Spaces:
Runtime error
Runtime error
Upload 2 files
Browse files- 10_fairy_tale_book.py +57 -0
- Requirements.txt +4 -0
10_fairy_tale_book.py
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
from transformers import Blip2Processor, Blip2ForConditionalGeneration
|
| 3 |
+
from PIL import Image
|
| 4 |
+
import torch
|
| 5 |
+
|
| 6 |
+
# ✅ 모델 로딩
|
| 7 |
+
processor = Blip2Processor.from_pretrained("Salesforce/blip2-flan-t5-xl")
|
| 8 |
+
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xl").to("cuda" if torch.cuda.is_available() else "cpu")
|
| 9 |
+
|
| 10 |
+
# 📌 Step 1: 이미지 설명 추출 (이미지 → 캡션)
|
| 11 |
+
def extract_caption(image):
|
| 12 |
+
inputs = processor(images=image, return_tensors="pt").to(model.device)
|
| 13 |
+
outputs = model.generate(**inputs, max_new_tokens=50)
|
| 14 |
+
caption = processor.tokenizer.decode(outputs[0], skip_special_tokens=True)
|
| 15 |
+
return caption
|
| 16 |
+
|
| 17 |
+
# 📌 Step 2: 캡션 기반 동화 프롬프트 생성
|
| 18 |
+
def build_prompt_from_caption(caption):
|
| 19 |
+
return (
|
| 20 |
+
f"Write a magical and fun children's fairytale based on this description: \"{caption}\". "
|
| 21 |
+
"Start with 'Once upon a time' and continue for at least 7 sentences. "
|
| 22 |
+
"Include characters, emotions, the setting, and a twist. Make it feel like a real story."
|
| 23 |
+
)
|
| 24 |
+
|
| 25 |
+
# 📌 Step 3: 동화 생성 (캡션 + 프롬프트 → 텍스트)
|
| 26 |
+
def generate_fairytale(image):
|
| 27 |
+
caption = extract_caption(image)
|
| 28 |
+
prompt = build_prompt_from_caption(caption)
|
| 29 |
+
inputs = processor(images=image, text=prompt, return_tensors="pt").to(model.device)
|
| 30 |
+
outputs = model.generate(
|
| 31 |
+
**inputs,
|
| 32 |
+
max_new_tokens=400,
|
| 33 |
+
do_sample=True,
|
| 34 |
+
temperature=0.95,
|
| 35 |
+
top_p=0.9,
|
| 36 |
+
)
|
| 37 |
+
story = processor.tokenizer.decode(outputs[0], skip_special_tokens=True)
|
| 38 |
+
return story
|
| 39 |
+
|
| 40 |
+
# 🌐 Gradio UI
|
| 41 |
+
with gr.Blocks() as demo:
|
| 42 |
+
gr.Markdown("## 🧚♂️ 이미지 기반 AI 동화 생성기\n사진을 업로드하면 동화로 바꿔드립니다!")
|
| 43 |
+
|
| 44 |
+
with gr.Row():
|
| 45 |
+
image_input = gr.Image(type="pil", label="📸 이미지 업로드")
|
| 46 |
+
|
| 47 |
+
with gr.Row():
|
| 48 |
+
generate_button = gr.Button("✨ 동화 만들기")
|
| 49 |
+
|
| 50 |
+
with gr.Row():
|
| 51 |
+
output_text = gr.Textbox(label="📖 생성된 동화", lines=10)
|
| 52 |
+
|
| 53 |
+
generate_button.click(fn=generate_fairytale, inputs=[image_input], outputs=[output_text])
|
| 54 |
+
|
| 55 |
+
# 실행
|
| 56 |
+
if __name__ == "__main__":
|
| 57 |
+
demo.launch(share=True)
|
Requirements.txt
ADDED
|
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
gradio
|
| 2 |
+
transformers
|
| 3 |
+
torch
|
| 4 |
+
pillow
|