Spaces:
Running
on
Zero
Running
on
Zero
Upload 3 files
Browse files- README.md +8 -17
- app.py +385 -997
- requirements.txt +10 -42
README.md
CHANGED
|
@@ -1,21 +1,12 @@
|
|
| 1 |
---
|
| 2 |
-
title:
|
|
|
|
|
|
|
|
|
|
| 3 |
sdk: gradio
|
| 4 |
-
emoji: 📷
|
| 5 |
sdk_version: 5.29.1
|
| 6 |
app_file: app.py
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
- Upscaling
|
| 12 |
-
- Restoring
|
| 13 |
-
- Image-to-Image
|
| 14 |
-
- Image-2-Image
|
| 15 |
-
- Img-to-Img
|
| 16 |
-
- Img-2-Img
|
| 17 |
-
- language models
|
| 18 |
-
- LLMs
|
| 19 |
-
short_description: Restore blurred or small images with prompt
|
| 20 |
-
suggested_hardware: zero-a10g
|
| 21 |
-
---
|
|
|
|
| 1 |
---
|
| 2 |
+
title: Wan 2 2 First Last Frame
|
| 3 |
+
emoji: 💻
|
| 4 |
+
colorFrom: purple
|
| 5 |
+
colorTo: gray
|
| 6 |
sdk: gradio
|
|
|
|
| 7 |
sdk_version: 5.29.1
|
| 8 |
app_file: app.py
|
| 9 |
+
pinned: false
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
app.py
CHANGED
|
@@ -1,1058 +1,446 @@
|
|
| 1 |
import os
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
import gradio as gr
|
| 3 |
-
import
|
| 4 |
import numpy as np
|
| 5 |
-
import
|
| 6 |
-
import einops
|
| 7 |
-
import copy
|
| 8 |
-
import math
|
| 9 |
-
import time
|
| 10 |
import random
|
| 11 |
-
|
|
|
|
| 12 |
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
except:
|
| 16 |
-
class spaces():
|
| 17 |
-
def GPU(*args, **kwargs):
|
| 18 |
-
def decorator(function):
|
| 19 |
-
return lambda *dummy_args, **dummy_kwargs: function(*dummy_args, **dummy_kwargs)
|
| 20 |
-
return decorator
|
| 21 |
|
| 22 |
-
|
| 23 |
-
|
| 24 |
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
import pillow_heif
|
| 31 |
-
|
| 32 |
-
pillow_heif.register_heif_opener()
|
| 33 |
|
| 34 |
-
|
| 35 |
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
hf_hub_download(repo_id="camenduru/SUPIR", filename="SUPIR-v0Q.ckpt", local_dir="yushan777_SUPIR")
|
| 40 |
-
hf_hub_download(repo_id="RunDiffusion/Juggernaut-XL-Lightning", filename="Juggernaut_RunDiffusionPhoto2_Lightning_4Steps.safetensors", local_dir="RunDiffusion_Juggernaut-XL-Lightning")
|
| 41 |
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
parser.add_argument("--ip", type=str, default='127.0.0.1')
|
| 45 |
-
parser.add_argument("--port", type=int, default='6688')
|
| 46 |
-
parser.add_argument("--no_llava", action='store_true', default=True)#False
|
| 47 |
-
parser.add_argument("--use_image_slider", action='store_true', default=False)#False
|
| 48 |
-
parser.add_argument("--log_history", action='store_true', default=False)
|
| 49 |
-
parser.add_argument("--loading_half_params", action='store_true', default=False)#False
|
| 50 |
-
parser.add_argument("--use_tile_vae", action='store_true', default=True)#False
|
| 51 |
-
parser.add_argument("--encoder_tile_size", type=int, default=512)
|
| 52 |
-
parser.add_argument("--decoder_tile_size", type=int, default=64)
|
| 53 |
-
parser.add_argument("--load_8bit_llava", action='store_true', default=False)
