demo update

.gitattributes

@@ -41,3 +41,6 @@ examples/ai-valley.jpg filter=lfs diff=lfs merge=lfs -text
 examples/istock-motorbike-rice-field.jpg filter=lfs diff=lfs merge=lfs -text
 examples/omni-misty-forest.jpg filter=lfs diff=lfs merge=lfs -text
 examples/omni-horse-trail.jpg filter=lfs diff=lfs merge=lfs -text
+examples/omni-greek-square.jpg filter=lfs diff=lfs merge=lfs -text
+examples/istock-country-road.jpg filter=lfs diff=lfs merge=lfs -text
+examples/istock-skier.jpg filter=lfs diff=lfs merge=lfs -text

app.py

@@ -539,6 +539,32 @@ EXAMPLES = [
         "idyllic, with greenery and natural elements dominating the scene.",
         "orbit_left",
     ],
+    [
+        str(HERE / "examples" / "omni-greek-square.jpg"),
+        "A lively ancient Greek market square with stone-paved pathways, wooden crates, and "
+        "fabric-draped stalls full of clay pottery. The setting features classical architecture, "
+        "including a prominent colonnaded building on the left and a vast cityscape stretching into "
+        "the distance. The clear weather and bright sunlight enhance the vividness of the scene, "
+        "creating a dynamic and immersive atmosphere.",
+        "crane_up_fwd",
+    ],
+    [
+        str(HERE / "examples" / "istock-country-road.jpg"),
+        "A white car driving on a winding road surrounded by lush green grass, seen from an aerial "
+        "perspective. The road is narrow and curves gently through the landscape. The grass on "
+        "either side is vibrant and well-maintained, with some patches of darker green. The overall "
+        "scene is serene and picturesque, with the car travelling through the greenery.",
+        "dolly_in",
+    ],
+    [
+        str(HERE / "examples" / "istock-skier.jpg"),
+        "A snowy mountain landscape under a clear blue sky. A person is skiing down a well-groomed "
+        "slope, leaving tracks in the snow. The skier is dressed in dark clothing and is using ski "
+        "poles for balance. In the background, there are snow-covered mountains with rocky "
+        "outcrops. A ski lift with red support towers is visible to the left, and a small building "
+        "is seen at the bottom of the slope.",
+        "orbit_left",
+    ],
     [
         str(HERE / "examples" / "ai-airmountains.jpg"),
         "A vast sky filled with multiple floating islands of different sizes, suspended above a "
@@ -592,6 +618,20 @@ CHARACTER_EXAMPLES = [
     ],
 ]
 
