Fix issues about ZeroGPU and examples (#2)

- Delete .DS_Store and __pycache__ (c5e30ab8e6a0071c198b2b6a7cfe16b45de6c673)
- Add .gitignore (ecda16070478ee8ba8913bba520fc1d9fb4c80c1)
- Apply formatter to app.py and requirements.txt (33cb6e3e0847e54a8fce301fb39b6340beb34b1e)
- Clean up (5719c29c84ab2296306a92e39b5cc38f57a5bdb9)
- Remove unused import (c37294a743eb4be148f3d8d65c30d423efefc24c)
- Use huggingface_hub to download models (634953839cbc8dd333a76dd302f0c241b9c2f491)
- format (1109e54a30722d3a0bec94c8a4e98941b478555f)
- Change how gr.State is used (49f5f360f59a322ca015e5189a43a2b665a8a112)
- Add error handling (de8155f9965bbe0a4e2d70c508ca29a01d802a6d)
- Add error handling (0e7235bc177bb3a74a4f22a2b2a90bc6dbcfb781)
- Add error handling (5336236acbd5cf22c374422ddbddf6458842781d)
- Process examples when loaded (661a9c1678c0d5bc9e0f952e541d4b83951d1a2e)


Co-authored-by: hysts <[email protected]>

.gitignore

@@ -0,0 +1,162 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/latest/usage/project/#working-with-version-control
+.pdm.toml
+.pdm-python
+.pdm-build/
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/

app.py

@@ -1,35 +1,35 @@
+import json
 import os
-import sys
-
-
-print("Installing correct gradio version...")
-os.system("pip uninstall -y gradio")
-os.system("pip install gradio==4.38.1")
-print("Installing Finished!")
-
+import uuid
 
+import cv2
 import gradio as gr
 import numpy as np
-import cv2
-import uuid
+import spaces
 import torch
 import torchvision
-import json
-import spaces
-
-from PIL import Image
+from diffusers import AutoencoderKL, DDIMScheduler
+from einops import rearrange
+from huggingface_hub import hf_hub_download
 from omegaconf import OmegaConf
-from einops import rearrange, repeat
-from torchvision import transforms,utils
+from PIL import Image
+from torchvision import transforms
 from transformers import CLIPTextModel, CLIPTokenizer
-from diffusers import AutoencoderKL, DDIMScheduler
 
-from pipelines.pipeline_imagecoductor import ImageConductorPipeline
 from modules.unet import UNet3DConditionFlowModel
-from utils.gradio_utils import ensure_dirname, split_filename, visualize_drag, image2pil
-from utils.utils import create_image_controlnet, create_flow_controlnet, interpolate_trajectory, load_weights, load_model, bivariate_Gaussian, save_videos_grid
+from pipelines.pipeline_imagecoductor import ImageConductorPipeline
+from utils.gradio_utils import ensure_dirname, image2pil, split_filename, visualize_drag
 from utils.lora_utils import add_LoRA_to_controlnet
-from utils.visualizer import Visualizer, vis_flow_to_video
+from utils.utils import (
+    bivariate_Gaussian,
+    create_flow_controlnet,
+    create_image_controlnet,
+    interpolate_trajectory,
+    load_model,
+    load_weights,
+)
+from utils.visualizer import vis_flow_to_video
+
 #### Description ####
 title = r"""<h1 align="center">CustomNet: Object Customization with Variable-Viewpoints in Text-to-Image Diffusion Models</h1>"""
 
@@ -41,7 +41,7 @@ head = r"""
                             <a href='https://liyaowei-stu.github.io/project/ImageConductor/'><img src='https://img.shields.io/badge/Project_Page-ImgaeConductor-green' alt='Project Page'></a>
                             <a href='https://arxiv.org/pdf/2406.15339'><img src='https://img.shields.io/badge/Paper-Arxiv-blue'></a>
                             <a href='https://github.com/liyaowei-stu/ImageConductor'><img src='https://img.shields.io/badge/Code-Github-orange'></a>
-                            
+
 
                         </div>
                         </br>
@@ -49,7 +49,6 @@ head = r"""
 """
 
 
-
 descriptions = r"""
 Official Gradio Demo for <a href='https://github.com/liyaowei-stu/ImageConductor'><b>Image Conductor: Precision Control for Interactive Video Synthesis</b></a>.<br>
 🧙Image Conductor enables precise, fine-grained control for generating motion-controllable videos from images, advancing the practical application of interactive video synthesis.<br>
@@ -66,7 +65,7 @@ instructions = r"""
             """
 
 citation = r"""
-If Image Conductor is helpful, please help to ⭐ the <a href='https://github.com/liyaowei-stu/ImageConductor' target='_blank'>Github Repo</a>. Thanks! 
+If Image Conductor is helpful, please help to ⭐ the <a href='https://github.com/liyaowei-stu/ImageConductor' target='_blank'>Github Repo</a>. Thanks!
