Update compact try-on demo UI

.gitattributes

@@ -1,35 +1,2 @@
-*.7z filter=lfs diff=lfs merge=lfs -text
-*.arrow filter=lfs diff=lfs merge=lfs -text
-*.bin filter=lfs diff=lfs merge=lfs -text
-*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
-*.ftz filter=lfs diff=lfs merge=lfs -text
-*.gz filter=lfs diff=lfs merge=lfs -text
-*.h5 filter=lfs diff=lfs merge=lfs -text
-*.joblib filter=lfs diff=lfs merge=lfs -text
-*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
-*.onnx filter=lfs diff=lfs merge=lfs -text
-*.ot filter=lfs diff=lfs merge=lfs -text
-*.parquet filter=lfs diff=lfs merge=lfs -text
-*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
-*.pt filter=lfs diff=lfs merge=lfs -text
-*.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* filter=lfs diff=lfs merge=lfs -text
-*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text
-*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
-*.xz filter=lfs diff=lfs merge=lfs -text
-*.zip filter=lfs diff=lfs merge=lfs -text
-*.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.jpg filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,3 @@
+.DS_Store
+__pycache__/
+*.py[cod]

app.py

@@ -1,47 +1,180 @@
 from __future__ import annotations
 
+import base64
 import random
+import time
 import traceback
+from pathlib import Path
 
 import gradio as gr
 import spaces
-from PIL import Image
+from PIL import Image, ImageDraw, ImageFilter, ImageFont
 
 from inference import run_tryon
-from preprocess import load_garment_images, resize_for_demo
+from preprocess import resize_for_demo
 
 
-def _validate_inputs(person_image: Image.Image | None, garment_files) -> list[Image.Image]:
-    if person_image is None:
-        raise gr.Error("Please upload a person image.")
+APP_DIR = Path(__file__).resolve().parent
+LOGO_PATH = APP_DIR / "assets" / "tryon_logo_compact.png"
+SHOWCASE_DIR = APP_DIR / "assets" / "showcase"
+SHOWCASE_IMAGES = [SHOWCASE_DIR / f"case_{idx}.jpg" for idx in range(4)]
 
-    garment_images = load_garment_images(garment_files)
-    if not garment_images:
-        raise gr.Error("Please upload at least one garment reference image.")
+def _image_data_uri(path: Path) -> str:
+    if not path.exists():
+        return ""
+    mime = "image/jpeg" if path.suffix.lower() in {".jpg", ".jpeg"} else "image/png"
+    data = base64.b64encode(path.read_bytes()).decode("utf-8")
+    return f"data:{mime};base64,{data}"
 
