Upload 2 files

app.py

@@ -0,0 +1,84 @@
+
+import gradio as gr
+import torch
+from diffusers import AutoPipelineForText2Image, DDIMScheduler
+from transformers import CLIPVisionModelWithProjection
+from diffusers.utils import load_image
+import os
+from PIL import Image
+
+STYLE_MAP = {
+    "pixar": [
+        "https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/style_ziggy/img0.png",
+        "https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/style_ziggy/img1.png",
+        "https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/style_ziggy/img2.png",
+        "https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/style_ziggy/img3.png",
+        "https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/style_ziggy/img4.png"
+    ]
+}
+
+torch_dtype = torch.float16
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+image_encoder = CLIPVisionModelWithProjection.from_pretrained(
+    "h94/IP-Adapter",
+    subfolder="models/image_encoder",
+    torch_dtype=torch_dtype,
+)
+
+pipeline = AutoPipelineForText2Image.from_pretrained(
+    "stabilityai/stable-diffusion-xl-base-1.0",
+    torch_dtype=torch_dtype,
+    image_encoder=image_encoder,
+    variant="fp16"
+).to(device)
+
+pipeline.scheduler = DDIMScheduler.from_config(pipeline.scheduler.config)
+pipeline.load_ip_adapter(
+    "h94/IP-Adapter",
+    subfolder="sdxl_models",
+    weight_name=[
+        "ip-adapter-plus_sdxl_vit-h.safetensors",
+        "ip-adapter-plus-face_sdxl_vit-h.safetensors"
+    ]
+)
+pipeline.set_ip_adapter_scale([0.7, 0.3])
+pipeline.enable_model_cpu_offload()
+
+os.makedirs("outputs", exist_ok=True)
+
+def generate_storybook(data):
+    character_image_url = data["character_image_url"]
+    style = data["style"]
+    scenes = data["scenes"]
+
+    face_image = load_image(character_image_url)
+    style_images = [load_image(url) for url in STYLE_MAP[style]]
+
+    result_paths = []
+    for i, prompt in enumerate(scenes):
+        image = pipeline(
+            prompt=prompt,
+            ip_adapter_image=[style_images, face_image],
+            negative_prompt="blurry, bad anatomy",
+            width=768,
+            height=1024,
+            guidance_scale=7.5,
+            num_inference_steps=30,
+            generator=torch.Generator(device).manual_seed(i + 42)
+        ).images[0]
+
+        path = f"outputs/scene_{i+1}.png"
+        image.save(path)
+        result_paths.append(path)
+
+    return result_paths
+
+iface = gr.Interface(
+    fn=generate_storybook,
+    inputs=gr.JSON(),
+    outputs=gr.JSON(),
+    title="AI Storybook Generator"
+)
+
+iface.launch()

requirements.txt

@@ -0,0 +1,7 @@
+torch
+diffusers
+transformers
+accelerate
+safetensors
+opencv-python
+gradio