Update app.py

app.py

@@ -1,40 +1,65 @@
+import torch, cv2, os, numpy as np
 import gradio as gr
-from diffusers import StableDiffusionPipeline, DiffusionPipeline
-import torch
-from PIL import Image
+from diffusers import StableDiffusionPipeline
+from modelscope.pipelines import pipeline
+from modelscope.outputs import OutputKeys
 
-# Load Text ➤ Image model
-text2img = StableDiffusionPipeline.from_pretrained(
-    "runwayml/stable-diffusion-v1-5", torch_dtype=torch.float16
-).to("cuda")
+device = "cuda" if torch.cuda.is_available() else "cpu"
 
-# Load Image ➤ Video model
-img2vid = DiffusionPipeline.from_pretrained(
-    "damo-vilab/image-to-video", torch_dtype=torch.float16, variant="fp16"
-).to("cuda")
+# -------- Text -> Image --------
+def text_to_image(prompt, ratio="9:16"):
+    pipe = StableDiffusionPipeline.from_pretrained(
+        "stabilityai/stable-diffusion-2-1", torch_dtype=torch.float16
+    ).to(device)
 
-def generate_image(prompt):
-    image = text2img(prompt).images[0]
-    return image
+    width, height = (576, 1024) if ratio=="9:16" else (1024, 576)
+    image = pipe(prompt, width=width, height=height).images[0]
+    filename = "generated.png"
+    image.save(filename)
+    return filename
 
-def generate_video(image):
-    video_frames = img2vid(image).frames
-    return video_frames  # Gradio will auto-render as video
+# -------- Cinematic Pan Video --------
+def cinematic_video(image, ratio="9:16"):
+    size = (576, 1024) if ratio=="9:16" else (1024, 576)
+    img = cv2.imread(image)
+    img = cv2.resize(img, size)
+    frames = []
+    for i in range(40):
+        dx, dy = int(i*0.5), int(i*0.3)
+        M = np.float32([[1, 0, dx], [0, 1, dy]])
+        frame = cv2.warpAffine(img, M, size)
+        frames.append(frame)
+    out_video = "cinematic_pan.mp4"
+    out = cv2.VideoWriter(out_video, cv2.VideoWriter_fourcc(*'mp4v'), 12, size)
+    for f in frames:
+        out.write(f)
+    out.release()
+    return out_video
 
-with gr.Blocks() as demo:
-    gr.Markdown("## 🎨 Text ➤ Image ➤ Video Generator")
-
-    with gr.Row():
-        prompt = gr.Textbox(label="📝 Enter Prompt")
-        img_output = gr.Image(label="🖼️ Generated Image")
-        btn_img = gr.Button("Generate Image")
+# -------- Realistic AnimateDiff Video --------
+def realistic_video(image, ratio="9:16"):
+    text2video = pipeline('image-to-video', model='damo-vilab/modelscope-damo-text-to-video-synthesis', device=device)
+    result = text2video({'image': image})
+    video_path = result[OutputKeys.OUTPUT_VIDEO]
+    os.rename(video_path, "realistic.mp4")
+    return "realistic.mp4"
 
-    with gr.Row():
-        img_input = gr.Image(label="📥 Drop Image Here")
-        vid_output = gr.Video(label="🎬 Generated Video")
-        btn_vid = gr.Button("Generate Video")
+# -------- Wrapper: Text -> Videos --------
+def text_to_videos(prompt, ratio):
+    img = text_to_image(prompt, ratio)
+    vid1 = cinematic_video(img, ratio)
+    vid2 = realistic_video(img, ratio)
+    return vid1, vid2
 
-    btn_img.click(fn=generate_image, inputs=prompt, outputs=img_output)
-    btn_vid.click(fn=generate_video, inputs=img_input, outputs=vid_output)
+# -------- Gradio UI --------
+with gr.Blocks() as demo:
+    gr.Markdown("## 🎬 Text → Video App (Cinematic + Realistic)")
+    prompt = gr.Textbox(label="Enter your text")
+    ratio = gr.Radio(["9:16", "16:9"], value="16:9", label="Aspect Ratio")
+    btn = gr.Button("Generate Videos")
+    out_vid1 = gr.Video(label="🎥 Cinematic Pan Motion")
+    out_vid2 = gr.Video(label="🎥 Realistic AnimateDiff")
+    
+    btn.click(text_to_videos, [prompt, ratio], [out_vid1, out_vid2])
 
 demo.launch()