the two recordings. Then try a second voice.

If you know the field: the data holds 58.6 hours from 9 speakers, with vowels on about half the words, so the model sees both marked and plain text. The voice prompt was kept in training, which is why built-in voices and cloning still work.

Egyptian Arabic text with vowel marks
Fine-tuned OmniVoice model
Built-in voice or reference clip
Spoken audio

Good to know: The model changes pronunciation control, not the voices themselves.

• Training data: 17,546 clips, 58.6 hours, 9 speakers
• Language: Egyptian Arabic
• Size: 613M parameters, about 2.5 GB in F32
• Voices: 19 built in, plus cloning from a reference clip
• License: Apache 2.0
• Cost to run: none named
• Data it reaches: model files downloaded to your computer, and any reference clip you provide
• Not for: dialects other than Egyptian Arabic, or cloning a voice you have no right to use

On Hugging Face, 14 November 2026

• Task: text to speech
• Parameters: 613M
• Weights: about 2.5 GB in F32; about 306 MB at 4-bit (estimate)
• Formats: Safetensors
• Downloads in the last 30 days: 150
• Likes: 6
• License: apache-2.0
• Languages: ar
• Based on: mohammedaly22/VoiceTut-TTS
• Tags: safetensors, omnivoice, text-to-speech, arabic, egyptian-arabic, tashkeel, diacritics

Links

• Model: https://huggingface.co/masterofaudio2077/voicetut-egyptian-tashkeel
• Files: https://huggingface.co/masterofaudio2077/voicetut-egyptian-tashkeel/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download masterofaudio2077/voicetut-egyptian-tashkeel

Then follow the usage in its model card. Nothing in this kit runs by itself.