app.py

import subprocess
subprocess.run(["pip", "install", "gradio", "--upgrade"])
subprocess.run(["pip", "install", "transformers"])
subprocess.run(["pip", "install", "torchaudio", "--upgrade"])

import numpy as np
import gradio as gr
from transformers import WhisperProcessor, WhisperForConditionalGeneration

# Load Whisper ASR model and processor
model_name = "openai/whisper-small"
processor = WhisperProcessor.from_pretrained(model_name, sampling_rate=44100)
model = WhisperForConditionalGeneration.from_pretrained(model_name)
forced_decoder_ids = processor.get_decoder_prompt_ids(language="italian", task="transcribe")



def transcribe_audio(input_audio):
    input_audio_np = np.array(input_audio[0].data)
    input_features = processor(input_audio_np, return_tensors="pt").input_features

    # Generate token ids
    predicted_ids = model.generate(input_features, forced_decoder_ids=forced_decoder_ids)

    # Decode token ids to text
    transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)
    
    return transcription[0]

iface = gr.Interface(
    fn=transcribe_audio,
    inputs=gr.Audio(sources=["microphone"], label="Speak"),
    outputs="text",
    live=True
)

iface.launch()