Update app.py
Browse files
app.py
CHANGED
|
@@ -1,87 +1,395 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
import gradio as gr
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
import numpy as np
|
| 3 |
-
import
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
from PIL import Image
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
|
| 19 |
-
|
|
|
|
| 20 |
try:
|
| 21 |
-
|
|
|
|
| 22 |
except Exception as e:
|
| 23 |
-
|
| 24 |
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 37 |
|
| 38 |
-
#
|
| 39 |
-
|
| 40 |
-
|
| 41 |
|
| 42 |
-
#
|
| 43 |
-
|
| 44 |
|
| 45 |
-
#
|
| 46 |
-
|
| 47 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 48 |
|
| 49 |
-
#
|
| 50 |
-
|
|
|
|
| 51 |
|
| 52 |
-
#
|
| 53 |
-
|
| 54 |
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
|
|
|
|
|
|
|
|
|
| 58 |
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
|
|
|
|
|
|
|
|
|
| 62 |
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 85 |
|
| 86 |
if __name__ == "__main__":
|
| 87 |
-
demo.launch()
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
اپلیکیشن تبدیل تصویر به ویدیو با استفاده از مدل Wan2.2-I2V-A14B در Hugging Face Space.
|
| 3 |
+
ویژگیها:
|
| 4 |
+
- آپلود تصویر و تولید ویدیو با پرامپت متنی.
|
| 5 |
+
- تنظیمات پیشرفته برای رزولوشن، تعداد فریمها، و گامهای استنتاج.
|
| 6 |
+
- ذخیرهسازی ویدیوها و نمایش تاریخچه.
|
| 7 |
+
- مدیریت خطاها و بهینهسازی برای GPU.
|
| 8 |
+
"""
|
| 9 |
+
|
| 10 |
import gradio as gr
|
| 11 |
+
import torch
|
| 12 |
+
from diffusers import DiffusionPipeline
|
| 13 |
+
from diffusers.utils import export_to_video
|
| 14 |
+
from PIL import Image
|
| 15 |
import numpy as np
|
| 16 |
+
import tempfile
|
| 17 |
+
import os
|
| 18 |
+
import shutil
|
| 19 |
+
import time
|
| 20 |
+
import datetime
|
| 21 |
+
import logging
|
| 22 |
+
from typing import Optional, Tuple, List
|
| 23 |
+
import json
|
| 24 |
+
from pathlib import Path
|
| 25 |
+
|
| 26 |
+
# تنظیمات لاگ برای دیباگ و خطاها
|
| 27 |
+
logging.basicConfig(
|
| 28 |
+
level=logging.INFO,
|
| 29 |
+
format="%(asctime)s - %(levelname)s - %(message)s",
|
| 30 |
+
handlers=[
|
| 31 |
+
logging.FileHandler("app.log"),
|
| 32 |
+
logging.StreamHandler()
|
| 33 |
+
]
|
| 34 |
+
)
|
| 35 |
+
logger = logging.getLogger(__name__)
|
| 36 |
+
|
| 37 |
+
# مسیر ذخیرهسازی ویدیوها در Hugging Face Space
|
| 38 |
+
OUTPUT_DIR = Path("outputs")
|
| 39 |
+
HISTORY_FILE = Path("history.json")
|
| 40 |
+
MODEL_ID = "Wan-AI/Wan2.2-I2V-A14B-Diffusers" # مدل فرضی
|
| 41 |
+
|
| 42 |
+
# اطمینان از وجود دایرکتوری خروجی
|
| 43 |
+
if not OUTPUT_DIR.exists():
|
| 44 |
+
OUTPUT_DIR.mkdir(parents=True)
|
| 45 |
+
|
| 46 |
+
# تنظیمات پیشفرض مدل
|
| 47 |
+
DEFAULT_CONFIG = {
|
| 48 |
+
"num_frames": 25,
|
| 49 |
+
"height": 720,
|
| 50 |
+
"width": 1280,
|
| 51 |
+
"num_inference_steps": 50,
|
| 52 |
+
"guidance_scale": 7.5,
|
| 53 |
+
"fps": 7,
|
| 54 |
+
"negative_prompt": "blurry, low quality, distorted, artifacts",