|
| 54 |
-
args = parser.parse_args()
|
| 55 |
|
| 56 |
input_image_debug_value = [None]
|
|
|
|
|
|
|
| 57 |
prompt_debug_value = [None]
|
| 58 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 59 |
|
| 60 |
-
|
| 61 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 62 |
|
| 63 |
-
# Load SUPIR
|
| 64 |
-
model, default_setting = create_SUPIR_model(args.opt, SUPIR_sign='Q', load_default_setting=True)
|
| 65 |
-
if args.loading_half_params:
|
| 66 |
-
model = model.half()
|
| 67 |
-
if args.use_tile_vae:
|
| 68 |
-
model.init_tile_vae(encoder_tile_size=args.encoder_tile_size, decoder_tile_size=args.decoder_tile_size)
|
| 69 |
-
model = model.to(SUPIR_device)
|
| 70 |
-
model.first_stage_model.denoise_encoder_s1 = copy.deepcopy(model.first_stage_model.denoise_encoder)
|
| 71 |
-
model.current_model = 'v0-Q'
|
| 72 |
-
ckpt_Q, ckpt_F = load_QF_ckpt(args.opt)
|
| 73 |
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
|
|
|
| 78 |
|
| 79 |
-
|
| 80 |
-
if
|
| 81 |
-
|
| 82 |
-
return seed
|
| 83 |
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
0,
|
| 88 |
-
None,
|
| 89 |
-
None,
|
| 90 |
-
"Cinematic, High Contrast, highly detailed, taken using a Canon EOS R camera, hyper detailed photo - realistic maximum detail, 32k, Color Grading, ultra HD, extreme meticulous detailing, skin pore detailing, hyper sharpness, perfect without deformations.",
|
| 91 |
-
"painting, oil painting, illustration, drawing, art, sketch, anime, cartoon, CG Style, 3D render, unreal engine, blurring, aliasing, pixel, unsharp, weird textures, ugly, dirty, messy, worst quality, low quality, frames, watermark, signature, jpeg artifacts, deformed, lowres, over-smooth",
|
| 92 |
-
1,
|
| 93 |
-
1024,
|
| 94 |
-
1,
|
| 95 |
-
2,
|
| 96 |
-
50,
|
| 97 |
-
-1.0,
|
| 98 |
-
1.,
|
| 99 |
-
default_setting.s_cfg_Quality if torch.cuda.device_count() > 0 else 1.0,
|
| 100 |
-
True,
|
| 101 |
-
random.randint(0, max_64_bit_int),
|
| 102 |
-
5,
|
| 103 |
-
1.003,
|
| 104 |
-
"Wavelet",
|
| 105 |
-
"fp32",
|
| 106 |
-
"fp32",
|
| 107 |
-
1.0,
|
| 108 |
-
True,
|
| 109 |
-
default_setting.spt_linear_CFG_Quality if torch.cuda.device_count() > 0 else 1.0,
|
| 110 |
-
False,
|
| 111 |
-
0.,
|
| 112 |
-
"v0-Q",
|
| 113 |
-
"input",
|
| 114 |
-
179
|
| 115 |
-
]
|
| 116 |
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
| 128 |
-
|
| 129 |
-
|
| 130 |
-
|
| 131 |
-
|
| 132 |
-
return None, None, gr.update(interactive = False)
|
| 133 |
-
torch.cuda.set_device(SUPIR_device)
|
| 134 |
-
LQ = HWC3(np.array(Image.open(input_image)))
|
| 135 |
-
LQ = fix_resize(LQ, 512)
|
| 136 |
-
# stage1
|
| 137 |
-
LQ = np.array(LQ) / 255 * 2 - 1
|
| 138 |
-
LQ = torch.tensor(LQ, dtype=torch.float32).permute(2, 0, 1).unsqueeze(0).to(SUPIR_device)[:, :3, :, :]
|
| 139 |
|
| 140 |
-
|
| 141 |
-
|
|
|
|
| 142 |
|
| 143 |
-
|
| 144 |
-
|
| 145 |
-
|
| 146 |
-
LQ = LQ / 255.0
|
| 147 |
-
LQ = np.power(LQ, gamma_correction)
|
| 148 |
-
LQ *= 255.0
|
| 149 |
-
LQ = LQ.round().clip(0, 255).astype(np.uint8)
|
