app.py

import gradio as gr
from transformers import BlipProcessor, BlipForConditionalGeneration
from gtts import gTTS
import pygame

model_id = "dblasko/blip-dalle3-img2prompt"
model = BlipForConditionalGeneration.from_pretrained(model_id)
processor = BlipProcessor.from_pretrained(model_id)

def generate_caption(image):
    inputs = processor(images=image, return_tensors="pt")
    pixel_values = inputs.pixel_values

    generated_ids = model.generate(pixel_values=pixel_values, max_length=50)
    generated_caption = processor.batch_decode(generated_ids, skip_special_tokens=True, temperature=0.8, top_k=40, top_p=0.9)[0]

    # Convert text to speech and save as audio file
    tts = gTTS(text=generated_caption, lang='en')
    audio_path = "generated_audio.mp3"
    tts.save(audio_path)

    return generated_caption, audio_path

def play_audio(audio_path):
    pygame.mixer.init()
    pygame.mixer.music.load(audio_path)
    pygame.mixer.music.play()
    pygame.event.wait()

# Create a Gradio interface with an image input, a textbox output, a button, and an audio player
demo = gr.Interface(
    fn=generate_caption, 
    inputs=gr.Image(), 
    outputs=[
        gr.Textbox(label="Generated caption"), 
        gr.Button("Convert to Audio", play_audio),
    ]
)
demo.launch()