Upload app.py with huggingface_hub

app.py

@@ -380,7 +380,6 @@ def preprocess_image(input: Image.Image) -> Image.Image:
     return output
 
 
-@spaces.GPU(duration=60)
 def preprocess_images(images):
     """
     Preprocess a list of input images for multi-image conditioning.
@@ -483,8 +482,18 @@ def split_image(image: Image.Image) -> List[Image.Image]:
     return [preprocess_image(image) for image in images]
 
 
-@spaces.GPU(duration=120)
+def preprocess_for_generate(multiimages):
+    """Preprocess images before GPU generation. No GPU needed — background
+    removal is a network call to the BRIA RMBG Space."""
+    if not multiimages:
+        raise gr.Error("Please upload images or select an example first.")
+    images = [image[0] for image in multiimages]
+    return [preprocess_image(img) for img in images]
+
+
+@spaces.GPU(duration=180)
 def image_to_3d(
+    processed_images,
     seed,
     resolution,
     ss_guidance_strength,
@@ -499,19 +508,11 @@ def image_to_3d(
     tex_slat_guidance_rescale,
     tex_slat_sampling_steps,
     tex_slat_rescale_t,
-    multiimages,
     multiimage_algo,
     tex_multiimage_algo,
     req: gr.Request,
     progress=gr.Progress(track_tqdm=True),
 ):
-    if not multiimages:
-        raise gr.Error("Please upload images or select an example first.")
-
-    # Preprocess images (background removal, cropping, etc.)
-    images = [image[0] for image in multiimages]
-    processed_images = [preprocess_image(img) for img in images]
-
     # --- Sampling ---
     outputs, latents = pipeline.run_multi_image(
         processed_images,
@@ -546,15 +547,19 @@ def image_to_3d(
     )
     mesh = outputs[0]
     mesh.simplify(16777216)  # nvdiffrast limit
-    images = render_utils.render_snapshot(mesh, resolution=1024, r=2, fov=36, nviews=STEPS, envmap=envmap)
+    render_images = render_utils.render_snapshot(mesh, resolution=1024, r=2, fov=36, nviews=STEPS, envmap=envmap)
     state = pack_state(latents)
     torch.cuda.empty_cache()
+    return state, render_images
+
 
-    # --- HTML Construction ---
-    # The Stack of 48 Images - encode in parallel for speed
+def build_preview_html(state, render_images):
+    """Encode rendered images to base64 and build the HTML preview.
+    No GPU needed — pure CPU work."""
+    # Encode 48 images in parallel
     def encode_preview_image(args):
         m_idx, s_idx, render_key = args
-        img_base64 = image_to_base64(Image.fromarray(images[render_key][s_idx]))
+        img_base64 = image_to_base64(Image.fromarray(render_images[render_key][s_idx]))
         return (m_idx, s_idx, img_base64)
 
     encode_tasks = [
@@ -587,7 +592,6 @@ def image_to_3d(
     btns_html = ""
     for idx, mode in enumerate(MODES):
         active_class = "active" if idx == DEFAULT_MODE else ""
-        # Note: onclick calls the JS function defined in Head
         btns_html += f"""
             <img src="{mode['icon_base64']}"
                  class="mode-btn {active_class}"
@@ -626,7 +630,7 @@ def image_to_3d(
     return state, full_html
 
 
-@spaces.GPU(duration=120)
+@spaces.GPU(duration=180)
 def extract_glb(
     state,
     decimation_target,
@@ -675,18 +679,19 @@ def extract_glb(
 