| 150 |
-
print('<<== stage1_process')
|
| 151 |
-
return LQ, gr.update(visible = True)
|
| 152 |
|
| 153 |
-
def stage2_process_example(*args, **kwargs):
|
| 154 |
-
[result_slider, result_gallery, restore_information, reset_btn, warning, dummy_button] = restore_in_Xmin(*args, **kwargs)
|
| 155 |
-
#outputs_folder = './outputs/'
|
| 156 |
-
outputs_folder = './tmp/'
|
| 157 |
-
os.makedirs(outputs_folder, exist_ok=True)
|
| 158 |
-
output_filename = os.path.join(outputs_folder, datetime.now().strftime("%Y-%m-%d_%H-%M-%S") + '.png')
|
| 159 |
-
print(output_filename)
|
| 160 |
-
iio.imwrite(output_filename, result_slider[1], format="png")
|
| 161 |
-
return [gr.update(visible = True, value=output_filename), warning, dummy_button, gr.skip()]
|
| 162 |
|
| 163 |
-
def
|
| 164 |
-
|
| 165 |
-
|
| 166 |
-
|
| 167 |
-
|
| 168 |
-
|
| 169 |
-
|
| 170 |
-
|
| 171 |
-
|
| 172 |
-
|
| 173 |
-
|
| 174 |
-
|
| 175 |
-
|
| 176 |
-
|
| 177 |
-
|
| 178 |
-
|
| 179 |
-
|
| 180 |
-
|
| 181 |
-
|
| 182 |
-
|
| 183 |
-
|
| 184 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 185 |
prompt,
|
| 186 |
-
|
| 187 |
-
|
| 188 |
-
|
| 189 |
-
|
| 190 |
-
|
| 191 |
-
|
| 192 |
-
|
| 193 |
-
|
| 194 |
-
s_stage2,
|
| 195 |
-
s_cfg,
|
| 196 |
-
randomize_seed,
|
| 197 |
-
seed,
|
| 198 |
-
s_churn,
|
| 199 |
-
s_noise,
|
| 200 |
-
color_fix_type,
|
| 201 |
-
diff_dtype,
|
| 202 |
-
ae_dtype,
|
| 203 |
-
gamma_correction,
|
| 204 |
-
linear_CFG,
|
| 205 |
-
spt_linear_CFG,
|
| 206 |
-
linear_s_stage2,
|
| 207 |
-
spt_linear_s_stage2,
|
| 208 |
-
model_select,
|
| 209 |
-
output_format,
|
| 210 |
-
allocation
|
| 211 |
):
|
| 212 |
-
|
| 213 |
-
|
| 214 |
-
|
| 215 |
-
|
| 216 |
-
|
| 217 |
-
|
| 218 |
-
|
| 219 |
-
|
| 220 |
-
|
| 221 |
-
|
| 222 |
-
|
| 223 |
-
|
| 224 |
-
|
| 225 |
-
|
| 226 |
-
|
| 227 |
-
|
| 228 |
-
|
| 229 |
-
|
| 230 |
-
|
| 231 |
-
|
| 232 |
-
|
| 233 |
-
|
| 234 |
-
|
| 235 |
-
print("linear_s_stage2: " + str(linear_s_stage2))
|
| 236 |
-
print("spt_linear_CFG: " + str(spt_linear_CFG))
|
| 237 |
-
print("spt_linear_s_stage2: " + str(spt_linear_s_stage2))
|
| 238 |
-
print("model_select: " + str(model_select))
|
| 239 |
-
print("GPU time allocation: " + str(allocation) + " min")
|
| 240 |
-
print("output_format: " + str(output_format))
|
| 241 |
-
|
| 242 |
-
if input_image_debug_value[0] is not None or prompt_debug_value[0] is not None or upscale_debug_value[0] is not None:
|
| 243 |
-
denoise_image = noisy_image = input_image_debug_value[0]
|
| 244 |
-
a_prompt = prompt_debug_value[0]
|
| 245 |
-
upscale = upscale_debug_value[0]
|
| 246 |
-
allocation = min(allocation * 60 * 100, 600)
|
| 247 |
-
seed = random.randint(0, max_64_bit_int)
|
| 248 |
-
|
| 249 |
-
input_format = re.sub(r"^.*\.([^\.]+)$", r"\1", noisy_image)
|
| 250 |
-
|
| 251 |
-
if input_format not in ['png', 'webp', 'jpg', 'jpeg', 'gif', 'bmp', 'avif']:
|
| 252 |
-
gr.Warning('Invalid image format. Please first convert into *.png, *.webp, *.jpg, *.jpeg, *.gif, *.bmp, *.heic or *.avif.')
