Fix: System-level monkeypatch for Gradio stability

app.py

@@ -1,8 +1,8 @@
 import sys
 import os
 
-# --- PARCHES DE COMPATIBILIDAD CRÍTICOS ---
-# 1. Parche para audioop (Python 3.13)
+# --- PARCHE QUIRÚRGICO DE NIVEL SISTEMA ---
+# 1. Parche de audioop para Python 3.13
 try:
     import audioop
 except ImportError:
@@ -12,15 +12,14 @@ except ImportError:
     except ImportError:
         pass
 
-# 2. Parche para HfFolder (removido en hf_hub nuevo)
-import huggingface_hub
-if not hasattr(huggingface_hub, "HfFolder"):
-    class MockHfFolder:
-        @staticmethod
-        def get_token(): return os.getenv("HF_TOKEN")
-        @staticmethod
-        def save_token(token): pass
-    huggingface_hub.HfFolder = MockHfFolder
+# 2. Parche para el bug de Gradio (TypeError: argument of type 'bool' is not iterable)
+import gradio_client.utils as client_utils
+original_get_type = client_utils.get_type
+def patched_get_type(schema):
+    if isinstance(schema, bool):
+        return "str"  # Evita el colapso si schema es un booleano
+    return original_get_type(schema)
+client_utils.get_type = patched_get_type
 # ------------------------------------------
 
 import spaces
@@ -36,6 +35,9 @@ LTX_MODEL  = "Lightricks/LTX-Video"
 LTX_NSFW_LORA = "Lora-Daddy/Ltx2.3-real-nudity-early-alpha-30k-steps"
 NEG_DEFAULT = "blurry, low quality, bad anatomy, deformed, ugly, watermark, text"
 
+pipe_t2i = None
+pipe_video = None
+
 def load_t2i(lora_id=None, lora_scale=1.0):
     from diffusers import StableDiffusionXLPipeline
     pipe = StableDiffusionXLPipeline.from_pretrained(
@@ -56,11 +58,12 @@ def load_video():
     except: pass
     return pipe
 
-@spaces.GPU(duration=120)
+@spaces.GPU(duration=100)
 def generate_t2i(prompt, neg, lora_id, lora_scale, w, h):
     pipe = load_t2i(lora_id, lora_scale).to("cuda")
     img = pipe(prompt=prompt, negative_prompt=neg, num_inference_steps=30, 
-               guidance_scale=7.0, width=int(w), height=int(h)).images[0]
+               guidance_scale=7.0, width=int(w), height=int(h), 
+               generator=torch.Generator("cuda").manual_seed(42)).images[0]
     return img
 
 @spaces.GPU(duration=200)
@@ -68,7 +71,7 @@ def generate_video(prompt, init_image, lora_scale):
     from diffusers.utils import export_to_video
     pipe = load_video().to("cuda")
     kwargs = {"prompt": prompt, "negative_prompt": NEG_DEFAULT, "num_frames": 49, 
-              "num_inference_steps": 30}
+              "num_inference_steps": 30, "generator": torch.Generator("cuda").manual_seed(42)}
     if init_image is not None:
         kwargs["image"] = Image.fromarray(init_image).convert("RGB").resize((768, 512))
     if lora_scale > 0:
@@ -78,6 +81,7 @@ def generate_video(prompt, init_image, lora_scale):
     export_to_video(output.frames[0], tmp.name, fps=24)
     return tmp.name
 
+# UI
 with gr.Blocks(title="Image Utility v2.1") as demo:
     gr.HTML("<h1 style='text-align:center;'>🛠 Image Processing Utility v2.1.4</h1>")
     with gr.Tabs():