app.py

import gradio as gr
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import torch
import librosa
import os

# Model details
models = {
    "m3hrdadfi/wav2vec2-large-xlsr-persian-v3": None,
    "jonatasgrosman/wav2vec2-large-xlsr-53-persian": None,
    "AlirezaSaei/wav2vec2-large-xlsr-persian-fine-tuned": None
}

# Load models and processors
def load_model(model_name):
    model = Wav2Vec2ForCTC.from_pretrained(model_name)
    processor = Wav2Vec2Processor.from_pretrained(model_name)
    return model, processor

def transcribe(audio, model_name):
    if models[model_name] is None:
        models[model_name] = load_model(model_name)
    model, processor = models[model_name]

    audio_data, _ = librosa.load(audio, sr=16000)
    input_values = processor(audio_data, sampling_rate=16000, return_tensors="pt", padding=True).input_values

    with torch.no_grad():
        logits = model(input_values).logits

    predicted_ids = torch.argmax(logits, dim=-1)
    transcription = processor.batch_decode(predicted_ids)[0]
    return transcription

# Read HTML banner
if os.path.exists("banner.html"):
    with open("banner.html", "r", encoding="utf-8") as file:
        banner = file.read()
else:
    banner = "<h1 style='color: red; text-align: center;'>Banner file not found!</h1>"

# Gradio app
with gr.Blocks() as demo:
    gr.HTML(banner)

    gr.Markdown("""
    <h1 style="color: #4CAF50; text-align: center;">Persian Speech-to-Text Models</h1>
    <p style="text-align: center;">Test the best Persian STT models in one place!</p>
    """)

    with gr.Row():
        with gr.Column():
            # Audio input (upload or record)
            audio_input = gr.Audio(
                type="filepath", label="Upload or Record Audio"
            )

            model_dropdown = gr.Dropdown(
                choices=list(models.keys()),
                label="Select Model",
                value="m3hrdadfi/wav2vec2-large-xlsr-persian-v3"
            )

            # Add Test Audio Button
            def use_test_audio():
                return "Test-Audio.ogg"

            test_audio_button = gr.Button("Use Test Audio")

        with gr.Column():
            output_text = gr.Textbox(
                label="Transcription", lines=5, placeholder="The transcription will appear here..."
            )

    transcribe_button = gr.Button("Transcribe")

    test_audio_button.click(
        fn=use_test_audio,
        inputs=[],
        outputs=[audio_input]
    )

    transcribe_button.click(
        fn=transcribe,
        inputs=[audio_input, model_dropdown],
        outputs=output_text
    )

    gr.Markdown("""
    <footer style="text-align: center; margin-top: 20px;">
        <p>Created with ❤️ using Gradio and Hugging Face</p>
    </footer>
    """)

demo.launch()