-    return garment_images
+
+def _showcase_marquee_html() -> str:
+    cards = []
+    for idx, path in enumerate(SHOWCASE_IMAGES):
+        uri = _image_data_uri(path)
+        if uri:
+            cards.append(f'<img src="{uri}" alt="Try-on showcase {idx + 1}" />')
+    row = "".join(cards)
+    return f"""
+    <div class="showcase-marquee" aria-label="效果展示">
+      <div class="showcase-track">{row}{row}</div>
+    </div>
+    """
+
+
+def _file_path(item) -> str | None:
+    if item is None:
+        return None
+    if isinstance(item, str):
+        return item
+    if isinstance(item, dict):
+        return item.get("path") or item.get("name")
+    return getattr(item, "path", None) or getattr(item, "name", None)
+
+
+def _parse_user_box(message) -> tuple[Image.Image, list[Image.Image], str]:
+    if not message:
+        raise gr.Error("请在输入框里上传图片并写下试穿指令。")
+
+    text = ""
+    files = []
+    if isinstance(message, dict):
+        text = message.get("text") or ""
+        files = message.get("files") or []
+    else:
+        text = str(message)
+
+    paths = [path for path in (_file_path(item) for item in files) if path]
+    if len(paths) < 2:
+        raise gr.Error("请至少上传 2 张图：第 1 张人物图，后面的图作为服装参考图。")
+
+    person = Image.open(paths[0]).convert("RGB")
+    garments = [Image.open(path).convert("RGB") for path in paths[1:]]
+    return person, garments, text.strip()
+
+
+def _font(size: int, bold: bool = False):
+    candidates = [
+        "/usr/share/fonts/opentype/noto/NotoSansCJK-Bold.ttc" if bold else "/usr/share/fonts/opentype/noto/NotoSansCJK-Regular.ttc",
+        "/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf" if bold else "/usr/share/fonts/truetype/dejavu/DejaVuSans.ttf",
+        "/System/Library/Fonts/Supplemental/Arial Bold.ttf" if bold else "/System/Library/Fonts/Supplemental/Arial.ttf",
+    ]
+    for path in candidates:
+        try:
+            return ImageFont.truetype(path, size)
+        except OSError:
+            continue
+    return ImageFont.load_default()
+
+
+def _cover(image: Image.Image, size: tuple[int, int]) -> Image.Image:
+    image = image.convert("RGB")
+    source_ratio = image.width / image.height
+    target_ratio = size[0] / size[1]
+    if source_ratio > target_ratio:
+        new_height = size[1]
+        new_width = int(new_height * source_ratio)
+    else:
+        new_width = size[0]
+        new_height = int(new_width / source_ratio)
+    resized = image.resize((new_width, new_height), Image.Resampling.LANCZOS)
+    left = (new_width - size[0]) // 2
+    top = (new_height - size[1]) // 2
+    return resized.crop((left, top, left + size[0], top + size[1]))
+
+
+def _rounded_image(image: Image.Image, radius: int) -> Image.Image:
+    mask = Image.new("L", image.size, 0)
+    draw = ImageDraw.Draw(mask)
+    draw.rounded_rectangle((0, 0, image.width, image.height), radius=radius, fill=255)
