Drone Transcripter 4

Drone Transcripter 4

Turns Egyptian Arabic speech into text, tuned on classroom lectures and also covering everyday speech and Arabic-English mixing. For anyone transcribing Egyptian lectures or recordings.

Made by AbdelrhmanX2 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: Drone Transcripter 4 is a speech-to-text model for Egyptian Arabic from Drone Academy. It is a full fine-tune of Whisper large-v3-turbo with 809M parameters. It was trained on about 329.5 hours of reviewed Egyptian audio, with lectures first and then everyday speech.

Who it suits: People who need written text from Egyptian Arabic audio, especially classroom lectures. It also copes with Modern Standard Arabic and speech that mixes Arabic and English.

Title: Who uses it, and how
Industry | Who uses it | How they use it | Example
Education | Teachers and students | Turn recorded lectures into text | A student transcribes a recorded class to review it later
Media | Caption editors | Draft captions for Egyptian Arabic audio | An editor transcribes an interview and then corrects the text
Software | App developers | Add Egyptian Arabic transcription to an app | A developer runs the model from a Python script
Research | Speech researchers | Compare Arabic recognition results across test sets | Check scores on lecture and everyday speech sets

Skip it for noisy phone calls or casual chat, where the card shows weaker results, and for languages other than Arabic.

Where to use it:

• Python with the transformers library, on a GPU
• whisper.cpp, on a CPU or with Metal or CUDA
• Windows, macOS or Linux for downloading the files

Use case | Who | What they get back
Lecture transcripts | Educators | Plain Egyptian Arabic text with no punctuation
Mixed-language speech | Media teams | Arabic script text with English words kept in Latin letters
Laptop transcription | Individuals | A smaller whisper.cpp model that runs without a GPU

How to set it up: You need Python and pip. For the Python route you need the transformers library and a CUDA GPU, and for the whisper.cpp route you need its command-line tool.

• Python and pip
• The transformers library, or whisper.cpp
• An audio clip, such as a WAV file

1. Install the Hugging Face hub library.
2. Download the repository.
3. Run the Python example, or use whisper.cpp with the smaller model file.

pip install -U huggingface_hub
hf download AbdelrhmanX2/drone-transcripter-4

Title: The three model files
File | Use
Transformers checkpoint folder | The Safetensors weights for Python use
f16 whisper.cpp file | Full-size model for whisper.cpp
q5_0 whisper.cpp file | About 574 MB and 2.8 times smaller, the one the card suggests for laptops

How to use it: Run the transformers pipeline on an audio clip and print the text, or call whisper.cpp with the smaller file. Both ask for Arabic.

from transformers import pipeline
asr = pipeline("automatic-speech-recognition", model="AbdelrhmanX2/drone-transcripter-4", device="cuda")
print(asr("clip.wav", generate_kwargs={"language": "ar", "task": "transcribe"})["text"])

whisper-cli -m ggml-drone-transcripter-v4-q5_0.bin -l ar -f clip.wav

Example: You point the script at a recording of an Egyptian classroom lecture, clip.wav. The model prints the spoken words as Arabic text without punctuation, and any English words stay in Latin letters.

For newcomers and experts: If you are new, record a short Egyptian lecture clip, run the Python example, and read the printed text. Expect to add punctuation yourself.

For people in the field, the card reports a 10.2 word error rate on held-out Egyptian lectures against 64.0 for the base model. Everyday speech is weaker, with 40.2 on Casablanca. Training ran two epochs of 4,712 steps on one A100, and songs and music were left out.

Record or pick an audio clip
Load the model in Python or whisper.cpp
Set the language to Arabic
Read the unpunctuated transcript

Good to know:

Title: Word error rate, lower is better
Egyptian lectures, this model: 10.2
Egyptian lectures, base model: 64
FLEURS ar_eg: 10.5
MGB-3 Egyptian: 15.3
Casablanca Egyptian: 40.2
GigaSpeechBench EGY: 48.46

• License: Apache 2.0, matching the base model, and the card asks you to cite Drone Academy
• Cost to run: none named, you use your own computer
• Data it reaches: it downloads files from Hugging Face, then reads the audio files you give it
• Not for: casual conversation or noisy phone audio, where results are weaker
• Output has no punctuation and is in Arabic script

On Hugging Face, 13 November 2026

• Task: automatic speech recognition
• Library: transformers
• Formats: Safetensors
• Downloads in the last 30 days: 0
• Likes: 5
• License: apache-2.0
• Languages: ar, arz
• Based on: openai/whisper-large-v3-turbo
• Tags: transformers, safetensors, automatic-speech-recognition, speech, egyptian-arabic, whisper, whisper.cpp, ggml, model-index

Links

• Model: https://huggingface.co/AbdelrhmanX2/drone-transcripter-4
• Files: https://huggingface.co/AbdelrhmanX2/drone-transcripter-4/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download AbdelrhmanX2/drone-transcripter-4

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