|
| 253 |
-
return None, None, None, None, None, gr.update(interactive = False)
|
| 254 |
-
|
| 255 |
-
if output_format == "input":
|
| 256 |
-
if noisy_image is None:
|
| 257 |
-
output_format = "png"
|
| 258 |
-
else:
|
| 259 |
-
output_format = input_format
|
| 260 |
-
print("final output_format: " + str(output_format))
|
| 261 |
-
|
| 262 |
-
if prompt is None:
|
| 263 |
-
prompt = ""
|
| 264 |
-
|
| 265 |
-
if a_prompt is None:
|
| 266 |
-
a_prompt = ""
|
| 267 |
-
|
| 268 |
-
if n_prompt is None:
|
| 269 |
-
n_prompt = ""
|
| 270 |
-
|
| 271 |
-
if prompt != "" and a_prompt != "":
|
| 272 |
-
a_prompt = prompt + ", " + a_prompt
|
| 273 |
-
else:
|
| 274 |
-
a_prompt = prompt + a_prompt
|
| 275 |
-
print("Final prompt: " + str(a_prompt))
|
| 276 |
-
|
| 277 |
-
denoise_image = np.array(Image.open(noisy_image if denoise_image is None else denoise_image))
|
| 278 |
-
|
| 279 |
-
if rotation == 90:
|
| 280 |
-
denoise_image = np.array(list(zip(*denoise_image[::-1])))
|
| 281 |
-
elif rotation == 180:
|
| 282 |
-
denoise_image = np.array(list(zip(*denoise_image[::-1])))
|
| 283 |
-
denoise_image = np.array(list(zip(*denoise_image[::-1])))
|
| 284 |
-
elif rotation == -90:
|
| 285 |
-
denoise_image = np.array(list(zip(*denoise_image))[::-1])
|
| 286 |
-
|
| 287 |
-
if 1 < downscale:
|
| 288 |
-
input_height, input_width, input_channel = denoise_image.shape
|
| 289 |
-
denoise_image = np.array(Image.fromarray(denoise_image).resize((input_width // downscale, input_height // downscale), Image.LANCZOS))
|
| 290 |
-
|
| 291 |
-
denoise_image = HWC3(denoise_image)
|
| 292 |
-
|
| 293 |
-
if torch.cuda.device_count() == 0:
|
| 294 |
-
gr.Warning('Set this space to GPU config to make it work.')
|
| 295 |
-
return [noisy_image, denoise_image], gr.update(label="Downloadable results in *." + output_format + " format", format = output_format, value = [denoise_image]), None, gr.update(visible=True)
|
| 296 |
-
|
| 297 |
-
if model_select != model.current_model:
|
| 298 |
-
print('load ' + model_select)
|
| 299 |
-
if model_select == 'v0-Q':
|
| 300 |
-
model.load_state_dict(ckpt_Q, strict=False)
|
| 301 |
-
elif model_select == 'v0-F':
|
| 302 |
-
model.load_state_dict(ckpt_F, strict=False)
|
| 303 |
-
model.current_model = model_select
|
| 304 |
-
|
| 305 |
-
model.ae_dtype = convert_dtype(ae_dtype)
|
| 306 |
-
model.model.dtype = convert_dtype(diff_dtype)
|
| 307 |
-
|
| 308 |
-
return restore_on_gpu(
|
| 309 |
-
noisy_image, denoise_image, prompt, a_prompt, n_prompt, num_samples, min_size, downscale, upscale, edm_steps, s_stage1, s_stage2, s_cfg, randomize_seed, seed, s_churn, s_noise, color_fix_type, diff_dtype, ae_dtype, gamma_correction, linear_CFG, linear_s_stage2, spt_linear_CFG, spt_linear_s_stage2, model_select, output_format, allocation
|
| 310 |
-
)
|
| 311 |
|
| 312 |