+    rounded = image.convert("RGBA")
+    rounded.putalpha(mask)
+    return rounded
+
+
+def _draw_badge(draw: ImageDraw.ImageDraw, xy: tuple[int, int], text: str):
+    font = _font(18, bold=True)
+    left, top = xy
+    bbox = draw.textbbox((0, 0), text, font=font)
+    width = bbox[2] - bbox[0] + 28
+    height = bbox[3] - bbox[1] + 14
+    draw.rounded_rectangle((left, top, left + width, top + height), radius=18, fill=(17, 23, 46))
+    draw.text((left + 14, top + 6), text, font=font, fill=(255, 255, 255))
+
+
+def _paste_panel(canvas: Image.Image, image: Image.Image, box: tuple[int, int, int, int], label: str):
+    left, top, right, bottom = box
+    panel = _cover(image, (right - left, bottom - top))
+    panel = _rounded_image(panel, radius=18)
+    canvas.alpha_composite(panel, (left, top))
+    draw = ImageDraw.Draw(canvas)
+    draw.rounded_rectangle((left, top, right, bottom), radius=18, outline=(228, 228, 228), width=2)
+    _draw_badge(draw, (left + 14, top + 12), label)
+
+
+def _result_card(person: Image.Image, garment: Image.Image, result: Image.Image) -> Image.Image:
+    width, height = 1180, 680
+    card_left, card_top, card_right, card_bottom = 60, 58, width - 60, height - 58
+
+    canvas = Image.new("RGBA", (width, height), (255, 255, 255, 0))
+    shadow = Image.new("RGBA", (width, height), (255, 255, 255, 0))
+    shadow_draw = ImageDraw.Draw(shadow)
+    shadow_draw.rounded_rectangle(
+        (card_left, card_top + 18, card_right, card_bottom + 18),
+        radius=58,
+        fill=(0, 0, 0, 55),
+    )
+    shadow = shadow.filter(ImageFilter.GaussianBlur(22))
+    canvas.alpha_composite(shadow)
+
+    draw = ImageDraw.Draw(canvas)
+    draw.rounded_rectangle((card_left, card_top, card_right, card_bottom), radius=58, fill=(250, 250, 250, 255))
+
+    gap = 24
+    left_col = (card_left + 34, card_top + 34, card_left + 420, card_bottom - 34)
+    right_panel = (left_col[2] + gap, left_col[1], card_right - 34, left_col[3])
+    left_h = (left_col[3] - left_col[1] - gap) // 2
+    person_panel = (left_col[0], left_col[1], left_col[2], left_col[1] + left_h)
+    garment_panel = (left_col[0], person_panel[3] + gap, left_col[2], left_col[3])
+
+    _paste_panel(canvas, person, person_panel, "User Image")
+    _paste_panel(canvas, garment, garment_panel, "Clothing Image")
+    _paste_panel(canvas, result, right_panel, "Virtual Try-On Result")
+
+    draw = ImageDraw.Draw(canvas)
+    draw.ellipse((card_left + 32, card_bottom - 98, card_left + 132, card_bottom + 2), fill=(194, 194, 194, 220))
+    draw.text((card_left + 62, card_bottom - 63), "编辑", font=_font(28, bold=True), fill=(255, 255, 255))
+    draw.ellipse((card_right - 116, card_bottom - 98, card_right - 16, card_bottom + 2), fill=(178, 178, 178, 230))
+    draw.text((card_right - 82, card_bottom - 73), "↥", font=_font(42, bold=True), fill=(255, 255, 255))
+
+    return canvas.convert("RGB")
 