 with gr.Blocks(theme=gr.themes.Soft(primary_hue="orange", neutral_hue="slate"), css=css, head=head) as demo:
     gr.HTML("""
-    <div style="display: flex; align-items: center; gap: 20px; margin-bottom: 10px;">
-        <a href="https://www.opsiclear.com" target="_blank">
-            <img src="https://www.opsiclear.com/assets/logos/Logo_v2_compact_name.svg" alt="OpsiClear" style="height: 80px;">
+    <div style="display: flex; align-items: flex-start; gap: 24px; padding: 16px 20px; margin-bottom: 12px; border-radius: 12px; background: var(--block-background-fill); border: 1px solid var(--border-color-primary);">
+        <a href="https://www.opsiclear.com" target="_blank" style="flex-shrink: 0; display: flex; align-items: center;">
+            <img src="https://www.opsiclear.com/assets/logos/Logo_v2_compact_name.svg" alt="OpsiClear"
+                 style="width: 72px; height: 72px; object-fit: contain;">
         </a>
-        <div>
-            <h2 style="margin: 0;">Multi-View to 3D with <a href="https://microsoft.github.io/TRELLIS.2" target="_blank">TRELLIS.2</a></h2>
-            <ul style="margin: 5px 0; padding-left: 20px;">
+        <div style="min-width: 0;">
+            <h2 style="margin: 0 0 8px 0; font-size: 1.25rem; line-height: 1.3;">Multi-View to 3D with <a href="https://microsoft.github.io/TRELLIS.2" target="_blank" style="text-decoration: none; color: var(--color-accent);">TRELLIS.2</a></h2>
+            <ul style="margin: 0; padding-left: 18px; font-size: 0.9rem; line-height: 1.6; color: var(--body-text-color-subdued, var(--body-text-color));">
                 <li>Upload multiple images from different viewpoints to create a 3D asset with multi-image conditioning.</li>
                 <li>Click an example below to load a pre-made multi-view set, or upload your own images.</li>
                 <li>Click <b>Generate</b> to create the 3D model, then <b>Extract GLB</b> to export.</li>
-                <li style="color: #e67300;"><b>⚠️ Note:</b> Generation quality is highly sensitive to parameters. Adjust settings in Advanced Settings if results are unsatisfactory.</li>
-                <li style="color: #cc3333;"><b>⚠️ Non-Commercial:</b> This space uses models with licenses that <b>forbid commercial use</b> (BRIA RMBG-2.0: CC BY-NC 4.0, nvdiffrast/nvdiffrec: NVIDIA Source Code License).</li>
+                <li style="color: #e67300;"><b>Note:</b> Generation quality is highly sensitive to parameters. Adjust settings in Advanced Settings if results are unsatisfactory.</li>
+                <li style="color: #cc3333;"><b>Non-Commercial:</b> This space uses models with licenses that <b>forbid commercial use</b> (BRIA RMBG-2.0: CC BY-NC 4.0, nvdiffrast/nvdiffrec: NVIDIA Source Code License).</li>
             </ul>
         </div>
     </div>
@@ -745,6 +750,8 @@ with gr.Blocks(theme=gr.themes.Soft(primary_hue="orange", neutral_hue="slate"),
             )
 
     output_buf = gr.State()
+    processed_buf = gr.State()
+    render_buf = gr.State()
 
 
     # Handlers
@@ -760,15 +767,23 @@ with gr.Blocks(theme=gr.themes.Soft(primary_hue="orange", neutral_hue="slate"),
         get_seed,
         inputs=[randomize_seed, seed],
         outputs=[seed],
+    ).then(
+        preprocess_for_generate,
+        inputs=[multiimage_prompt],
+        outputs=[processed_buf],
     ).then(
         image_to_3d,
         inputs=[
-            seed, resolution,
+            processed_buf, seed, resolution,
             ss_guidance_strength, ss_guidance_rescale, ss_sampling_steps, ss_rescale_t,
             shape_slat_guidance_strength, shape_slat_guidance_rescale, shape_slat_sampling_steps, shape_slat_rescale_t,
             tex_slat_guidance_strength, tex_slat_guidance_rescale, tex_slat_sampling_steps, tex_slat_rescale_t,
-            multiimage_prompt, multiimage_algo, tex_multiimage_algo
+            multiimage_algo, tex_multiimage_algo
         ],
+        outputs=[output_buf, render_buf],
+    ).then(
+        build_preview_html,
+        inputs=[output_buf, render_buf],
         outputs=[output_buf, preview_output],
     )