def get_duration(
|
| 313 |
-
|
| 314 |
-
|
| 315 |
prompt,
|
| 316 |
-
|
| 317 |
-
|
| 318 |
-
|
| 319 |
-
|
| 320 |
-
|
| 321 |
-
upscale,
|
| 322 |
-
edm_steps,
|
| 323 |
-
s_stage1,
|
| 324 |
-
s_stage2,
|
| 325 |
-
s_cfg,
|
| 326 |
-
randomize_seed,
|
| 327 |
seed,
|
| 328 |
-
|
| 329 |
-
|
| 330 |
-
|
| 331 |
-
diff_dtype,
|
| 332 |
-
ae_dtype,
|
| 333 |
-
gamma_correction,
|
| 334 |
-
linear_CFG,
|
| 335 |
-
spt_linear_CFG,
|
| 336 |
-
linear_s_stage2,
|
| 337 |
-
spt_linear_s_stage2,
|
| 338 |
-
model_select,
|
| 339 |
-
output_format,
|
| 340 |
-
allocation
|
| 341 |
):
|
| 342 |
-
return
|
| 343 |
|
| 344 |
@spaces.GPU(duration=get_duration)
|
| 345 |
-
def
|
| 346 |
-
|
| 347 |
-
|
| 348 |
prompt,
|
| 349 |
-
|
| 350 |
-
|
| 351 |
-
|
| 352 |
-
|
| 353 |
-
|
| 354 |
-
upscale,
|
| 355 |
-
edm_steps,
|
| 356 |
-
s_stage1,
|
| 357 |
-
s_stage2,
|
| 358 |
-
s_cfg,
|
| 359 |
-
randomize_seed,
|
| 360 |
seed,
|
| 361 |
-
|
| 362 |
-
|
| 363 |
-
|
| 364 |
-
diff_dtype,
|
| 365 |
-
ae_dtype,
|
| 366 |
-
gamma_correction,
|
| 367 |
-
linear_CFG,
|
| 368 |
-
spt_linear_CFG,
|
| 369 |
-
linear_s_stage2,
|
| 370 |
-
spt_linear_s_stage2,
|
| 371 |
-
model_select,
|
| 372 |
-
output_format,
|
| 373 |
-
allocation
|
| 374 |
):
|
| 375 |
-
start = time.time()
|
| 376 |
-
print('restore ==>>')
|
| 377 |
-
|
| 378 |
-
torch.cuda.set_device(SUPIR_device)
|
| 379 |
-
|
| 380 |
-
with torch.no_grad():
|
| 381 |
-
input_image = upscale_image(input_image, upscale, unit_resolution=32, min_size=min_size)
|
| 382 |
-
LQ = np.array(input_image) / 255.0
|
| 383 |
-
LQ = np.power(LQ, gamma_correction)
|
| 384 |
-
LQ *= 255.0
|
| 385 |
-
LQ = LQ.round().clip(0, 255).astype(np.uint8)
|
| 386 |
-
LQ = LQ / 255 * 2 - 1
|
| 387 |
-
LQ = torch.tensor(LQ, dtype=torch.float32).permute(2, 0, 1).unsqueeze(0).to(SUPIR_device)[:, :3, :, :]
|
| 388 |
-
captions = ['']
|
| 389 |
-
|
| 390 |
-
samples = model.batchify_sample(LQ, captions, num_steps=edm_steps, restoration_scale=s_stage1, s_churn=s_churn,
|
| 391 |
-
s_noise=s_noise, cfg_scale=s_cfg, control_scale=s_stage2, seed=seed,
|
| 392 |
-
num_samples=num_samples, p_p=a_prompt, n_p=n_prompt, color_fix_type=color_fix_type,
|
| 393 |
-
use_linear_CFG=linear_CFG, use_linear_control_scale=linear_s_stage2,
|
| 394 |
-
cfg_scale_start=spt_linear_CFG, control_scale_start=spt_linear_s_stage2)
|
| 395 |
-
|
| 396 |
-
x_samples = (einops.rearrange(samples, 'b c h w -> b h w c') * 127.5 + 127.5).cpu().numpy().round().clip(
|
| 397 |
-
0, 255).astype(np.uint8)
|
| 398 |
-
results = [x_samples[i] for i in range(num_samples)]
|
| 399 |
-
torch.cuda.empty_cache()
|
| 400 |
-
|
| 401 |
-
# All the results have the same size
|
| 402 |
-
input_height, input_width, input_channel = np.array(input_image).shape
|
| 403 |
-
result_height, result_width, result_channel = np.array(results[0]).shape