 
 @spaces.GPU(size="xlarge", duration=180)
-def predict(
-    person_image: Image.Image,
-    garment_files,
-    garment_type: str,
-    primary_garment_index: int,
-    prompt: str,
-    prompt_enhancer: bool,
-    negative_prompt: str,
-    seed: int,
-    randomize_seed: bool,
-    num_inference_steps: int,
-    guidance_scale: float,
-    crop_and_paste: bool,
-):
+def predict_from_box(message):
     try:
-        garment_images = _validate_inputs(person_image, garment_files)
-
-        if randomize_seed:
-            seed = random.randint(0, 2**31 - 1)
+        started_at = time.time()
+        person_image, garment_images, prompt = _parse_user_box(message)
+        seed = random.randint(0, 2**31 - 1)
 
         person = resize_for_demo(person_image, max_side=1024)
         garments = [resize_for_demo(img, max_side=1024) for img in garment_images]
@@ -49,18 +182,21 @@ def predict(
         result, used_index, enhanced_prompt, pe_status, status = run_tryon(
             person_image=person,
             garment_images=garments,
-            garment_type=garment_type,
-            primary_garment_index=int(primary_garment_index),
+            garment_type="full_body",
+            primary_garment_index=0,
             seed=int(seed),
-            num_inference_steps=int(num_inference_steps),
-            guidance_scale=float(guidance_scale),
-            auto_crop=bool(crop_and_paste),
+            num_inference_steps=30,
+            guidance_scale=6.0,
+            auto_crop=False,
             prompt=prompt,
-            prompt_enhancer=bool(prompt_enhancer),
-            negative_prompt=negative_prompt,
+            prompt_enhancer=True,
+            negative_prompt="",
         )
 
-        return result, int(seed), used_index, enhanced_prompt, pe_status, status
+        result_card = _result_card(person, garments[used_index], result)
+        elapsed = max(1, int(time.time() - started_at))
+        display_status = f"Thought for {elapsed}s > {status}"
+        return result_card, int(seed), used_index, enhanced_prompt, pe_status, display_status
 
     except gr.Error:
         raise
@@ -69,111 +205,213 @@ def predict(
         raise gr.Error(f"Inference failed: {exc}") from exc
 