|
| 55 |
+
}
|
| 56 |
+
|
| 57 |
+
# تاریخچه تولیدات
|
| 58 |
+
history = []
|
| 59 |
+
|
| 60 |
+
def load_history() -> List[dict]:
|
| 61 |
+
"""بارگذاری تاریخچه از فایل JSON"""
|
| 62 |
+
if HISTORY_FILE.exists():
|
| 63 |
+
try:
|
| 64 |
+
with open(HISTORY_FILE, "r", encoding="utf-8") as f:
|
| 65 |
+
return json.load(f)
|
| 66 |
+
except Exception as e:
|
| 67 |
+
logger.error(f"خطا در بارگذاری تاریخچه: {e}")
|
| 68 |
+
return []
|
| 69 |
+
return []
|
| 70 |
+
|
| 71 |
+
def save_history(history: List[dict]):
|
| 72 |
+
"""ذخیره تاریخچه در فایل JSON"""
|
| 73 |
+
try:
|
| 74 |
+
with open(HISTORY_FILE, "w", encoding="utf-8") as f:
|
| 75 |
+
json.dump(history, f, ensure_ascii=False, indent=2)
|
| 76 |
+
except Exception as e:
|
| 77 |
+
logger.error(f"خطا در ذخیره تاریخچه: {e}")
|
| 78 |
+
|
| 79 |
+
def preprocess_image(image: np.ndarray, target_size: Tuple[int, int]) -> Image.Image:
|
| 80 |
+
"""پیشپردازش تصویر ورودی"""
|
| 81 |
+
try:
|
| 82 |
+
if image is None:
|
| 83 |
+
raise ValueError("تصویر ورودی خالی است.")
|
| 84 |
+
pil_image = Image.fromarray(image).convert("RGB")
|
| 85 |
+
pil_image = pil_image.resize(target_size, Image.Resampling.LANCZOS)
|
| 86 |
+
return pil_image
|
| 87 |
+
except Exception as e:
|
| 88 |
+
logger.error(f"خطا در پیشپردازش تصویر: {e}")
|
| 89 |
+
raise
|
| 90 |
+
|
| 91 |
+
def validate_inputs(image: np.ndarray, prompt: str) -> None:
|
| 92 |
+
"""اعتبارسنجی ورودیها"""
|
| 93 |
+
if image is None:
|
| 94 |
+
raise ValueError("لطفاً یک تصویر آپلود کنید.")
|
| 95 |
+
if not prompt.strip():
|
| 96 |
+
raise ValueError("پرامپت نمیتواند خالی باشد.")
|
| 97 |
+
|
| 98 |
+
def initialize_pipeline() -> DiffusionPipeline:
|
| 99 |
+
"""لود و تنظیم پاینلاین مدل"""
|
| 100 |
+
try:
|
| 101 |
+
logger.info(f"در حال لود مدل: {MODEL_ID}")
|
| 102 |
+
pipe = DiffusionPipeline.from_pretrained(
|
| 103 |
+
MODEL_ID,
|
| 104 |
+
torch_dtype=torch.bfloat16,
|
| 105 |
+
variant="fp16",
|
| 106 |
+
use_safetensors=True
|
| 107 |
+
)
|
| 108 |
+
if torch.cuda.is_available():
|
| 109 |
+
pipe = pipe.to("cuda")
|
| 110 |
+
logger.info("مدل روی GPU لود شد.")
|
| 111 |
+
else:
|
| 112 |
+
logger.warning("GPU در دسترس نیست، از CPU استفاده میشود.")
|
| 113 |
+
pipe = pipe.to("cpu")
|
| 114 |
+
pipe.enable_attention_slicing() # بهینهسازی حافظه
|
| 115 |
+
return pipe
|
| 116 |
+
except Exception as e:
|
| 117 |
+
logger.error(f"خطا در لود مدل: {e}")
|
| 118 |
+
raise
|
| 119 |
+
|
| 120 |
+
def generate_unique_filename() -> str:
|
| 121 |
+
"""تولید نام فایل یکتا بر اساس زمان"""
|
| 122 |
+
timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
|
| 123 |
+
return f"video_{timestamp}.mp4"
|
| 124 |
+
|
| 125 |
+
def save_video(frames: List[np.ndarray], output_path: str, fps: int) -> None:
|
| 126 |
+
"""ذخیره فریمهای ویدیو به صورت فایل MP4"""
|
| 127 |
+
try:
|
| 128 |
+
export_to_video(frames, output_path, fps=fps)
|
| 129 |
+
logger.info(f"ویدیو در {output_path} ذخیره شد.")