|
| 404 |
-
|
| 405 |
-
print('<<== restore')
|
| 406 |
-
end = time.time()
|
| 407 |
-
secondes = int(end - start)
|
| 408 |
-
minutes = math.floor(secondes / 60)
|
| 409 |
-
secondes = secondes - (minutes * 60)
|
| 410 |
-
hours = math.floor(minutes / 60)
|
| 411 |
-
minutes = minutes - (hours * 60)
|
| 412 |
-
information = ("Start the process again if you want a different result. " if randomize_seed else "") + \
|
| 413 |
-
"If you don't get the image you wanted, add more details in the « Image description ». " + \
|
| 414 |
-
"The image" + (" has" if len(results) == 1 else "s have") + " been generated in " + \
|
| 415 |
-
((str(hours) + " h, ") if hours != 0 else "") + \
|
| 416 |
-
((str(minutes) + " min, ") if hours != 0 or minutes != 0 else "") + \
|
| 417 |
-
str(secondes) + " sec. " + \
|
| 418 |
-
"The new image resolution is " + str(result_width) + \
|
| 419 |
-
" pixels large and " + str(result_height) + \
|
| 420 |
-
" pixels high, so a resolution of " + f'{result_width * result_height:,}' + " pixels."
|
| 421 |
-
print(information)
|
| 422 |
-
try:
|
| 423 |
-
print("Initial resolution: " + f'{input_width * input_height:,}')
|
| 424 |
-
print("Final resolution: " + f'{result_width * result_height:,}')
|
| 425 |
-
print("edm_steps: " + str(edm_steps))
|
| 426 |
-
print("num_samples: " + str(num_samples))
|
| 427 |
-
print("downscale: " + str(downscale))
|
| 428 |
-
print("Estimated minutes: " + f'{(((result_width * result_height**(1/1.75)) * input_width * input_height * (edm_steps**(1/2)) * (num_samples**(1/2.5)))**(1/2.5)) / 25000:,}')
|
| 429 |
-
except Exception as e:
|
| 430 |
-
print('Exception of Estimation')
|
| 431 |
-
|
| 432 |
-
# Only one image can be shown in the slider
|
| 433 |
-
return [noisy_image] + [results[0]], gr.update(label="Downloadable results in *." + output_format + " format", format = output_format, value = results), gr.update(value = information, visible = True), gr.update(visible=True), gr.update(visible=False), gr.update(interactive = False)
|
| 434 |
-
|
| 435 |
-
def load_and_reset(param_setting):
|
| 436 |
-
print('load_and_reset ==>>')
|
| 437 |
-
if torch.cuda.device_count() == 0:
|
| 438 |
-
gr.Warning('Set this space to GPU config to make it work.')
|
| 439 |
-
return None, None, None, None, None, None, None, None, None, None, None, None, None, None
|
| 440 |
-
edm_steps = default_setting.edm_steps
|
| 441 |
-
s_stage2 = 1.0
|
| 442 |
-
s_stage1 = -1.0
|
| 443 |
-
s_churn = 5
|
| 444 |
-
s_noise = 1.003
|
| 445 |
-
a_prompt = 'Cinematic, High Contrast, highly detailed, taken using a Canon EOS R camera, hyper detailed photo - ' \
|
| 446 |
-
'realistic maximum detail, 32k, Color Grading, ultra HD, extreme meticulous detailing, skin pore ' \
|
| 447 |
-
'detailing, hyper sharpness, perfect without deformations.'