 
-with gr.Blocks(title="Virtual Try-On ZeroGPU Demo") as demo:
-    gr.Markdown(
-        """
-        # Virtual Try-On Demo
+CSS = """
+:root {
+  --tryon-border: #dedede;
+  --tryon-muted: #6f6f6f;
+}
+body, .gradio-container {
+  background: #fff !important;
+  color: #111 !important;
+  font-family: Inter, ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif;
+}
+.gradio-container {
+  max-width: none !important;
+  min-height: 100vh;
+  overflow-x: hidden;
+}
+.tryon-topbar {
+  position: fixed;
+  z-index: 20;
+  top: 24px;
+  left: 34px;
+  display: flex;
+  align-items: center;
+}
+.tryon-logo {
+  width: 124px;
+  height: 124px;
+  border-radius: 26px;
+  object-fit: cover;
+  box-shadow: 0 10px 30px rgba(0,0,0,.08);
+}
+.tryon-hero {
+  min-height: auto;
+  display: flex;
+  flex-direction: column;
+  justify-content: center;
+  align-items: center;
+  padding: 92px 20px 24px;
+}
+.tryon-title {
+  font-size: 32px;
+  line-height: 1.2;
+  font-weight: 760;
+  letter-spacing: 0;
+  text-align: center;
+  margin-bottom: 18px;
+}
+.prompt-shell {
+  max-width: 920px;
+  width: min(920px, calc(100vw - 36px));
+  margin: 0 auto;
+}
+.prompt-shell .wrap,
+.prompt-shell .block {
+  border-radius: 32px !important;
+  border-color: #b9d2ff !important;
+  box-shadow: 0 16px 44px rgba(0,0,0,.07) !important;
+}
+.prompt-shell textarea,
+.prompt-shell [contenteditable="true"] {
+  min-height: 138px !important;
+  border-radius: 32px !important;
+  border: 1px solid #b9d2ff !important;
+  box-shadow: 0 16px 44px rgba(0,0,0,.07) !important;
+  font-size: 22px !important;
+  line-height: 1.45 !important;
+  padding: 28px 88px 56px 28px !important;
+}
+.prompt-shell label {
+  display: none !important;
+}
+.prompt-shell button[title*="Upload"],
+.prompt-shell button[aria-label*="Upload"],
+.prompt-shell button[title*="上传"],
+.prompt-shell button[aria-label*="上传"] {
+  position: absolute !important;
+  left: 26px !important;
+  bottom: 20px !important;
+  width: 42px !important;
+  height: 42px !important;
+  border-radius: 999px !important;
+  background: transparent !important;
+  border: 0 !important;
+  font-size: 28px !important;
+}
+.prompt-shell button[type="submit"],
+.prompt-shell button[aria-label*="Submit"],
+.prompt-shell button[title*="Submit"] {
+  position: absolute !important;
+  right: 24px !important;
+  bottom: 18px !important;
+  width: 48px !important;
+  height: 48px !important;
+  border-radius: 999px !important;
+  font-size: 22px !important;
+  background: #f3f3f3 !important;
+  color: #111 !important;
+  border: 0 !important;
+}
+.workspace {
+  max-width: 1180px;
+  margin: 0 auto;
+  padding: 8px 24px 18px;
+}
+.result-card-output {
+  border: 0 !important;
+  background: transparent !important;
+}
+.result-card-output img {
+  border-radius: 34px !important;
+  max-height: 620px !important;
+  object-fit: contain !important;
+  box-shadow: none !important;
+}
+.result-meta textarea,
+.result-meta input {
+  color: #8b8b8b !important;
+  border: 0 !important;
+  background: transparent !important;
+}
+.section-title {
+  max-width: 1360px;
+  margin: 0 auto 14px;
+  padding: 0 24px;
+  font-size: 22px;
+  font-weight: 700;
+}
+.showcase-wrap {
+  width: 100%;
+  overflow: hidden;
+  padding: 0 0 30px;
+}
+.mode-preview img, .showcase-marquee img {
+  border-radius: 14px !important;
+}
+.showcase-marquee {
+  width: 100%;
+  overflow: hidden;
+  mask-image: linear-gradient(to right, transparent, black 7%, black 93%, transparent);
+}
+.showcase-track {
+  display: flex;
+  width: max-content;
+  gap: 24px;
+  animation: showcase-scroll 48s linear infinite;
+}
+.showcase-track:hover {
+  animation-play-state: paused;
+}
+.showcase-track img {
+  width: min(42vw, 560px);
+  min-width: 420px;
+  aspect-ratio: 16 / 9;
+  object-fit: cover;
+  box-shadow: 0 12px 36px rgba(0,0,0,.08);
+}
+@keyframes showcase-scroll {
+  from { transform: translateX(0); }
+  to { transform: translateX(calc(-50% - 12px)); }
+}
+@media (max-width: 760px) {
+  .tryon-topbar { top: 14px; left: 16px; }
+  .tryon-logo { width: 78px; height: 78px; border-radius: 18px; }
+  .tryon-hero { padding-top: 104px; }
+  .tryon-title { font-size: 24px; }
+  .prompt-shell textarea, .prompt-shell [contenteditable="true"] { font-size: 18px !important; min-height: 126px !important; }
+  .showcase-track img { width: 320px; min-width: 320px; }
+}
+"""
+
 
