from speechbrain.inference.ASR import EncoderASR
import gradio as gr

model = EncoderASR.from_hparams("speechbrain/asr-wav2vec2-dvoice-wolof")

# def transcribe(audio):
#   return model.transcribe_file(audio.name)

def transcribe(audio):
    if isinstance(audio, str):  # If input is a file path
        return model.transcribe_file(audio.name)
    else:  # If input is audio data from microphone
        return model.transcribe_array(audio)


demo = gr.Interface(fn=transcribe, inputs=["file", "microphone"], outputs="text", 
                   title="Transcription audio en wolof latin by Papa Séga", 
             description="Ce modèle transcrit un fichier audio en wolof en texte en utilisant l'alphabet latin.",              
             input_label="Audio en wolof", 
             output_label="Transcription alphabet latin"
                   )

demo.launch()