 [![GitHub Stars](https://img.shields.io/github/stars/liyaowei-stu%2FImageConductor)](https://github.com/liyaowei-stu/ImageConductor)
 ---
 
@@ -75,7 +74,7 @@ If Image Conductor is helpful, please help to ⭐ the <a href='https://github.co
 If our work is useful for your research, please consider citing:
 ```bibtex
 @misc{li2024imageconductor,
-    title={Image Conductor: Precision Control for Interactive Video Synthesis}, 
+    title={Image Conductor: Precision Control for Interactive Video Synthesis},
     author={Li, Yaowei and Wang, Xintao and Zhang, Zhaoyang and Wang, Zhouxia and Yuan, Ziyang and Xie, Liangbin and Zou, Yuexian and Shan, Ying},
     year={2024},
     eprint={2406.15339},
@@ -90,46 +89,19 @@ If you have any questions, please feel free to reach me out at <b>[email protected]
 
 # """
 
-os.makedirs("models/personalized")
-os.makedirs("models/sd1-5")
-
-if not os.path.exists("models/flow_controlnet.ckpt"):
-    os.system(f'wget -q https://huggingface.co/TencentARC/ImageConductor/resolve/main/flow_controlnet.ckpt?download=true -P models/')
-    os.system(f'mv models/flow_controlnet.ckpt?download=true models/flow_controlnet.ckpt')
-    print("flow_controlnet Download!", )
-
-if not os.path.exists("models/image_controlnet.ckpt"):
-    os.system(f'wget -q https://huggingface.co/TencentARC/ImageConductor/resolve/main/image_controlnet.ckpt?download=true -P models/')
-    os.system(f'mv models/image_controlnet.ckpt?download=true models/image_controlnet.ckpt')
-    print("image_controlnet Download!", )
+flow_controlnet_path = hf_hub_download("TencentARC/ImageConductor", "flow_controlnet.ckpt")
+image_controlnet_path = hf_hub_download("TencentARC/ImageConductor", "image_controlnet.ckpt")
+unet_path = hf_hub_download("TencentARC/ImageConductor", "unet.ckpt")
 
-if not os.path.exists("models/unet.ckpt"):
-    os.system(f'wget -q https://huggingface.co/TencentARC/ImageConductor/resolve/main/unet.ckpt?download=true -P models/')
-    os.system(f'mv models/unet.ckpt?download=true models/unet.ckpt')
-    print("unet Download!", )
+helloobjects_path = hf_hub_download("TencentARC/ImageConductor", "helloobjects_V12c.safetensors")
+tusun_path = hf_hub_download("TencentARC/ImageConductor", "TUSUN.safetensors")
 
-    
+os.makedirs("models/sd1-5", exist_ok=True)
+sd15_config_path = hf_hub_download("runwayml/stable-diffusion-v1-5", "config.json", subfolder="unet")
 if not os.path.exists("models/sd1-5/config.json"):
-    os.system(f'wget -q https://huggingface.co/runwayml/stable-diffusion-v1-5/resolve/main/unet/config.json?download=true -P models/sd1-5/')
-    os.system(f'mv models/sd1-5/config.json?download=true  models/sd1-5/config.json')
-    print("config Download!", )
-
-
+    os.symlink(sd15_config_path, "models/sd1-5/config.json")
 if not os.path.exists("models/sd1-5/unet.ckpt"):
-    os.system(f'cp -r models/unet.ckpt  models/sd1-5/unet.ckpt')
-
-# os.system(f'wget https://huggingface.co/runwayml/stable-diffusion-v1-5/resolve/main/unet/diffusion_pytorch_model.bin?download=true -P models/sd1-5/')
-
-if not os.path.exists("models/personalized/helloobjects_V12c.safetensors"):
-    os.system(f'wget -q https://huggingface.co/TencentARC/ImageConductor/resolve/main/helloobjects_V12c.safetensors?download=true -P models/personalized')
-    os.system(f'mv models/personalized/helloobjects_V12c.safetensors?download=true models/personalized/helloobjects_V12c.safetensors')
-    print("helloobjects_V12c Download!", )
-
-
-if not os.path.exists("models/personalized/TUSUN.safetensors"):
-    os.system(f'wget -q https://huggingface.co/TencentARC/ImageConductor/resolve/main/TUSUN.safetensors?download=true -P models/personalized')
-    os.system(f'mv models/personalized/TUSUN.safetensors?download=true models/personalized/TUSUN.safetensors')
-    print("TUSUN Download!", )
+    os.symlink(unet_path, "models/sd1-5/unet.ckpt")
 
 # mv1 = os.system(f'mv /usr/local/lib/python3.10/site-packages/gradio/helpers.py /usr/local/lib/python3.10/site-packages/gradio/helpers_bkp.py')
 # mv2 = os.system(f'mv helpers.py /usr/local/lib/python3.10/site-packages/gradio/helpers.py')
@@ -145,128 +117,135 @@ if not os.path.exists("models/personalized/TUSUN.safetensors"):
 # - - - - - examples  - - - - -  #
 
 image_examples = [
-    ["__asset__/images/object/turtle-1.jpg", 
-     "a sea turtle gracefully swimming over a coral reef in the clear blue ocean.", 
-     "object",
-     11318446767408804497,
-     "",
-     "turtle",
-     "__asset__/turtle.mp4"
-     ],
-    
-    ["__asset__/images/object/rose-1.jpg", 
-     "a red rose engulfed in flames.", 
-     "object",
-     6854275249656120509,
-     "",
-     "rose",
-     "__asset__/rose.mp4"
-     ],
-    
-    ["__asset__/images/object/jellyfish-1.jpg", 
-     "intricate detailing,photorealism,hyperrealistic, glowing jellyfish mushroom, flying, starry sky, bokeh, golden ratio composition.", 
-     "object",
-     17966188172968903484,
-     "HelloObject",
-     "jellyfish",
-     "__asset__/jellyfish.mp4"
-     ],
-    
-    
-    ["__asset__/images/camera/lush-1.jpg", 
-     "detailed craftsmanship, photorealism, hyperrealistic, roaring waterfall, misty spray, lush greenery, vibrant rainbow, golden ratio composition.", 
-     "camera",
-     7970487946960948963,
-     "HelloObject",
-     "lush",
-     "__asset__/lush.mp4",
-     ],
-    
-    ["__asset__/images/camera/tusun-1.jpg", 
-     "tusuncub with its mouth open, blurry, open mouth, fangs, photo background, looking at viewer, tongue, full body, solo, cute and lovely, Beautiful and realistic eye details, perfect anatomy, Nonsense, pure background, Centered-Shot, realistic photo, photograph, 4k, hyper detailed, DSLR, 24 Megapixels, 8mm Lens, Full Frame, film grain, Global Illumination, studio Lighting, Award Winning Photography, diffuse reflection, ray tracing.", 
-     "camera",
-     996953226890228361,
-     "TUSUN",
-     "tusun",
-     "__asset__/tusun.mp4"
-     ],
-    
-    ["__asset__/images/camera/painting-1.jpg", 
-     "A oil painting.", 
-     "camera",
-     16867854766769816385,
-     "",
-     "painting",
-     "__asset__/painting.mp4"
-     ],
+    [
+        "__asset__/images/object/turtle-1.jpg",
+        "a sea turtle gracefully swimming over a coral reef in the clear blue ocean.",
+        "object",
+        11318446767408804497,
