Upload app.py with huggingface_hub

app.py

@@ -27,8 +27,11 @@ pipe = JoyImageEditPlusPipeline.from_pretrained(
 pipe.enable_model_cpu_offload()
 
 
-def _estimate_duration(images, prompt, num_inference_steps, guidance_scale, seed, negative_prompt, progress):
-    """Estimate the GPU duration needed based on step count and image count."""
+def _estimate_duration(images, prompt, num_inference_steps, guidance_scale, seed, negative_prompt, randomize_seed, *args, **kwargs):
+    """Estimate the GPU duration needed based on step count and image count.
+
+    Accepts *args/**kwargs because Gradio passes progress= positionally.
+    """
     n_imgs = len(images) if images else 0
     # ~1s per step for 16B transformer + 8B text encoder + VAE, plus overhead
     return min(300, 60 + int(num_inference_steps * 3.5) + n_imgs * 10)
@@ -67,8 +70,11 @@ def generate(
     if isinstance(images, list):
         pil_images = []
         for img in images:
-            if isinstance(img, dict) and "path" in img:
-                pil_images.append(Image.open(img["path"]).convert("RGB"))
+            if isinstance(img, dict):
+                # Gallery passes {'image': filepath, 'caption': ...} or {'path': ...}
+                path = img.get("image") or img.get("path")
+                if path:
+                    pil_images.append(Image.open(path).convert("RGB"))
             elif isinstance(img, str):
                 pil_images.append(Image.open(img).convert("RGB"))
             elif isinstance(img, Image.Image):