+# A compact image-only strip for the left rail — every showcase frame, in the same order.
+# Clicking a thumbnail loads its image, prompt and camera move into the input boxes.
+QUICK_PICKS = CHARACTER_EXAMPLES + EXAMPLES
+QUICK_PICK_GALLERY = [
+    (row[0], Path(row[0]).stem.replace("istock-", "").replace("ai-", "").replace("omni-", "").replace("-", " "))
+    for row in QUICK_PICKS
+]
+
+
+def load_quick_pick(event: gr.SelectData) -> tuple[str, str, str]:
+    image, prompt, trajectory = QUICK_PICKS[event.index]
+    return image, prompt, trajectory
+
+
 CSS = """
 #col-container { margin: 0 auto; max-width: 1180px; }
 .dark .gradio-container { color: var(--body-text-color); }
@@ -610,68 +650,62 @@ with gr.Blocks() as demo:
             Drop in an image, pick a camera move, and (optionally) describe the scene — you get
             an 81-frame, 832x480, 16 fps clip that follows that trajectory. Leave the description
             blank and a small vision-language model writes one from your image automatically.
+
+            *To fit a ZeroGPU slot this Space uses a fast few-step, fp8 setting, which can
+            slightly soften quality; for the highest-fidelity results you're welcome to use the
+            full [reference code](https://github.com/TencentARC/SCoPE).*
             """
         )
         with gr.Row():
-            with gr.Column(scale=1):
-                image_input = gr.Image(
-                    label="First frame", type="filepath", height=300, sources=["upload", "clipboard"]
-                )
-                prompt_input = gr.Textbox(
-                    label="Scene description",
-                    placeholder="Describe what is in the image… (leave empty to auto-caption it)",
-                    lines=4,
-                    info="Optional — if left blank, Qwen2-VL writes a caption from your image.",
+            with gr.Column(scale=1, min_width=110):
+                quick_picks = gr.Gallery(
+                    value=QUICK_PICK_GALLERY,
+                    label="Quick picks — click to load",
+                    columns=1,
+                    height=600,
+                    object_fit="cover",
+                    allow_preview=False,
                 )
-                trajectory_input = gr.Dropdown(
-                    label="Camera move",
-                    choices=PRESETS,
-                    value="dolly_in",
-                )
-                run_button = gr.Button("Generate video", variant="primary")
-            with gr.Column(scale=1):
-                video_output = gr.Video(
-                    label="Generated video", autoplay=True, loop=True, height=300
-                )
-                path_preview = gr.Image(
-                    label="Camera path (start green, end red)",
-                    height=280,
-                    interactive=False,
-                )
-                status_output = gr.Markdown()
-
-        gr.Examples(
-            examples=EXAMPLES,
-            inputs=[image_input, prompt_input, trajectory_input],
-            outputs=[video_output, seed_input, status_output],
-            fn=generate,
-            cache_examples=True,
-            cache_mode="lazy",
-            label="Examples — click a row to load its image, prompt and camera move",
-            examples_per_page=8,
-        )
-        gr.Examples(
-            examples=CHARACTER_EXAMPLES,
-            inputs=[image_input, prompt_input, trajectory_input],
-            outputs=[video_output, seed_input, status_output],
-            fn=generate,
-            cache_examples=True,
-            cache_mode="lazy",
-            label="Examples with characters",
-            examples_per_page=8,
-        )
+            with gr.Column(scale=5):
+                with gr.Row():
+                    with gr.Column(scale=1):
+                        image_input = gr.Image(
+                            label="First frame", type="filepath", height=300,
+                            sources=["upload", "clipboard"],
+                        )
+                        prompt_input = gr.Textbox(
+                            label="Scene description",
+                            placeholder="Describe what is in the image… (leave empty to auto-caption it)",
+                            lines=4,
+                            info="Optional — if left blank, Qwen2-VL writes a caption from your image.",
+                        )
+                        trajectory_input = gr.Dropdown(
+                            label="Camera move", choices=PRESETS, value="dolly_in",
+                        )
+                        run_button = gr.Button("Generate video", variant="primary")
+                    with gr.Column(scale=1):
+                        video_output = gr.Video(
+                            label="Generated video", autoplay=True, loop=True, height=300
+                        )
+                        path_preview = gr.Image(
+                            label="Camera path (start green, end red)",
+                            height=280,
+                            interactive=False,
+                        )
+                        status_output = gr.Markdown()
 
         with gr.Accordion("Advanced settings", open=False):
             with gr.Row():
                 steps_input = gr.Slider(
                     label="Sampling steps",
                     minimum=4,
-                    maximum=8,
+                    maximum=16,
                     step=1,
                     value=4,
                     info=(
                         "The distillation LoRA is trained for 4 steps (2 high-noise + "
-                        "2 low-noise) — about 68s. Each extra step adds ~14s of GPU time."
+                        "2 low-noise) — about 68s. Each extra step adds ~14s of GPU time; "
+                        "gains past ~8 steps are marginal for the distilled LoRA."
                     ),
                 )
                 motion_input = gr.Slider(
@@ -696,6 +730,27 @@ with gr.Blocks() as demo:
                 )
                 randomize_input = gr.Checkbox(label="Randomize seed", value=True)
 
+        gr.Examples(
+            examples=CHARACTER_EXAMPLES,
+            inputs=[image_input, prompt_input, trajectory_input],
+            outputs=[video_output, seed_input, status_output],
+            fn=generate,
+            cache_examples=True,
+            cache_mode="lazy",
+            label="Camera control with people in the scene — click a row to load it",
+            examples_per_page=8,
+        )
+        gr.Examples(
+            examples=EXAMPLES,
+            inputs=[image_input, prompt_input, trajectory_input],
+            outputs=[video_output, seed_input, status_output],
+            fn=generate,
+            cache_examples=True,
+            cache_mode="lazy",
+            label="More examples — click a row to load its image, prompt and camera move",
+            examples_per_page=8,
+        )
+
         gr.Markdown(
             """
             **Notes** · Camera paths are OpenCV camera-to-world matrices `[81, 3, 4]` relative to
@@ -709,6 +764,9 @@ with gr.Blocks() as demo:
             """
         )
 
+    quick_picks.select(
+        load_quick_pick, None, [image_input, prompt_input, trajectory_input], show_progress="hidden"
+    )
     preview_inputs = [trajectory_input, motion_input]
     trajectory_input.change(preview_path, preview_inputs, path_preview, show_progress="hidden")
     motion_input.change(preview_path, preview_inputs, path_preview, show_progress="hidden")