+        "",
+        "turtle",
+        "__asset__/turtle.mp4",
+    ],
+    [
+        "__asset__/images/object/rose-1.jpg",
+        "a red rose engulfed in flames.",
+        "object",
+        6854275249656120509,
+        "",
+        "rose",
+        "__asset__/rose.mp4",
+    ],
+    [
+        "__asset__/images/object/jellyfish-1.jpg",
+        "intricate detailing,photorealism,hyperrealistic, glowing jellyfish mushroom, flying, starry sky, bokeh, golden ratio composition.",
+        "object",
+        17966188172968903484,
+        "HelloObject",
+        "jellyfish",
+        "__asset__/jellyfish.mp4",
+    ],
+    [
+        "__asset__/images/camera/lush-1.jpg",
+        "detailed craftsmanship, photorealism, hyperrealistic, roaring waterfall, misty spray, lush greenery, vibrant rainbow, golden ratio composition.",
+        "camera",
+        7970487946960948963,
+        "HelloObject",
+        "lush",
+        "__asset__/lush.mp4",
+    ],
+    [
+        "__asset__/images/camera/tusun-1.jpg",
+        "tusuncub with its mouth open, blurry, open mouth, fangs, photo background, looking at viewer, tongue, full body, solo, cute and lovely, Beautiful and realistic eye details, perfect anatomy, Nonsense, pure background, Centered-Shot, realistic photo, photograph, 4k, hyper detailed, DSLR, 24 Megapixels, 8mm Lens, Full Frame, film grain, Global Illumination, studio Lighting, Award Winning Photography, diffuse reflection, ray tracing.",
+        "camera",
+        996953226890228361,
+        "TUSUN",
+        "tusun",
+        "__asset__/tusun.mp4",
+    ],
+    [
+        "__asset__/images/camera/painting-1.jpg",
+        "A oil painting.",
+        "camera",
+        16867854766769816385,
+        "",
+        "painting",
+        "__asset__/painting.mp4",
+    ],
 ]
 
 
 POINTS = {
-    'turtle': "__asset__/trajs/object/turtle-1.json",
-    'rose': "__asset__/trajs/object/rose-1.json",
-    'jellyfish': "__asset__/trajs/object/jellyfish-1.json",
-    'lush': "__asset__/trajs/camera/lush-1.json",
-    'tusun': "__asset__/trajs/camera/tusun-1.json",
-    'painting': "__asset__/trajs/camera/painting-1.json",
+    "turtle": "__asset__/trajs/object/turtle-1.json",
+    "rose": "__asset__/trajs/object/rose-1.json",
+    "jellyfish": "__asset__/trajs/object/jellyfish-1.json",
+    "lush": "__asset__/trajs/camera/lush-1.json",
+    "tusun": "__asset__/trajs/camera/tusun-1.json",
+    "painting": "__asset__/trajs/camera/painting-1.json",
 }
 
 IMAGE_PATH = {
-    'turtle': "__asset__/images/object/turtle-1.jpg",
-    'rose': "__asset__/images/object/rose-1.jpg",
-    'jellyfish': "__asset__/images/object/jellyfish-1.jpg",
-    'lush': "__asset__/images/camera/lush-1.jpg",
-    'tusun': "__asset__/images/camera/tusun-1.jpg",
-    'painting': "__asset__/images/camera/painting-1.jpg",
+    "turtle": "__asset__/images/object/turtle-1.jpg",
+    "rose": "__asset__/images/object/rose-1.jpg",
+    "jellyfish": "__asset__/images/object/jellyfish-1.jpg",
+    "lush": "__asset__/images/camera/lush-1.jpg",
+    "tusun": "__asset__/images/camera/tusun-1.jpg",
+    "painting": "__asset__/images/camera/painting-1.jpg",
 }
 
 
-
 DREAM_BOOTH = {
-    'HelloObject': 'models/personalized/helloobjects_V12c.safetensors',
+    "HelloObject": helloobjects_path,
 }
 
 LORA = {
-    'TUSUN': 'models/personalized/TUSUN.safetensors',
+    "TUSUN": tusun_path,
 }
 
 LORA_ALPHA = {
-    'TUSUN': 0.6,
+    "TUSUN": 0.6,
 }
 
 NPROMPT = {
-    "HelloObject": 'FastNegativeV2,(bad-artist:1),(worst quality, low quality:1.4),(bad_prompt_version2:0.8),bad-hands-5,lowres,bad anatomy,bad hands,((text)),(watermark),error,missing fingers,extra digit,fewer digits,cropped,worst quality,low quality,normal quality,((username)),blurry,(extra limbs),bad-artist-anime,badhandv4,EasyNegative,ng_deepnegative_v1_75t,verybadimagenegative_v1.3,BadDream,(three hands:1.6),(three legs:1.2),(more than two hands:1.4),(more than two legs,:1.2)'    
+    "HelloObject": "FastNegativeV2,(bad-artist:1),(worst quality, low quality:1.4),(bad_prompt_version2:0.8),bad-hands-5,lowres,bad anatomy,bad hands,((text)),(watermark),error,missing fingers,extra digit,fewer digits,cropped,worst quality,low quality,normal quality,((username)),blurry,(extra limbs),bad-artist-anime,badhandv4,EasyNegative,ng_deepnegative_v1_75t,verybadimagenegative_v1.3,BadDream,(three hands:1.6),(three legs:1.2),(more than two hands:1.4),(more than two legs,:1.2)"
 }
 
 output_dir = "outputs"
 ensure_dirname(output_dir)
 
+
 def points_to_flows(track_points, model_length, height, width):
     input_drag = np.zeros((model_length - 1, height, width, 2))
     for splited_track in track_points:
-        if len(splited_track) == 1: # stationary point
+        if len(splited_track) == 1:  # stationary point
             displacement_point = tuple([splited_track[0][0] + 1, splited_track[0][1] + 1])
             splited_track = tuple([splited_track[0], displacement_point])
         # interpolate the track
         splited_track = interpolate_trajectory(splited_track, model_length)
         splited_track = splited_track[:model_length]
         if len(splited_track) < model_length:
-            splited_track = splited_track + [splited_track[-1]] * (model_length -len(splited_track))
+            splited_track = splited_track + [splited_track[-1]] * (model_length - len(splited_track))
         for i in range(model_length - 1):
             start_point = splited_track[i]
-            end_point = splited_track[i+1]
+            end_point = splited_track[i + 1]
             input_drag[i][int(start_point[1])][int(start_point[0])][0] = end_point[0] - start_point[0]
             input_drag[i][int(start_point[1])][int(start_point[0])][1] = end_point[1] - start_point[1]
     return input_drag
 
+
 class ImageConductor:
-    def __init__(self, device, unet_path, image_controlnet_path, flow_controlnet_path, height, width, model_length, lora_rank=64):
+    def __init__(
+        self, device, unet_path, image_controlnet_path, flow_controlnet_path, height, width, model_length, lora_rank=64
+    ):
         self.device = device
-        tokenizer    = CLIPTokenizer.from_pretrained("runwayml/stable-diffusion-v1-5", subfolder="tokenizer")
-        text_encoder = CLIPTextModel.from_pretrained("runwayml/stable-diffusion-v1-5", subfolder="text_encoder").to(device)
