Upload 6 files

README.md

@@ -1,14 +1,16 @@
 ---
-title: Z-Image + Watermark Remover V3
+title: Z-Image + Watermark Remover V4
 emoji: 🎨
 colorFrom: yellow
 colorTo: purple
 sdk: gradio
+sdk_version: 5.23.1
+python_version: "3.10"
 app_file: app.py
-pinned: true
+startup_duration_timeout: 30m
 ---
 
-# Z-Image + Watermark Remover V3
+# Z-Image + Watermark Remover V4
 
 Combined ZeroGPU Space:
 
@@ -18,11 +20,22 @@ Combined ZeroGPU Space:
 - Video watermark removal
 - Gradio API
 
-## V3 fixes
+## V4 dependency fix
+
+Gradio is **not** listed in `requirements.txt`. Hugging Face Spaces installs and manages Gradio from the `sdk_version` field above.
+
+This avoids the dependency conflict between:
+
+- Gradio 6.x -> `huggingface-hub >= 1.16`
+- Transformers 4.56/4.57 -> `huggingface-hub < 1.0`
+
+V4 uses Gradio 5.23.1 through the Space SDK and keeps `huggingface-hub < 1.0` for the Transformers/Florence-2 stack.
+
+## V3 fixes retained
 
 ### Florence-2 / Transformers compatibility
 
-The runtime uses `transformers>=4.56.1,<5` and applies a compatibility patch before loading Florence-2.
+The runtime uses `transformers==4.57.6` and applies a compatibility patch before loading Florence-2.
 
 The patch handles the `forced_bos_token_id` access used by older Florence-2 remote code.
 
@@ -50,8 +63,12 @@ The original audio stream is mapped back into the output when present.
 - `remove_image`
 - `remove_video`
 
-## ZeroGPU
+## Deployment
 
-Select ZeroGPU as the Space hardware. GPU functions use `@spaces.GPU`.
+1. Create/select a Gradio Space.
+2. Select ZeroGPU hardware.
+3. Upload the files from this ZIP to the Space root.
+4. If the Space has an older cached environment, use Factory Reboot after committing.
+5. Optionally add `HF_TOKEN` as a Space Secret for authenticated Hub access.
 
-Add `HF_TOKEN` in Space Settings -> Variables and secrets if you want authenticated Hugging Face Hub access.
+Do not add `gradio`, `gradio-client`, `spaces`, or `huggingface-hub` to `requirements.txt`.

app.py

@@ -16,8 +16,11 @@ print("Z-Image pipeline loaded!")
 from watermark import preview, remove_image, process_video
 
 @spaces.GPU(duration=60)
-def generate_image(prompt, height, width, steps, seed, randomize_seed,
+def generate_image(prompt: str, height: int, width: int, steps: int,
+                   seed: int, randomize_seed: bool,
                    progress=gr.Progress(track_tqdm=True)):
+    """Generate an image with Z-Image-Turbo and return the image plus seed."""
+
     if randomize_seed:
         seed = torch.randint(0, 2**32 - 1, (1,)).item()
     seed = int(seed)
@@ -33,21 +36,30 @@ def generate_image(prompt, height, width, steps, seed, randomize_seed,
     return image, seed
 
 @spaces.GPU(duration=60)
-def wm_preview(image, detection_prompt="watermark", max_bbox_percent=30):
+def wm_preview(image, detection_prompt: str = "watermark",
+                max_bbox_percent: float = 30):
+    """Detect likely watermark regions and return a preview with boxes."""
+
     return preview(image, float(max_bbox_percent), detection_prompt or "watermark")
 
 @spaces.GPU(duration=90)
-def wm_remove_image(image, detection_prompt="watermark",
-                     max_bbox_percent=30, transparent=False):
+def wm_remove_image(image, detection_prompt: str = "watermark",
+                     max_bbox_percent: float = 30,
+                     transparent: bool = False):
+    """Detect and remove watermark regions from an image."""
+
     return remove_image(
         image, float(max_bbox_percent),
         detection_prompt or "watermark", bool(transparent)
     )
 
 @spaces.GPU(duration=300)
-def wm_remove_video(video_path, detection_prompt="watermark",
-                    max_bbox_percent=30, detection_skip=5,
-                    fade_in=2, fade_out=2, progress=gr.Progress()):
+def wm_remove_video(video_path: str, detection_prompt: str = "watermark",
+                    max_bbox_percent: float = 30, detection_skip: int = 5,
+                    fade_in: int = 2, fade_out: int = 2,
+                    progress=gr.Progress()):
+    """Detect and remove watermarks from a video while preserving audio."""
+
     if not video_path:
         raise gr.Error("Please upload a video.")
     return process_video(
@@ -145,4 +157,8 @@ with gr.Blocks(title="Z-Image + Watermark Remover V3") as demo:
             )
 
 if __name__ == "__main__":
-    demo.launch()
+    print("Starting Gradio Space...")
+    print(f"CUDA available: {torch.cuda.is_available()}")
+    if torch.cuda.is_available():
+        print(f"CUDA device: {torch.cuda.get_device_name(torch.cuda.current_device())}")
+    demo.launch(server_name="0.0.0.0", server_port=7860)

requirements.txt

@@ -1,12 +1,13 @@
-# ZeroGPU runtime provides gradio/spaces/huggingface_hub.
-# Do not pin those platform-managed packages.
+# Gradio, spaces, and huggingface_hub are platform-managed by Hugging Face Spaces.
+# Do NOT add them here.
 
 diffusers
-transformers>=4.56.1,<5
+transformers==4.57.6
 accelerate
 peft
 safetensors
 sentencepiece
+huggingface-hub<1.0
 
 opencv-python-headless>=4.8,<4.12
 numpy<2