app.py

import gradio as gr
from transformers import pipeline

captioner = pipeline(model="microsoft/git-base")
tts = pipeline(model="suno/bark-small")

def predict(image):
    text = captioner(image)[0]["generated_text"]
    audio = tts(text)["audio"]
    return text, audio

demo = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="pil"),
    outputs=['text', gr.Audio()]
)

demo.launch()

# gr.Interface.load("models/ronniet/git-base-env").launch()
# gr.Interface.load("models/microsoft/git-base").launch()