-        vae          = AutoencoderKL.from_pretrained("runwayml/stable-diffusion-v1-5", subfolder="vae").to(device)
+        tokenizer = CLIPTokenizer.from_pretrained("runwayml/stable-diffusion-v1-5", subfolder="tokenizer")
+        text_encoder = CLIPTextModel.from_pretrained("runwayml/stable-diffusion-v1-5", subfolder="text_encoder").to(
+            device
+        )
+        vae = AutoencoderKL.from_pretrained("runwayml/stable-diffusion-v1-5", subfolder="vae").to(device)
         inference_config = OmegaConf.load("configs/inference/inference.yaml")
-        unet = UNet3DConditionFlowModel.from_pretrained_2d("models/sd1-5/", unet_additional_kwargs=OmegaConf.to_container(inference_config.unet_additional_kwargs))
+        unet = UNet3DConditionFlowModel.from_pretrained_2d(
+            "models/sd1-5/", unet_additional_kwargs=OmegaConf.to_container(inference_config.unet_additional_kwargs)
+        )
 
         self.vae = vae
 
@@ -287,15 +266,14 @@ class ImageConductor:
 
         self.pipeline = ImageConductorPipeline(
             unet=unet,
-            vae=vae, 
-            tokenizer=tokenizer, 
-            text_encoder=text_encoder, 
+            vae=vae,
+            tokenizer=tokenizer,
+            text_encoder=text_encoder,
             scheduler=DDIMScheduler(**OmegaConf.to_container(inference_config.noise_scheduler_kwargs)),
             image_controlnet=image_controlnet,
             flow_controlnet=flow_controlnet,
         ).to(device)
 
-        
         self.height = height
         self.width = width
         # _, model_step, _ = split_filename(model_path)
@@ -307,40 +285,51 @@ class ImageConductor:
         self.blur_kernel = blur_kernel
 
     @spaces.GPU(duration=180)
-    def run(self, first_frame_path, tracking_points, prompt, drag_mode, negative_prompt, seed, randomize_seed, guidance_scale, num_inference_steps, personalized, examples_type):
+    def run(
+        self,
+        first_frame_path,
+        tracking_points,
+        prompt,
+        drag_mode,
+        negative_prompt,
+        seed,
+        randomize_seed,
+        guidance_scale,
+        num_inference_steps,
+        personalized,
+    ):
         print("Run!")
-        if examples_type != "":
-            ### for adapting high version gradio
-            tracking_points = gr.State([])
-            first_frame_path = IMAGE_PATH[examples_type]
-            points = json.load(open(POINTS[examples_type]))
-            tracking_points.value.extend(points)
-            print("example first_frame_path", first_frame_path)
-            print("example tracking_points", tracking_points.value)
-            
-        original_width, original_height=384, 256
-        if isinstance(tracking_points, list):
-            input_all_points = tracking_points
-        else:
-            input_all_points = tracking_points.value
-        
+
+        original_width, original_height = 384, 256
+        input_all_points = tracking_points
+
         print("input_all_points", input_all_points)
-        resized_all_points = [tuple([tuple([float(e1[0]*self.width/original_width), float(e1[1]*self.height/original_height)]) for e1 in e]) for e in input_all_points]
+        resized_all_points = [
+            tuple(
+                [
+                    tuple([float(e1[0] * self.width / original_width), float(e1[1] * self.height / original_height)])
+                    for e1 in e
+                ]
+            )
+            for e in input_all_points
+        ]
 
         dir, base, ext = split_filename(first_frame_path)
-        id = base.split('_')[-1]
-        
-        
-        visualized_drag, _ = visualize_drag(first_frame_path, resized_all_points, drag_mode, self.width, self.height, self.model_length)
+        id = base.split("_")[-1]
+
+        visualized_drag, _ = visualize_drag(
+            first_frame_path, resized_all_points, drag_mode, self.width, self.height, self.model_length
+        )
 
-        ## image condition        
-        image_transforms = transforms.Compose([
+        ## image condition
+        image_transforms = transforms.Compose(
+            [
                 transforms.RandomResizedCrop(
-                    (self.height, self.width), (1.0, 1.0), 
-                    ratio=(self.width/self.height, self.width/self.height)
+                    (self.height, self.width), (1.0, 1.0), ratio=(self.width / self.height, self.width / self.height)
                 ),
                 transforms.ToTensor(),
-            ])
+            ]
+        )
 
         image_paths = [first_frame_path]
         controlnet_images = [(image_transforms(Image.open(path).convert("RGB"))) for path in image_paths]
@@ -349,205 +338,296 @@ class ImageConductor:
         num_controlnet_images = controlnet_images.shape[2]
         controlnet_images = rearrange(controlnet_images, "b c f h w -> (b f) c h w")
         self.vae.to(device)
-        controlnet_images = self.vae.encode(controlnet_images * 2. - 1.).latent_dist.sample() * 0.18215
+        controlnet_images = self.vae.encode(controlnet_images * 2.0 - 1.0).latent_dist.sample() * 0.18215
         controlnet_images = rearrange(controlnet_images, "(b f) c h w -> b c f h w", f=num_controlnet_images)
 
         # flow condition
         controlnet_flows = points_to_flows(resized_all_points, self.model_length, self.height, self.width)
-        for i in range(0, self.model_length-1):
+        for i in range(0, self.model_length - 1):
             controlnet_flows[i] = cv2.filter2D(controlnet_flows[i], -1, self.blur_kernel)
-        controlnet_flows = np.concatenate([np.zeros_like(controlnet_flows[0])[np.newaxis, ...], controlnet_flows], axis=0)  # pad the first frame with zero flow
+        controlnet_flows = np.concatenate(
+            [np.zeros_like(controlnet_flows[0])[np.newaxis, ...], controlnet_flows], axis=0
+        )  # pad the first frame with zero flow
         os.makedirs(os.path.join(output_dir, "control_flows"), exist_ok=True)
-        trajs_video = vis_flow_to_video(controlnet_flows, num_frames=self.model_length) # T-1 x H x W x 3
-        torchvision.io.write_video(f'{output_dir}/control_flows/sample-{id}-train_flow.mp4', trajs_video, fps=8, video_codec='h264', options={'crf': '10'})
-        controlnet_flows = torch.from_numpy(controlnet_flows)[None][:, :self.model_length, ...]
