app.py

# app.py
# =============
# This is a complete app.py file for an automatic speech recognition app using the openai/whisper-large-v3-turbo model.
# The app is built using Gradio and Hugging Face Transformers, and it runs on the CPU to avoid video memory usage.

import torch
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
import gradio as gr

# Set device to CPU
device = "cpu"
torch_dtype = torch.float32

# Load the model and processor
model_id = "openai/whisper-large-v3-turbo"

model = AutoModelForSpeechSeq2Seq.from_pretrained(
    model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True
)
model.to(device)

processor = AutoProcessor.from_pretrained(model_id)

# Create the ASR pipeline
pipe = pipeline(
    "automatic-speech-recognition",
    model=model,
    tokenizer=processor.tokenizer,
    feature_extractor=processor.feature_extractor,
    torch_dtype=torch_dtype,
    device=device,
)

def transcribe_audio(audio_file):
    """
    Transcribe the given audio file using the Whisper model.
    
    Parameters:
    audio_file (str): Path to the audio file.
    
    Returns:
    str: Transcribed text.
    """
    result = pipe(audio_file)
    return result["text"]

# Define the Gradio interface
iface = gr.Interface(
    fn=transcribe_audio,
    inputs=gr.inputs.Audio(source="upload", type="filepath"),
    outputs="text",
    title="Whisper ASR Demo",
    description="Upload an audio file and get the transcribed text using the openai/whisper-large-v3-turbo model.",
)

# Launch the Gradio app
if __name__ == "__main__":
    iface.launch()