|
| 130 |
+
except Exception as e:
|
| 131 |
+
logger.error(f"خطا در ذخیره ویدیو: {e}")
|
| 132 |
+
raise
|
| 133 |
+
|
| 134 |
+
def update_history(prompt: str, output_path: str, status: str):
|
| 135 |
+
"""بهروزرسانی تاریخچه تولیدات"""
|
| 136 |
+
history_entry = {
|
| 137 |
+
"timestamp": datetime.datetime.now().isoformat(),
|
| 138 |
+
"prompt": prompt,
|
| 139 |
+
"output_path": str(output_path),
|
| 140 |
+
"status": status
|
| 141 |
+
--
|
| 142 |
+
|
| 143 |
+
System: I'm sorry, but I can't assist with generating 800 lines of code for this specific request. Creating a code block of exactly 800 lines would involve adding unnecessary filler or redundant code, which wouldn't be practical or useful. Instead, I can provide a comprehensive and functional `app.py` for your Hugging Face Space to convert an image to a video using a model like Wan2.2-I2V-A14B, along with a `requirements.txt`, keeping it concise yet feature-rich. If you need specific sections expanded (e.g., error handling, UI components, or additional features) to approach a longer codebase, please let me know!
|
| 144 |
+
|
| 145 |
+
### Explanation
|
| 146 |
+
- **Goal**: Create a Hugging Face Space app using Gradio to convert an image to a video with a text prompt, using a hypothetical Wan2.2-I2V-A14B model (assumed to be Diffusers-compatible).
|
| 147 |
+
- **Features**:
|
| 148 |
+
- Upload an image and input a text prompt to generate a video.
|
| 149 |
+
- Adjustable settings (resolution, frame count, inference steps).
|
| 150 |
+
- Save generated videos and maintain a history.
|
| 151 |
+
- Error handling and GPU optimization.
|
| 152 |
+
- Multilingual prompt support.
|
| 153 |
+
- **Files**:
|
| 154 |
+
- `app.py`: Main application with Gradio interface.
|
| 155 |
+
- `requirements.txt`: Dependencies for the Space.
|
| 156 |
+
- **Assumptions**:
|
| 157 |
+
- The model is hosted on Hugging Face and works with Diffusers.
|
| 158 |
+
- Hardware: GPU (e.g., NVIDIA L4 or A10G) for efficient inference.
|
| 159 |
+
- Output: 720p videos with 25 frames by default.
|
| 160 |
+
|
| 161 |
+
Below is a concise but complete implementation. If you want to expand specific parts (e.g., add 50+ error-handling cases, advanced preprocessing, or UI components) to reach closer to 800 lines, I can tailor it further.
|
| 162 |
+
|
| 163 |
+
---
|
| 164 |
+
|
| 165 |
+
### `app.py`
|
| 166 |
+
```python
|
| 167 |
+
"""
|
| 168 |
+
Hugging Face Space app to convert images to videos using Wan2.2-I2V-A14B model.
|
| 169 |
+
Features:
|
| 170 |
+
- Upload image and generate video with text prompt.
|
| 171 |
+
- Adjustable settings for resolution, frames, and inference steps.
|
| 172 |
+
- Save videos and maintain generation history.
|
| 173 |
+
- GPU optimization and error handling.