-        controlnet_flows =  rearrange(controlnet_flows, "b f h w c-> b c f h w").float().to(device)
+        trajs_video = vis_flow_to_video(controlnet_flows, num_frames=self.model_length)  # T-1 x H x W x 3
+        torchvision.io.write_video(
+            f"{output_dir}/control_flows/sample-{id}-train_flow.mp4",
+            trajs_video,
+            fps=8,
+            video_codec="h264",
+            options={"crf": "10"},
+        )
+        controlnet_flows = torch.from_numpy(controlnet_flows)[None][:, : self.model_length, ...]
+        controlnet_flows = rearrange(controlnet_flows, "b f h w c-> b c f h w").float().to(device)
 
-        dreambooth_model_path = DREAM_BOOTH.get(personalized, '')
-        lora_model_path = LORA.get(personalized, '')
+        dreambooth_model_path = DREAM_BOOTH.get(personalized, "")
+        lora_model_path = LORA.get(personalized, "")
         lora_alpha = LORA_ALPHA.get(personalized, 0.6)
         self.pipeline = load_weights(
             self.pipeline,
-            dreambooth_model_path      = dreambooth_model_path,
-            lora_model_path            = lora_model_path,
-            lora_alpha                 = lora_alpha,
+            dreambooth_model_path=dreambooth_model_path,
+            lora_model_path=lora_model_path,
+            lora_alpha=lora_alpha,
         ).to(device)
-        
-        if NPROMPT.get(personalized, '') != '':
-            negative_prompt =  NPROMPT.get(personalized)
-        
+
+        if NPROMPT.get(personalized, "") != "":
+            negative_prompt = NPROMPT.get(personalized)
+
         if randomize_seed:
             random_seed = torch.seed()
         else:
             seed = int(seed)
             random_seed = seed
         torch.manual_seed(random_seed)
-        torch.cuda.manual_seed_all(random_seed)  
+        torch.cuda.manual_seed_all(random_seed)
         print(f"current seed: {torch.initial_seed()}")
         sample = self.pipeline(
-                    prompt,
-                    negative_prompt     = negative_prompt,
-                    num_inference_steps = num_inference_steps,
-                    guidance_scale      = guidance_scale,
-                    width               = self.width,
-                    height              = self.height,
-                    video_length        = self.model_length,
-                    controlnet_images = controlnet_images, # 1 4 1 32 48
-                    controlnet_image_index = [0], 
-                    controlnet_flows  = controlnet_flows,# [1, 2, 16, 256, 384]
-                    control_mode = drag_mode,
-                    eval_mode = True,
-                ).videos
-
-        outputs_path = os.path.join(output_dir, f'output_{i}_{id}.mp4')
-        vis_video = (rearrange(sample[0], 'c t h w -> t h w c') * 255.).clip(0, 255)
-        torchvision.io.write_video(outputs_path, vis_video, fps=8, video_codec='h264', options={'crf': '10'})
-        
+            prompt,
+            negative_prompt=negative_prompt,
+            num_inference_steps=num_inference_steps,
+            guidance_scale=guidance_scale,
+            width=self.width,
+            height=self.height,
+            video_length=self.model_length,
+            controlnet_images=controlnet_images,  # 1 4 1 32 48
+            controlnet_image_index=[0],
+            controlnet_flows=controlnet_flows,  # [1, 2, 16, 256, 384]
+            control_mode=drag_mode,
+            eval_mode=True,
+        ).videos
+
+        outputs_path = os.path.join(output_dir, f"output_{i}_{id}.mp4")
+        vis_video = (rearrange(sample[0], "c t h w -> t h w c") * 255.0).clip(0, 255)
+        torchvision.io.write_video(outputs_path, vis_video, fps=8, video_codec="h264", options={"crf": "10"})
+
         # outputs_path = os.path.join(output_dir, f'output_{i}_{id}.gif')
         # save_videos_grid(sample[0][None], outputs_path)
         print("Done!")