-        Upload one person image and one or more garment reference images.
-        The model uses Picture 1 as the garment reference and Picture 2 as the person image.
-        If you upload multiple garments, the selected primary garment index is used.
+with gr.Blocks(title="JoyAI Virtual Try-On", css=CSS) as demo:
+    logo_uri = _image_data_uri(LOGO_PATH)
+    gr.HTML(
+        f"""
+        <div class="tryon-topbar">
+          <img class="tryon-logo" src="{logo_uri}" alt="JoyAI Try-On" />
+        </div>
         """
     )
 
-    with gr.Row():
-        with gr.Column(scale=1):
-            person_image = gr.Image(
-                label="Person Image",
-                type="pil",
-                sources=["upload", "clipboard"],
-            )
-
-            garment_files = gr.File(
-                label="Garment Reference Images",
-                file_count="multiple",
-                file_types=["image"],
-                type="filepath",
-            )
-
-            garment_type = gr.Radio(
-                choices=["upper_body", "lower_body", "dress", "full_body"],
-                value="upper_body",
-                label="Garment Type",
-            )
-
-            primary_garment_index = gr.Slider(
-                minimum=0,
-                maximum=9,
-                value=0,
-                step=1,
-                label="Primary Garment Index",
-            )
-
-            prompt = gr.Textbox(
-                label="Original Try-On Prompt",
-                value="",
-                lines=3,
-                placeholder="Leave empty to use the default prompt for the selected garment type.",
-            )
-            prompt_enhancer = gr.Checkbox(value=True, label="Prompt Enhancer")
-
-            with gr.Accordion("Advanced Settings", open=False):
-                seed = gr.Number(value=42, precision=0, label="Seed")
-                randomize_seed = gr.Checkbox(value=True, label="Randomize Seed")
-                num_inference_steps = gr.Slider(
-                    minimum=10,
-                    maximum=80,
-                    value=30,
-                    step=1,
-                    label="Inference Steps",
-                )
-                guidance_scale = gr.Slider(
-                    minimum=0.0,
-                    maximum=15.0,
-                    value=6.0,
-                    step=0.1,
-                    label="Guidance Scale",
-                )
-                crop_and_paste = gr.Checkbox(
-                    value=False,
-                    label="Enable crop-and-paste when body bbox tools are available",
-                )
-                negative_prompt = gr.Textbox(
-                    value="",
-                    label="Negative Prompt",
-                    lines=3,
-                    placeholder="Leave empty to use the default negative prompt from tryon_infer.py.",
-                )
-
-            run_button = gr.Button("Generate Try-On", variant="primary")
-
-        with gr.Column(scale=1):
-            output_image = gr.Image(label="Try-On Result", type="pil")
+    with gr.Column(elem_classes=["tryon-hero"]):
+        gr.HTML('<div class="tryon-title">开始虚拟试穿之旅，放入图像或写指令来看效果</div>')
+        user_box = gr.MultimodalTextbox(
+            label="",
+            file_count="multiple",
+            file_types=["image"],
+            placeholder="描述你想要的试穿效果",
+            submit_btn="↑",
+            elem_classes=["prompt-shell"],
+        )
+
+    with gr.Column(elem_classes=["workspace"]):
+        status = gr.Textbox(label="", interactive=False, elem_classes=["result-meta"])
+        output_image = gr.Image(label="", type="pil", elem_classes=["result-card-output"])
+        enhanced_prompt = gr.Textbox(label="PE 后完整文本", interactive=False, lines=3, elem_classes=["result-meta"])
+        pe_status = gr.Textbox(label="Prompt Enhancer 状态", interactive=False, elem_classes=["result-meta"])
+        with gr.Accordion("运行信息", open=False):
             used_seed = gr.Number(label="Used Seed", precision=0)
             used_garment_index = gr.Number(label="Used Garment Index", precision=0)
-            enhanced_prompt = gr.Textbox(label="Enhanced Prompt Used", interactive=False, lines=5)
-            pe_status = gr.Textbox(label="Prompt Enhancer Status", interactive=False)
-            status = gr.Textbox(label="Status", interactive=False)
-
-    run_button.click(
-        fn=predict,
-        inputs=[
-            person_image,
-            garment_files,
-            garment_type,
-            primary_garment_index,
-            prompt,
-            prompt_enhancer,
-            negative_prompt,
-            seed,
-            randomize_seed,
-            num_inference_steps,
-            guidance_scale,
-            crop_and_paste,
-        ],
-        outputs=[output_image, used_seed, used_garment_index, enhanced_prompt, pe_status, status],
-        api_name="tryon",
-    )
+
+    gr.HTML('<div class="section-title">效果展示</div>')
+    with gr.Column(elem_classes=["showcase-wrap"]):
+        gr.HTML(_showcase_marquee_html())
+
+    predict_outputs = [output_image, used_seed, used_garment_index, enhanced_prompt, pe_status, status]
+
+    user_box.submit(fn=predict_from_box, inputs=[user_box], outputs=predict_outputs, api_name="tryon")
 