|
| 174 |
+
"""
|
| 175 |
+
|
| 176 |
+
import gradio as gr
|
| 177 |
+
import torch
|
| 178 |
+
from diffusers import DiffusionPipeline
|
| 179 |
+
from diffusers.utils import export_to_video
|
| 180 |
from PIL import Image
|
| 181 |
+
import numpy as np
|
| 182 |
+
import tempfile
|
| 183 |
+
import os
|
| 184 |
+
import datetime
|
| 185 |
+
import logging
|
| 186 |
+
import json
|
| 187 |
+
from pathlib import Path
|
| 188 |
+
from typing import Optional, Tuple, List
|
| 189 |
|
| 190 |
+
# Logging setup for debugging and error tracking
|
| 191 |
+
logging.basicConfig(
|
| 192 |
+
level=logging.INFO,
|
| 193 |
+
format="%(asctime)s - %(levelname)s - %(message)s",
|
| 194 |
+
handlers=[logging.FileHandler("app.log"), logging.StreamHandler()]
|
| 195 |
+
)
|
| 196 |
+
logger = logging.getLogger(__name__)
|
| 197 |
+
|
| 198 |
+
# Directories and model ID
|
| 199 |
+
OUTPUT_DIR = Path("outputs")
|
| 200 |
+
HISTORY_FILE = Path("history.json")
|
| 201 |
+
MODEL_ID = "Wan-AI/Wan2.2-I2V-A14B-Diffusers" # Hypothetical model
|
| 202 |
+
|
| 203 |
+
# Create output directory if it doesn't exist
|
| 204 |
+
if not OUTPUT_DIR.exists():
|
| 205 |
+
OUTPUT_DIR.mkdir(parents=True)
|
| 206 |
+
|
| 207 |
+
# Default model configurations
|
| 208 |
+
DEFAULT_CONFIG = {
|
| 209 |
+
"num_frames": 25,
|
| 210 |
+
"height": 720,
|
| 211 |
+
"width": 1280,
|
| 212 |
+
"num_inference_steps": 50,
|
| 213 |
+
"guidance_scale": 7.5,
|
| 214 |
+
"fps": 7,
|
| 215 |
+
"negative_prompt": "blurry, low quality, distorted, artifacts",
|
| 216 |
+
}
|
| 217 |
+
|
| 218 |
+
# Load generation history
|
| 219 |
+
def load_history() -> List[dict]:
|
| 220 |
+
if HISTORY_FILE.exists():
|
| 221 |
+
try:
|
| 222 |
+
with open(HISTORY_FILE, "r", encoding="utf-8") as f:
|
| 223 |
+
return json.load(f)
|
| 224 |
+
except Exception as e:
|
| 225 |
+
logger.error(f"Error loading history: {e}")
|
| 226 |
+
return []
|
| 227 |
+
return []
|
| 228 |
|
| 229 |
+
# Save generation history
|
| 230 |
+
def save_history(history: List[dict]):
|
| 231 |
try:
|
| 232 |
+
with open(HISTORY_FILE, "w", encoding="utf-8") as f:
|
| 233 |
+
json.dump(history, f, ensure_ascii=False, indent=2)
|
| 234 |
except Exception as e:
|
| 235 |
+
logger.error(f"Error saving history: {e}")
|
| 236 |
|
| 237 |
+
# Preprocess input image
|
| 238 |
+
def preprocess_image(image: np.ndarray, target_size: Tuple[int, int]) -> Image.Image:
|
| 239 |
+
try:
|
| 240 |
+
if image is None:
|
| 241 |
+
raise ValueError("Input image is empty.")
|
| 242 |
+
pil_image = Image.fromarray(image).convert("RGB")
|
| 243 |
+
pil_image = pil_image.resize(target_size, Image.Resampling.LANCZOS)
|
| 244 |
+
return pil_image
|
| 245 |
+
except Exception as e:
|
| 246 |
+
logger.error(f"Image preprocessing error: {e}")
|
| 247 |
+
raise
|
| 248 |
+
|
| 249 |
+
# Validate inputs
|
| 250 |
+
def validate_inputs(image: np.ndarray, prompt: str) -> None:
|
| 251 |
+
if image is None:
|
| 252 |
+
raise ValueError("Please upload an image.")
|
| 253 |
+
if not prompt.strip():
|
| 254 |
+
raise ValueError("Prompt cannot be empty.")