-        return {output_image: visualized_drag, output_video: outputs_path}
+        return visualized_drag, outputs_path
 
 
 def reset_states(first_frame_path, tracking_points):
-    first_frame_path = gr.State()
-    tracking_points = gr.State([])
-    return {input_image:None, first_frame_path_var: first_frame_path, tracking_points_var: tracking_points}
+    first_frame_path = None
+    tracking_points = []
+    return {input_image: None, first_frame_path_var: first_frame_path, tracking_points_var: tracking_points}
 
 
 def preprocess_image(image, tracking_points):
     image_pil = image2pil(image.name)
     raw_w, raw_h = image_pil.size
-    resize_ratio = max(384/raw_w, 256/raw_h)
+    resize_ratio = max(384 / raw_w, 256 / raw_h)
     image_pil = image_pil.resize((int(raw_w * resize_ratio), int(raw_h * resize_ratio)), Image.BILINEAR)
-    image_pil = transforms.CenterCrop((256, 384))(image_pil.convert('RGB'))
+    image_pil = transforms.CenterCrop((256, 384))(image_pil.convert("RGB"))
     id = str(uuid.uuid4())[:4]
     first_frame_path = os.path.join(output_dir, f"first_frame_{id}.jpg")
     image_pil.save(first_frame_path, quality=95)
-    tracking_points =  gr.State([])
-    return {input_image: first_frame_path, first_frame_path_var: first_frame_path, tracking_points_var: tracking_points, personalized:""}
+    tracking_points = []
+    return {
+        input_image: first_frame_path,
+        first_frame_path_var: first_frame_path,
+        tracking_points_var: tracking_points,
+        personalized: "",
+    }
+
+
+def add_tracking_points(
+    tracking_points, first_frame_path, drag_mode, evt: gr.SelectData
+):  # SelectData is a subclass of EventData
+    if drag_mode == "object":
+        color = (255, 0, 0, 255)
+    elif drag_mode == "camera":
+        color = (0, 0, 255, 255)
+
+    print(f"You selected {evt.value} at {evt.index} from {evt.target}")
+    if not tracking_points:
+        tracking_points = [[]]
+    tracking_points[-1].append(evt.index)
+
+    transparent_background = Image.open(first_frame_path).convert("RGBA")
+    w, h = transparent_background.size
+    transparent_layer = np.zeros((h, w, 4))
+
+    for track in tracking_points:
+        if len(track) > 1:
+            for i in range(len(track) - 1):
+                start_point = track[i]
+                end_point = track[i + 1]
+                vx = end_point[0] - start_point[0]
+                vy = end_point[1] - start_point[1]
+                arrow_length = np.sqrt(vx**2 + vy**2)
+                if i == len(track) - 2:
+                    cv2.arrowedLine(
+                        transparent_layer, tuple(start_point), tuple(end_point), color, 2, tipLength=8 / arrow_length
+                    )
+                else:
+                    cv2.line(
+                        transparent_layer,
+                        tuple(start_point),
+                        tuple(end_point),
+                        color,
+                        2,
+                    )
+        else:
+            cv2.circle(transparent_layer, tuple(track[0]), 5, color, -1)
+
+    transparent_layer = Image.fromarray(transparent_layer.astype(np.uint8))
+    trajectory_map = Image.alpha_composite(transparent_background, transparent_layer)
+    return {tracking_points_var: tracking_points, input_image: trajectory_map}
+
 
+def preprocess_example_image(image_path, tracking_points, drag_mode):
+    image_pil = image2pil(image_path)
+    raw_w, raw_h = image_pil.size
+    resize_ratio = max(384 / raw_w, 256 / raw_h)
+    image_pil = image_pil.resize((int(raw_w * resize_ratio), int(raw_h * resize_ratio)), Image.BILINEAR)
+    image_pil = transforms.CenterCrop((256, 384))(image_pil.convert("RGB"))
+    id = str(uuid.uuid4())[:4]
+    first_frame_path = os.path.join(output_dir, f"first_frame_{id}.jpg")
+    image_pil.save(first_frame_path, quality=95)
 
-def add_tracking_points(tracking_points, first_frame_path, drag_mode, evt: gr.SelectData):  # SelectData is a subclass of EventData
-    if drag_mode=='object':
+    if drag_mode == "object":
         color = (255, 0, 0, 255)
-    elif drag_mode=='camera':
+    elif drag_mode == "camera":
         color = (0, 0, 255, 255)
 
-    if not isinstance(tracking_points ,list):
-        print(f"You selected {evt.value} at {evt.index} from {evt.target}")
-        tracking_points.value[-1].append(evt.index)
-        print(tracking_points.value)
-        tracking_points_values =  tracking_points.value
-    else:
-        try:
-            tracking_points[-1].append(evt.index)
-        except Exception as e:
-            tracking_points.append([])
-            tracking_points[-1].append(evt.index)
-            print(f"Solved Error: {e}")
-        
-        tracking_points_values = tracking_points
-        
-    
-    transparent_background = Image.open(first_frame_path).convert('RGBA')
+    transparent_background = Image.open(first_frame_path).convert("RGBA")
     w, h = transparent_background.size
     transparent_layer = np.zeros((h, w, 4))
-    
-    for track in tracking_points_values:
+
+    for track in tracking_points:
         if len(track) > 1:
-            for i in range(len(track)-1):
+            for i in range(len(track) - 1):
                 start_point = track[i]
-                end_point = track[i+1]
+                end_point = track[i + 1]
                 vx = end_point[0] - start_point[0]
                 vy = end_point[1] - start_point[1]
                 arrow_length = np.sqrt(vx**2 + vy**2)
-                if i == len(track)-2:
-                    cv2.arrowedLine(transparent_layer, tuple(start_point), tuple(end_point), color, 2, tipLength=8 / arrow_length)
+                if i == len(track) - 2:
+                    cv2.arrowedLine(
+                        transparent_layer, tuple(start_point), tuple(end_point), color, 2, tipLength=8 / arrow_length
+                    )
                 else:
-                    cv2.line(transparent_layer, tuple(start_point), tuple(end_point), color, 2,)
+                    cv2.line(
+                        transparent_layer,
+                        tuple(start_point),
+                        tuple(end_point),
+                        color,
+                        2,
+                    )
         else:
             cv2.circle(transparent_layer, tuple(track[0]), 5, color, -1)
 
     transparent_layer = Image.fromarray(transparent_layer.astype(np.uint8))
     trajectory_map = Image.alpha_composite(transparent_background, transparent_layer)
-    return {tracking_points_var: tracking_points, input_image: trajectory_map}
+
+    return trajectory_map, first_frame_path
 
 
 def add_drag(tracking_points):
-    if not isinstance(tracking_points ,list):
-        # print("before", tracking_points.value)
-        tracking_points.value.append([])
-        # print(tracking_points.value)
-    else:
+    if not tracking_points or tracking_points[-1]:
         tracking_points.append([])
     return {tracking_points_var: tracking_points}
-    
+
 
 def delete_last_drag(tracking_points, first_frame_path, drag_mode):
-    if drag_mode=='object':