 
 if __name__ == "__main__":

app_v1.py

@@ -0,0 +1,180 @@
+from __future__ import annotations
+
+import random
+import traceback
+
+import gradio as gr
+import spaces
+from PIL import Image
+
+from inference import run_tryon
+from preprocess import load_garment_images, resize_for_demo
+
+
+def _validate_inputs(person_image: Image.Image | None, garment_files) -> list[Image.Image]:
+    if person_image is None:
+        raise gr.Error("Please upload a person image.")
+
+    garment_images = load_garment_images(garment_files)
+    if not garment_images:
+        raise gr.Error("Please upload at least one garment reference image.")
+
+    return garment_images
+
+
+@spaces.GPU(size="xlarge", duration=180)
+def predict(
+    person_image: Image.Image,
+    garment_files,
+    garment_type: str,
+    primary_garment_index: int,
+    prompt: str,
+    prompt_enhancer: bool,
+    negative_prompt: str,
+    seed: int,
+    randomize_seed: bool,
+    num_inference_steps: int,
+    guidance_scale: float,
+    crop_and_paste: bool,
+):
+    try:
+        garment_images = _validate_inputs(person_image, garment_files)
+
+        if randomize_seed:
+            seed = random.randint(0, 2**31 - 1)
+
+        person = resize_for_demo(person_image, max_side=1024)
+        garments = [resize_for_demo(img, max_side=1024) for img in garment_images]
+
+        result, used_index, enhanced_prompt, pe_status, status = run_tryon(
+            person_image=person,
+            garment_images=garments,
+            garment_type=garment_type,
+            primary_garment_index=int(primary_garment_index),
+            seed=int(seed),
+            num_inference_steps=int(num_inference_steps),
+            guidance_scale=float(guidance_scale),
+            auto_crop=bool(crop_and_paste),
+            prompt=prompt,
+            prompt_enhancer=bool(prompt_enhancer),
+            negative_prompt=negative_prompt,
+        )
+
+        return result, int(seed), used_index, enhanced_prompt, pe_status, status
+
+    except gr.Error:
+        raise
+    except Exception as exc:
+        traceback.print_exc()
+        raise gr.Error(f"Inference failed: {exc}") from exc
+
+
+with gr.Blocks(title="Virtual Try-On ZeroGPU Demo") as demo:
+    gr.Markdown(
+        """
+        # Virtual Try-On Demo
+
+        Upload one person image and one or more garment reference images.
+        The model uses Picture 1 as the garment reference and Picture 2 as the person image.
+        If you upload multiple garments, the selected primary garment index is used.
+        """
+    )
+
+    with gr.Row():
+        with gr.Column(scale=1):
+            person_image = gr.Image(
+                label="Person Image",
+                type="pil",
+                sources=["upload", "clipboard"],
+            )
+
+            garment_files = gr.File(
+                label="Garment Reference Images",
+                file_count="multiple",
+                file_types=["image"],
+                type="filepath",
+            )
+
+            garment_type = gr.Radio(
+                choices=["upper_body", "lower_body", "dress", "full_body"],
+                value="upper_body",
+                label="Garment Type",
+            )
+
+            primary_garment_index = gr.Slider(
+                minimum=0,
+                maximum=9,
+                value=0,
+                step=1,
+                label="Primary Garment Index",
+            )
+
+            prompt = gr.Textbox(
+                label="Original Try-On Prompt",
+                value="",
+                lines=3,
+                placeholder="Leave empty to use the default prompt for the selected garment type.",
+            )
+            prompt_enhancer = gr.Checkbox(value=True, label="Prompt Enhancer")
+
+            with gr.Accordion("Advanced Settings", open=False):
+                seed = gr.Number(value=42, precision=0, label="Seed")
+                randomize_seed = gr.Checkbox(value=True, label="Randomize Seed")
+                num_inference_steps = gr.Slider(
+                    minimum=10,
+                    maximum=80,
+                    value=30,
+                    step=1,
+                    label="Inference Steps",
+                )
+                guidance_scale = gr.Slider(
+                    minimum=0.0,
+                    maximum=15.0,
+                    value=6.0,
+                    step=0.1,
+                    label="Guidance Scale",
+                )
+                crop_and_paste = gr.Checkbox(
+                    value=False,
+                    label="Enable crop-and-paste when body bbox tools are available",
+                )
+                negative_prompt = gr.Textbox(
+                    value="",
+                    label="Negative Prompt",
+                    lines=3,
+                    placeholder="Leave empty to use the default negative prompt from tryon_infer.py.",
+                )
+
+            run_button = gr.Button("Generate Try-On", variant="primary")
+
+        with gr.Column(scale=1):
+            output_image = gr.Image(label="Try-On Result", type="pil")
+            used_seed = gr.Number(label="Used Seed", precision=0)
+            used_garment_index = gr.Number(label="Used Garment Index", precision=0)
+            enhanced_prompt = gr.Textbox(label="Enhanced Prompt Used", interactive=False, lines=5)
+            pe_status = gr.Textbox(label="Prompt Enhancer Status", interactive=False)
+            status = gr.Textbox(label="Status", interactive=False)
+
+    run_button.click(
+        fn=predict,
+        inputs=[
+            person_image,
+            garment_files,
+            garment_type,
+            primary_garment_index,
+            prompt,
+            prompt_enhancer,
+            negative_prompt,
+            seed,
+            randomize_seed,
+            num_inference_steps,
+            guidance_scale,
+            crop_and_paste,
+        ],
+        outputs=[output_image, used_seed, used_garment_index, enhanced_prompt, pe_status, status],
+        api_name="tryon",
+    )
+
+
+if __name__ == "__main__":
+    demo.queue(max_size=20).launch()

packages.txt

@@ -2,4 +2,4 @@ ffmpeg
 libgl1
 libglib2.0-0
 git
-
+fonts-noto-cjk