You can deploy this model and start earning money today!

ghayth123/
tunisian-asr-whisper_medium

Name: tunisian-asr-whisper_medium
Availability: InStock

pytorch

safetensors

whisper

Views

No views yet

Tunisian ASR Whisper Medium

This repository provides a fine-tuned version of OpenAI's Whisper Medium model for Tunisian Arabic Automatic Speech Recognition (ASR).
The model is trained on Tunisian dialect speech data and is ready for testing.

📌 Model Details

Base model: openai/whisper-medium
Language: Tunisian Arabic
Task: Automatic Speech Recognition (ASR)
License: MIT

🚀 Quick Test Example

You can quickly test the model on an audio file using the following Python snippet:

python

import torch
import torchaudio
from transformers import WhisperForConditionalGeneration, WhisperProcessor

# Load the fine-tuned model and processor
model = WhisperForConditionalGeneration.from_pretrained("ghayth123/tunisian-asr-whisper_medium/checkpoint")
processor = WhisperProcessor.from_pretrained("openai/whisper-medium")
model.eval()

# Load audio
audio_path = "example.wav"  # Replace with your audio file
waveform, sample_rate = torchaudio.load(audio_path)

# Resample to 16kHz if necessary
if sample_rate != 16000:
    waveform = torchaudio.transforms.Resample(orig_freq=sample_rate, new_freq=16000)(waveform)

# Convert to numpy and mono if needed
speech = waveform[0].numpy()
if waveform.shape[0] > 1:
    speech = waveform.mean(dim=0).numpy()

# Prepare input and run inference
inputs = processor(speech, sampling_rate=16000, return_tensors="pt")
with torch.no_grad():
    predicted_ids = model.generate(inputs["input_features"])

# Decode and print the transcription
transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]
print("🎧 Transcription:", transcription)