| import gradio as gr | |
| import nemo.collections.asr as nemo_asr | |
| import numpy as np | |
| # Load the pre-trained Kabyle ASR model | |
| asr_model = nemo_asr.models.EncDecRNNTBPEModel.from_pretrained("nvidia/stt_kab_conformer_transducer_large") | |
| # Function to transcribe the audio input | |
| def transcribe(audio): | |
| # Extract audio data and sample rate | |
| audio_data, sample_rate = audio | |
| # Convert audio data to numpy array if it's not already | |
| if isinstance(audio_data, np.ndarray): | |
| audio_data = np.array(audio_data) | |
| # Transcribe the audio | |
| return asr_model.transcribe([audio_data]) | |
| # Create the Gradio interface with audio input and text output | |
| iface = gr.Interface(fn=transcribe, inputs="audio", outputs="text") | |
| # Launch the Gradio interface | |
| iface.launch() | |