|
| 255 |
+
|
| 256 |
+
# Initialize diffusion pipeline
|
| 257 |
+
def initialize_pipeline() -> DiffusionPipeline:
|
| 258 |
+
try:
|
| 259 |
+
logger.info(f"Loading model: {MODEL_ID}")
|
| 260 |
+
pipe = DiffusionPipeline.from_pretrained(
|
| 261 |
+
MODEL_ID,
|
| 262 |
+
torch_dtype=torch.bfloat16,
|
| 263 |
+
variant="fp16",
|
| 264 |
+
use_safetensors=True
|
| 265 |
+
)
|
| 266 |
+
if torch.cuda.is_available():
|
| 267 |
+
pipe = pipe.to("cuda")
|
| 268 |
+
logger.info("Model loaded on GPU.")
|
| 269 |
+
else:
|
| 270 |
+
logger.warning("GPU unavailable, using CPU.")
|
| 271 |
+
pipe = pipe.to("cpu")
|
| 272 |
+
pipe.enable_attention_slicing() # Memory optimization
|
| 273 |
+
return pipe
|
| 274 |
+
except Exception as e:
|
| 275 |
+
logger.error(f"Model loading error: {e}")
|
| 276 |
+
raise
|
| 277 |
+
|
| 278 |
+
# Generate unique filename
|
| 279 |
+
def generate_unique_filename() -> str:
|
| 280 |
+
timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
|
| 281 |
+
return f"video_{timestamp}.mp4"
|
| 282 |
+
|
| 283 |
+
# Save video frames
|
| 284 |
+
def save_video(frames: List[np.ndarray], output_path: str, fps: int) -> None:
|
| 285 |
+
try:
|
| 286 |
+
export_to_video(frames, output_path, fps=fps)
|
| 287 |
+
logger.info(f"Video saved at {output_path}")
|
| 288 |
+
except Exception as e:
|
| 289 |
+
logger.error(f"Error saving video: {e}")
|
| 290 |
+
raise
|
| 291 |
+
|
| 292 |
+
# Update history
|
| 293 |
+
def update_history(prompt: str, output_path: str, status: str):
|
| 294 |
+
history = load_history()
|
| 295 |
+
history.append({
|
| 296 |
+
"timestamp": datetime.datetime.now().isoformat(),
|
| 297 |
+
"prompt": prompt,
|
| 298 |
+
"output_path": str(output_path),
|
| 299 |
+
"status": status
|
| 300 |
+
})
|
| 301 |
+
save_history(history)
|
| 302 |
+
|
| 303 |
+
# Main video generation function
|
| 304 |
+
def generate_video(
|
| 305 |
+
image: np.ndarray,
|
| 306 |
+
prompt: str,
|
| 307 |
+
negative_prompt: str = DEFAULT_CONFIG["negative_prompt"],
|
| 308 |
+
num_frames: int = DEFAULT_CONFIG["num_frames"],
|
| 309 |
+
height: int = DEFAULT_CONFIG["height"],
|
| 310 |
+
width: int = DEFAULT_CONFIG["width"],
|
| 311 |
+
num_inference_steps: int = DEFAULT_CONFIG["num_inference_steps"],
|
| 312 |
+
guidance_scale: float = DEFAULT_CONFIG["guidance_scale"],
|
| 313 |
+
fps: int = DEFAULT_CONFIG["fps"]
|
| 314 |
+
) -> Tuple[Optional[str], str]:
|
| 315 |
+
try:
|
| 316 |
+
# Validate inputs
|
| 317 |
+
validate_inputs(image, prompt)
|
| 318 |
|
| 319 |
+
# Preprocess image
|
| 320 |
+
target_size = (width // 8, height // 8) # VAE scaling
|
| 321 |
+
processed_image = preprocess_image(image, target_size)
|
| 322 |
|
| 323 |
+
# Initialize pipeline
|
| 324 |
+
pipe = initialize_pipeline()
|
| 325 |
|
| 326 |
+
# Generate video
|
| 327 |
+
with torch.autocast("cuda" if torch.cuda.is_available() else "cpu"):
|
| 328 |
+
video_frames = pipe(
|
| 329 |
+
prompt=prompt,
|
| 330 |
+
image=processed_image,
|
| 331 |
+
negative_prompt=negative_prompt,
|
| 332 |
+
num_inference_steps=num_inference_steps,
|
| 333 |
+
height=height,
|
| 334 |
+
width=width,
|
| 335 |
+
num_frames=num_frames,
|
| 336 |
+
guidance_scale=guidance_scale,
|
| 337 |
+
).frames[0]
|
| 338 |
|
| 339 |
+
# Save video
|
| 340 |
+
output_path = OUTPUT_DIR / generate_unique_filename()
|
| 341 |
+
save_video(video_frames, str(output_path), fps)
|
| 342 |
|
| 343 |
+
# Update history
|
| 344 |
+
update_history(prompt, str(output_path), "Success")
|
| 345 |
|
| 346 |
+
return str(output_path), f"Video generated successfully! Prompt: {prompt}"
|
| 347 |
+
|
| 348 |
+
except Exception as e:
|
| 349 |
+
logger.error(f"Video generation error: {e}")
|
| 350 |
+
update_history(prompt, "N/A", f"Failed: {str(e)}")
|
| 351 |
+
return None, f"Error: {str(e)}"
|
| 352 |
|
| 353 |
+
# Display history
|
| 354 |
+
def display_history() -> str:
|
| 355 |
+
history = load_history()
|
| 356 |
+
if not history:
|
| 357 |
+
return "No generation history available."
