Update app.py

app.py

@@ -1,13 +1,101 @@
+import torch
 import gradio as gr
+from diffusers import (
+    StableDiffusionPipeline,
+    StableDiffusionInstructPix2PixPipeline,
+    StableVideoDiffusionPipeline,
+    WanPipeline,
+)
+from diffusers.utils import export_to_video, load_image
 
-def greet(name):
-    return f"Hello, {name}!"
+device = "cuda" if torch.cuda.is_available() else "cpu"
+dtype = torch.float16 if device == "cuda" else torch.float32
 
+# Pipeline factory
+def make_pipe(cls, model_id, **kwargs):
+    pipe = cls.from_pretrained(model_id, torch_dtype=dtype, **kwargs)
+    pipe.enable_model_cpu_offload()
+    return pipe
+
+# Global model caches
+TXT2IMG_PIPE = None
+IMG2IMG_PIPE = None
+TXT2VID_PIPE = None
+IMG2VID_PIPE = None
+
+# Text → Image
+def generate_image_from_text(prompt):
+    global TXT2IMG_PIPE
+    if TXT2IMG_PIPE is None:
+        TXT2IMG_PIPE = make_pipe(
+            StableDiffusionPipeline,
+            "stabilityai/stable-diffusion-2-1-base"
+        ).to(device)
+    return TXT2IMG_PIPE(prompt, num_inference_steps=20).images[0]
+
+# Image → Image
+def generate_image_from_image_and_prompt(image, prompt):
+    global IMG2IMG_PIPE
+    if IMG2IMG_PIPE is None:
+        IMG2IMG_PIPE = make_pipe(
+            StableDiffusionInstructPix2PixPipeline,
+            "timbrooks/instruct-pix2pix"
+        ).to(device)
+    out = IMG2IMG_PIPE(prompt=prompt, image=image, num_inference_steps=8)
+    return out.images[0]
+
+# Text → Video
+def generate_video_from_text(prompt):
+    global TXT2VID_PIPE
+    if TXT2VID_PIPE is None:
+        TXT2VID_PIPE = make_pipe(
+            WanPipeline,
+            "Wan-AI/Wan2.1-T2V-1.3B-Diffusers"
+        ).to(device)
+    frames = TXT2VID_PIPE(prompt=prompt, num_frames=12).frames[0]
+    return export_to_video(frames, "/tmp/wan_video.mp4", fps=8)
+
+# Image → Video
+def generate_video_from_image(image):
+    global IMG2VID_PIPE
+    if IMG2VID_PIPE is None:
+        IMG2VID_PIPE = make_pipe(
+            StableVideoDiffusionPipeline,
+            "stabilityai/stable-video-diffusion-img2vid-xt",
+            variant="fp16" if dtype == torch.float16 else None
+        ).to(device)
+    image = load_image(image).resize((512, 288))
+    frames = IMG2VID_PIPE(image, num_inference_steps=16).frames[0]
+    return export_to_video(frames, "/tmp/svd_video.mp4", fps=8)
+
+# Gradio Interface
 with gr.Blocks() as demo:
-    name_input = gr.Textbox(label="Enter your name")
-    greet_button = gr.Button("Greet")
-    output_text = gr.Textbox(label="Greeting")
+    gr.Markdown("# 🧠 Lightweight Any‑to‑Any AI Playground")
+
+    with gr.Tab("Text → Image"):
+        text_prompt = gr.Textbox(label="Prompt")
+        output_image = gr.Image(label="Generated Image")
+        text2img_button = gr.Button("Generate")
+        text2img_button.click(generate_image_from_text, inputs=text_prompt, outputs=output_image)
+
+    with gr.Tab("Image → Image"):
+        input_image = gr.Image(label="Input Image")
+        edit_prompt = gr.Textbox(label="Edit Prompt")
+        edited_image = gr.Image(label="Edited Image")
+        img2img_button = gr.Button("Generate")
+        img2img_button.click(generate_image_from_image_and_prompt, inputs=[input_image, edit_prompt], outputs=edited_image)
+
+    with gr.Tab("Text → Video"):
+        video_prompt = gr.Textbox(label="Prompt")
+        video_output = gr.Video(label="Generated Video")
+        txt2vid_button = gr.Button("Generate")
+        txt2vid_button.click(generate_video_from_text, inputs=video_prompt, outputs=video_output)
 
-    greet_button.click(fn=greet, inputs=name_input, outputs=output_text)
+    with gr.Tab("Image → Video"):
+        video_input_img = gr.Image(label="Input Image")
+        anim_video_output = gr.Video(label="Animated Video")
+        img2vid_button = gr.Button("Animate")
+        img2vid_button.click(generate_video_from_image, inputs=video_input_img, outputs=anim_video_output)
 
-demo.launch()
+demo.queue()
+demo.launch(show_error=True)