+    if drag_mode == "object":
         color = (255, 0, 0, 255)
-    elif drag_mode=='camera':
+    elif drag_mode == "camera":
         color = (0, 0, 255, 255)
-    tracking_points.value.pop()
-    transparent_background = Image.open(first_frame_path).convert('RGBA')
+    if tracking_points:
+        tracking_points.pop()
+    transparent_background = Image.open(first_frame_path).convert("RGBA")
     w, h = transparent_background.size
     transparent_layer = np.zeros((h, w, 4))
-    for track in tracking_points.value:
+    for track in tracking_points:
         if len(track) > 1:
-            for i in range(len(track)-1):
+            for i in range(len(track) - 1):
                 start_point = track[i]
-                end_point = track[i+1]
+                end_point = track[i + 1]
                 vx = end_point[0] - start_point[0]
                 vy = end_point[1] - start_point[1]
                 arrow_length = np.sqrt(vx**2 + vy**2)
-                if i == len(track)-2:
-                    cv2.arrowedLine(transparent_layer, tuple(start_point), tuple(end_point), color, 2, tipLength=8 / arrow_length)
+                if i == len(track) - 2:
+                    cv2.arrowedLine(
+                        transparent_layer, tuple(start_point), tuple(end_point), color, 2, tipLength=8 / arrow_length
+                    )
                 else:
-                    cv2.line(transparent_layer, tuple(start_point), tuple(end_point), color, 2,)
+                    cv2.line(
+                        transparent_layer,
+                        tuple(start_point),
+                        tuple(end_point),
+                        color,
+                        2,
+                    )
         else:
             cv2.circle(transparent_layer, tuple(track[0]), 5, color, -1)
 
     transparent_layer = Image.fromarray(transparent_layer.astype(np.uint8))
     trajectory_map = Image.alpha_composite(transparent_background, transparent_layer)
     return {tracking_points_var: tracking_points, input_image: trajectory_map}
-    
+
 
 def delete_last_step(tracking_points, first_frame_path, drag_mode):
-    if drag_mode=='object':
+    if drag_mode == "object":
         color = (255, 0, 0, 255)
-    elif drag_mode=='camera':
+    elif drag_mode == "camera":
         color = (0, 0, 255, 255)
-    tracking_points.value[-1].pop()
-    transparent_background = Image.open(first_frame_path).convert('RGBA')
+    if tracking_points and tracking_points[-1]:
+        tracking_points[-1].pop()
+    transparent_background = Image.open(first_frame_path).convert("RGBA")
     w, h = transparent_background.size
     transparent_layer = np.zeros((h, w, 4))
-    for track in tracking_points.value:
+    for track in tracking_points:
+        if not track:
+            continue
         if len(track) > 1:
-            for i in range(len(track)-1):
+            for i in range(len(track) - 1):
                 start_point = track[i]
-                end_point = track[i+1]
+                end_point = track[i + 1]
                 vx = end_point[0] - start_point[0]
                 vy = end_point[1] - start_point[1]
                 arrow_length = np.sqrt(vx**2 + vy**2)
-                if i == len(track)-2:
-                    cv2.arrowedLine(transparent_layer, tuple(start_point), tuple(end_point), color, 2, tipLength=8 / arrow_length)
+                if i == len(track) - 2:
+                    cv2.arrowedLine(
+                        transparent_layer, tuple(start_point), tuple(end_point), color, 2, tipLength=8 / arrow_length
+                    )
                 else:
-                    cv2.line(transparent_layer, tuple(start_point), tuple(end_point), color, 2,)
+                    cv2.line(
+                        transparent_layer,
+                        tuple(start_point),
+                        tuple(end_point),
+                        color,
+                        2,
+                    )
         else:
-            cv2.circle(transparent_layer, tuple(track[0]), 5,color, -1)
+            cv2.circle(transparent_layer, tuple(track[0]), 5, color, -1)
 
     transparent_layer = Image.fromarray(transparent_layer.astype(np.uint8))
     trajectory_map = Image.alpha_composite(transparent_background, transparent_layer)
     return {tracking_points_var: tracking_points, input_image: trajectory_map}
 
 
-block = gr.Blocks(
-        theme=gr.themes.Soft(
-            radius_size=gr.themes.sizes.radius_none,
-            text_size=gr.themes.sizes.text_md
-        )
-        )
+def load_example(drag_mode, examples_type):
+    example_image_path = IMAGE_PATH[examples_type]
+    with open(POINTS[examples_type]) as f:
+        tracking_points = json.load(f)
+    tracking_points = np.round(tracking_points).astype(int).tolist()
+    trajectory_map, first_frame_path = preprocess_example_image(example_image_path, tracking_points, drag_mode)
+    return {input_image: trajectory_map, first_frame_path_var: first_frame_path, tracking_points_var: tracking_points}
+
+
+device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
+ImageConductor_net = ImageConductor(
+    device=device,
+    unet_path=unet_path,
+    image_controlnet_path=image_controlnet_path,
+    flow_controlnet_path=flow_controlnet_path,
+    height=256,
+    width=384,
+    model_length=16,
+)
+
+block = gr.Blocks(theme=gr.themes.Soft(radius_size=gr.themes.sizes.radius_none, text_size=gr.themes.sizes.text_md))
 with block:
     with gr.Row():
         with gr.Column():
@@ -557,66 +637,58 @@ with block:
 
     with gr.Accordion(label="🛠️ Instructions:", open=True, elem_id="accordion"):
         with gr.Row(equal_height=True):
-            gr.Markdown(instructions)      
-
-
-    # device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
-    device = torch.device("cuda")
-    unet_path = 'models/unet.ckpt'
-    image_controlnet_path = 'models/image_controlnet.ckpt'
-    flow_controlnet_path = 'models/flow_controlnet.ckpt'
-    ImageConductor_net = ImageConductor(device=device, 
-                                        unet_path=unet_path, 
-                                        image_controlnet_path=image_controlnet_path, 
-                                        flow_controlnet_path=flow_controlnet_path, 
-                                        height=256,
-                                        width=384,
-                                        model_length=16
-                                        )
-    first_frame_path_var = gr.State(value=None)
+            gr.Markdown(instructions)
+
+    first_frame_path_var = gr.State()
     tracking_points_var = gr.State([])
 
     with gr.Row():
         with gr.Column(scale=1):
-            image_upload_button = gr.UploadButton(label="Upload Image",file_types=["image"])
+            image_upload_button = gr.UploadButton(label="Upload Image", file_types=["image"])
             add_drag_button = gr.Button(value="Add Drag")
             reset_button = gr.Button(value="Reset")
             delete_last_drag_button = gr.Button(value="Delete last drag")
             delete_last_step_button = gr.Button(value="Delete last step")
