Update app.py

app.py

@@ -9,55 +9,64 @@ pipe = CogVideoXPipeline.from_pretrained(
     "THUDM/CogVideoX1.5-5B",
     torch_dtype=torch.bfloat16
 )
-# Enable on‐the‐fly CPU/GPU offloading; no manual .to("cuda")
 pipe.enable_model_cpu_offload()
-# Further reduce peak VRAM with VAE slicing
 pipe.vae.enable_slicing()
 
-# 2. Decorate only the generation function that needs GPU
-@spaces.GPU(duration=180)
-def generate_video(prompt: str) -> str:
+# 2. GPU-decorated generation function
+@spaces.GPU(duration=600)
+def generate_video(prompt: str, steps: int, frames: int, fps: int) -> str:
     """
-    Generates ~10s of video (161 frames @ default 16 FPS) from text.
+    Generates video from text with adjustable steps, frames, and fps.
     Returns path to the saved MP4 file.
     """
-    # Offloading moves weights automatically; just call the pipeline
+    # Run pipeline; offload handles device placement
     output = pipe(
         prompt=prompt,
-        num_inference_steps=3,
-        num_frames=50
+        num_inference_steps=steps,
+        num_frames=frames
     )
-    frames = output.frames[0]
+    frame_list = output.frames[0]
 
-    # Export frames to MP4 (H.264) for browser playback
-    return export_to_video(frames, "generated.mp4", fps=15)
+    # Export to MP4 for browser playback
+    return export_to_video(frame_list, "generated.mp4", fps=fps)
 
-# 3. Build the Gradio interface
-with gr.Blocks(title="CogVideoX Text-to-Video") as demo:
+# 3. Build the Gradio interface with sliders
+with gr.Blocks(title="CogVideoX Interactive Text-to-Video") as demo:
     gr.Markdown(
         """
-        # 🎞️ CogVideoX1.5-5B Text→Video
-        Enter a descriptive prompt below and click **Generate Video**
-        to create up to 10 seconds of high-quality video,
-        powered by ZeroGPU on Hugging Face Spaces.
+        # 🎞️ Interactive Text‑to‑Video Demo
+        Adjust the sliders below to control the number of diffusion steps,
+        total frames (length), and frames per second (fps) for your video.
         """
     )
-    with gr.Row():
+    with gr.Column():
         prompt_input = gr.Textbox(
             label="Prompt",
-            placeholder="e.g., A majestic eagle soaring above snowy peaks",
+            placeholder="A serene forest at dawn",
             lines=2
         )
+        steps_slider = gr.Slider(
+            minimum=1, maximum=100, step=1, value=25,
+            label="Inference Steps"
+        )
+        frames_slider = gr.Slider(
+            minimum=16, maximum=320, step=1, value=161,
+            label="Total Frames (approx. length)"
+        )
+        fps_slider = gr.Slider(
+            minimum=1, maximum=60, step=1, value=16,
+            label="Frames per Second (fps)"
+        )
         gen_button = gr.Button("Generate Video")
     video_output = gr.Video(label="Generated Video", format="mp4")
 
     gen_button.click(
         fn=generate_video,
-        inputs=prompt_input,
+        inputs=[prompt_input, steps_slider, frames_slider, fps_slider],
         outputs=video_output
     )
 
-# 4. Launch the app (SSR disabled to keep the process alive)
+# 4. Launch the app with SSR disabled
 if __name__ == "__main__":
     demo.launch(
         server_name="0.0.0.0",