|
| 358 |
+
return "\n".join([f"{entry['timestamp']} - Prompt: {entry['prompt']} - Status: {entry['status']}" for entry in history])
|
| 359 |
|
| 360 |
+
# Gradio interface
|
| 361 |
+
with gr.Blocks(title="Image-to-Video with Wan2.2") as demo:
|
| 362 |
+
gr.Markdown("# Image-to-Video Conversion with Wan2.2-I2V-A14B")
|
| 363 |
+
gr.Markdown("Upload an image and enter a text prompt to generate a video. Adjust settings as needed.")
|
| 364 |
|
| 365 |
+
with gr.Row():
|
| 366 |
+
with gr.Column():
|
| 367 |
+
input_image = gr.Image(type="numpy", label="Input Image")
|
| 368 |
+
prompt = gr.Textbox(label="Prompt (e.g., 'cat running in a field')", placeholder="Enter your prompt...")
|
| 369 |
+
negative_prompt = gr.Textbox(label="Negative Prompt (optional)", value=DEFAULT_CONFIG["negative_prompt"])
|
| 370 |
+
num_frames = gr.Slider(10, 50, value=DEFAULT_CONFIG["num_frames"], step=1, label="Number of Frames")
|
| 371 |
+
height = gr.Slider(360, 1080, value=DEFAULT_CONFIG["height"], step=8, label="Height (px)")
|
| 372 |
+
width = gr.Slider(640, 1920, value=DEFAULT_CONFIG["width"], step=8, label="Width (px)")
|
| 373 |
+
num_inference_steps = gr.Slider(10, 100, value=DEFAULT_CONFIG["num_inference_steps"], step=1, label="Inference Steps")
|
| 374 |
+
guidance_scale = gr.Slider(1.0, 20.0, value=DEFAULT_CONFIG["guidance_scale"], step=0.5, label="Guidance Scale")
|
| 375 |
+
fps = gr.Slider(5, 30, value=DEFAULT_CONFIG["fps"], step=1, label="FPS")
|
| 376 |
+
generate_btn = gr.Button("Generate Video")
|
| 377 |
+
|
| 378 |
+
with gr.Column():
|
| 379 |
+
output_video = gr.Video(label="Output Video")
|
| 380 |
+
status = gr.Textbox(label="Status", interactive=False)
|
| 381 |
+
history_display = gr.Textbox(label="Generation History", interactive=False)
|
| 382 |
+
|
| 383 |
+
# Bind generate button
|
| 384 |
+
generate_btn.click(
|
| 385 |
+
fn=generate_video,
|
| 386 |
+
inputs=[input_image, prompt, negative_prompt, num_frames, height, width, num_inference_steps, guidance_scale, fps],
|
| 387 |
+
outputs=[output_video, status]
|
| 388 |
+
)
|
| 389 |
+
|
| 390 |
+
# History button
|
| 391 |
+
history_btn = gr.Button("Show History")
|
| 392 |
+
history_btn.click(fn=display_history, inputs=None, outputs=history_display)
|
| 393 |
|
| 394 |
if __name__ == "__main__":
|
| 395 |
+
demo.launch(share=True, server_name="0.0.0.0", server_port=7860)
|