22

README.md

@@ -1,7 +1,7 @@
 ---
 title: Z-Image Turbo (ZIT) - Inpaint
 emoji: 💃
-colorFrom: blue
+colorFrom: red
 colorTo: yellow
 sdk: gradio
 sdk_version: 6.1.0

app.py

@@ -16,6 +16,7 @@ repo_utils.move_folder("repos/VideoX-Fun/config", "config")
 from videox_fun.pipeline import ZImageControlPipeline
 from videox_fun.models import ZImageControlTransformer2DModel
 from videox_fun.utils.utils import get_image_latent
+from controlnet_aux.processor import Processor
 
 #clone models
 repo_utils.clone_repo_if_not_exists("https://huggingface.co/Tongyi-MAI/Z-Image-Turbo", "models")
@@ -122,13 +123,13 @@ def inference(
         return None
     
     # rescale to prevent OOM
-    inpaint_image = edit_dict['background']
-    inpaint_image, width, height = image_utils.rescale_image(inpaint_image, 1, 8, max_size=1280)
+    input_image = edit_dict['background']
+    input_image, width, height = image_utils.rescale_image(inpaint_image, 1, 8, max_size=1280)
     sample_size = [height, width]
     
     print("DEBUG: control_image_torch")
-    if inpaint_image is not None:
-        inpaint_image = get_image_latent(inpaint_image, sample_size=sample_size)[:, :, 0]
+    if input_image is not None:
+        inpaint_image = get_image_latent(input_image, sample_size=sample_size)[:, :, 0]
     else:
         inpaint_image = torch.zeros([1, 3, sample_size[0], sample_size[1]])
 
@@ -138,6 +139,14 @@ def inference(
     else:
         mask_image = torch.ones([1, 1, sample_size[0], sample_size[1]]) * 255
 
+    print("DEBUG: control_image_torch")
+    processor = Processor('canny')
+    control_image, w, h = image_utils.rescale_image(input_image, 1, 8, max_size=1280)
+    control_image = control_image.resize((1024, 1024))
+    control_image = processor(control_image, to_pil=True)
+    control_image = control_image.resize((width, height))
+    control_image_torch = get_image_latent(control_image, sample_size=sample_size)[:, :, 0]
+
     # generation
     if randomize_seed: seed = random.randint(0, MAX_SEED)
     generator = torch.Generator().manual_seed(seed)
@@ -151,7 +160,7 @@ def inference(
         guidance_scale=guidance_scale,
         image               = inpaint_image,
         mask_image          = mask_image,
-        # control_image=control_image_torch,
+        control_image=control_image_torch,
         num_inference_steps=num_inference_steps,
         control_context_scale=control_context_scale,
     ).images[0]

static/header.html

@@ -4,9 +4,9 @@
     </h1>
     <div class="grid-container" >
         <p>
-            This space still in development, there will be some errors. Please be patient and check again later 🙏
-            <br>
-            Edit and guide image generation while preserving the original image quality using controlnet (pose maps, scribbling maps, and other features). 
+            <!-- This space still in development, there will be some errors. Please be patient and check again later 🙏
+            <br> -->
+            Transform your pictures by simply applying a mask and providing a text prompt.
             <br>
             If you like my work, please support me by visiting <a href="https://aisudo.com/" target="_blank">AiSudo</a> 😊
     </div>