-            
-            
 
         with gr.Column(scale=7):
             with gr.Row():
                 with gr.Column(scale=6):
-                    input_image = gr.Image(label="Input Image",
-                                        interactive=True,
-                                        height=300,
-                                        width=384,)
+                    input_image = gr.Image(
+                        label="Input Image",
+                        interactive=True,
+                        height=300,
+                        width=384,
+                    )
                 with gr.Column(scale=6):
-                    output_image = gr.Image(label="Motion Path",
-                                            interactive=False,
-                                            height=256,
-                                            width=384,)
+                    output_image = gr.Image(
+                        label="Motion Path",
+                        interactive=False,
+                        height=256,
+                        width=384,
+                    )
     with gr.Row():
         with gr.Column(scale=1):
-            prompt = gr.Textbox(value="a wonderful elf.", label="Prompt (highly-recommended)", interactive=True, visible=True)
+            prompt = gr.Textbox(
+                value="a wonderful elf.",
+                label="Prompt (highly-recommended)",
+                interactive=True,
+                visible=True,
+            )
             negative_prompt = gr.Text(
-                        label="Negative Prompt",
-                        max_lines=5,
-                        placeholder="Please input your negative prompt",
-                        value='worst quality, low quality, letterboxed',lines=1
-                    )
-            drag_mode = gr.Radio(['camera', 'object'], label='Drag mode: ', value='object', scale=2)
+                label="Negative Prompt",
+                max_lines=5,
+                placeholder="Please input your negative prompt",
+                value="worst quality, low quality, letterboxed",
+                lines=1,
+            )
+            drag_mode = gr.Radio(["camera", "object"], label="Drag mode: ", value="object", scale=2)
             run_button = gr.Button(value="Run")
 
             with gr.Accordion("More input params", open=False, elem_id="accordion1"):
                 with gr.Group():
-                    seed = gr.Textbox(
-                        label="Seed: ",  value=561793204,
-                    )
+                    seed = gr.Textbox(label="Seed: ", value=561793204)
                     randomize_seed = gr.Checkbox(label="Randomize seed", value=False)
-                
+
                 with gr.Group():
                     with gr.Row():
                         guidance_scale = gr.Slider(
@@ -633,24 +705,15 @@ with block:
                             step=1,
                             value=25,
                         )
-                        
+
                 with gr.Group():
-                    personalized = gr.Dropdown(label="Personalized", choices=["", 'HelloObject', 'TUSUN'], value="")
-                    examples_type = gr.Textbox(label="Examples Type (Ignore) ",  value="", visible=False)
+                    personalized = gr.Dropdown(label="Personalized", choices=["", "HelloObject", "TUSUN"], value="")
+                    examples_type = gr.Textbox(label="Examples Type (Ignore) ", value="", visible=False)
 
         with gr.Column(scale=7):
-            output_video = gr.Video(
-                                    label="Output Video", 
-                                    width=384, 
-                                    height=256)
-            # output_video = gr.Image(label="Output Video",
-            #                                 height=256,
-            #                                 width=384,)
-            
-            
-    with gr.Row():
-   
+            output_video = gr.Video(label="Output Video", width=384, height=256)
 
+    with gr.Row():
         example = gr.Examples(
             label="Input Example",
             examples=image_examples,
@@ -658,26 +721,65 @@ with block:
             examples_per_page=10,
             cache_examples=False,
         )
-        
-        
+
     with gr.Row():
         gr.Markdown(citation)
 
-    
-    image_upload_button.upload(preprocess_image, [image_upload_button, tracking_points_var], [input_image, first_frame_path_var, tracking_points_var, personalized])
+    image_upload_button.upload(
+        preprocess_image,
+        [image_upload_button, tracking_points_var],
+        [input_image, first_frame_path_var, tracking_points_var, personalized],
+    )
 
     add_drag_button.click(add_drag, tracking_points_var, tracking_points_var)
 
-    delete_last_drag_button.click(delete_last_drag, [tracking_points_var, first_frame_path_var, drag_mode], [tracking_points_var, input_image])
-
-    delete_last_step_button.click(delete_last_step, [tracking_points_var, first_frame_path_var, drag_mode], [tracking_points_var, input_image])
-
-    reset_button.click(reset_states, [first_frame_path_var, tracking_points_var], [input_image, first_frame_path_var, tracking_points_var])
-
-    input_image.select(add_tracking_points, [tracking_points_var, first_frame_path_var, drag_mode], [tracking_points_var, input_image])
-
-    run_button.click(ImageConductor_net.run, [first_frame_path_var, tracking_points_var, prompt, drag_mode, 
-                                            negative_prompt, seed, randomize_seed, guidance_scale, num_inference_steps, personalized, examples_type], 
-                                            [output_image, output_video])
+    delete_last_drag_button.click(
+        delete_last_drag,
+        [tracking_points_var, first_frame_path_var, drag_mode],
+        [tracking_points_var, input_image],
+    )
+
+    delete_last_step_button.click(
+        delete_last_step,
+        [tracking_points_var, first_frame_path_var, drag_mode],
+        [tracking_points_var, input_image],
+    )
+
+    reset_button.click(
+        reset_states,
+        [first_frame_path_var, tracking_points_var],
+        [input_image, first_frame_path_var, tracking_points_var],
+    )
+
+    input_image.select(
+        add_tracking_points,
+        [tracking_points_var, first_frame_path_var, drag_mode],
+        [tracking_points_var, input_image],
+    )
+
+    run_button.click(
+        ImageConductor_net.run,
+        [
+            first_frame_path_var,
+            tracking_points_var,
+            prompt,
+            drag_mode,
+            negative_prompt,
+            seed,
+            randomize_seed,
+            guidance_scale,
+            num_inference_steps,
+            personalized,
+        ],
+        [output_image, output_video],
+    )
+
+    examples_type.change(
+        fn=load_example,
+        inputs=[drag_mode, examples_type],
+        outputs=[input_image, first_frame_path_var, tracking_points_var],
+        api_name=False,
+        queue=False,
+    )
 
 block.queue().launch()