|
| 448 |
-
n_prompt = 'painting, oil painting, illustration, drawing, art, sketch, anime, cartoon, CG Style, ' \
|
| 449 |
-
'3D render, unreal engine, blurring, dirty, messy, worst quality, low quality, frames, watermark, ' \
|
| 450 |
-
'signature, jpeg artifacts, deformed, lowres, over-smooth'
|
| 451 |
-
color_fix_type = 'Wavelet'
|
| 452 |
-
spt_linear_s_stage2 = 0.0
|
| 453 |
-
linear_s_stage2 = False
|
| 454 |
-
linear_CFG = True
|
| 455 |
-
if param_setting == "Quality":
|
| 456 |
-
s_cfg = default_setting.s_cfg_Quality
|
| 457 |
-
spt_linear_CFG = default_setting.spt_linear_CFG_Quality
|
| 458 |
-
model_select = "v0-Q"
|
| 459 |
-
elif param_setting == "Fidelity":
|
| 460 |
-
s_cfg = default_setting.s_cfg_Fidelity
|
| 461 |
-
spt_linear_CFG = default_setting.spt_linear_CFG_Fidelity
|
| 462 |
-
model_select = "v0-F"
|
| 463 |
-
else:
|
| 464 |
-
raise NotImplementedError
|
| 465 |
-
gr.Info('The parameters are reset.')
|
| 466 |
-
print('<<== load_and_reset')
|
| 467 |
-
return edm_steps, s_cfg, s_stage2, s_stage1, s_churn, s_noise, a_prompt, n_prompt, color_fix_type, linear_CFG, \
|
| 468 |
-
spt_linear_CFG, linear_s_stage2, spt_linear_s_stage2, model_select
|
| 469 |
-
|
| 470 |
-
def log_information(result_gallery):
|
| 471 |
-
print('log_information')
|
| 472 |
-
if result_gallery is not None:
|
| 473 |
-
for i, result in enumerate(result_gallery):
|
| 474 |
-
print(result[0])
|
| 475 |
-
|
| 476 |
-
def on_select_result(result_slider, result_gallery, evt: gr.SelectData):
|
| 477 |
-
print('on_select_result')
|
| 478 |
-
if result_gallery is not None:
|
| 479 |
-
for i, result in enumerate(result_gallery):
|
| 480 |
-
print(result[0])
|
| 481 |
-
return [result_slider[0], result_gallery[evt.index][0]]
|
| 482 |
-
|
| 483 |
-
def on_render_image_example(result_example):
|
| 484 |
-
print('on_render_image_example')
|
| 485 |
-
return gr.update(value = result_example, visible = True)
|
| 486 |
-
|
| 487 |
-
title_html = """
|
| 488 |
-
<h1><center>SUPIR</center></h1>
|
| 489 |
-
<big><center>Upscale your images up to x10 freely, without account, without watermark and download it</center></big>
|
| 490 |
-
<center><big><big>🤸<big><big><big><big><big><big>🤸</big></big></big></big></big></big></big></big></center>
|
| 491 |
-
|
| 492 |
-
<p>This is an online demo of SUPIR, a practicing model scaling for photo-realistic image restoration.
|
| 493 |
-
The content added by SUPIR is <b><u>imagination, not real-world information</u></b>.
|
| 494 |
-
SUPIR is for beauty and illustration only.
|
| 495 |
-
Most of the processes last few minutes.
|
| 496 |
-
If you want to upscale AI-generated images, be noticed that <i>PixArt Sigma</i> space can directly generate 5984x5984 images.
|
| 497 |
-
Due to Gradio issues, the generated image is slightly less satured than the original.
|
| 498 |
-
Please leave a <a href="https://huggingface.co/spaces/Fabrice-TIERCELIN/SUPIR/discussions/new">message in discussion</a> if you encounter issues.
|
| 499 |
-
You can also use <a href="https://huggingface.co/spaces/gokaygokay/AuraSR">AuraSR</a> to upscale x4.
|
| 500 |
-
|
| 501 |
-
<p><center><a href="https://arxiv.org/abs/2401.13627">Paper</a>   <a href="http://supir.xpixel.group/">Project Page</a>   <a href="https://huggingface.co/blog/MonsterMMORPG/supir-sota-image-upscale-better-than-magnific-ai">Local Install Guide</a></center></p>
|
| 502 |
-
<p><center><a style="display:inline-block" href='https://github.com/Fanghua-Yu/SUPIR'><img alt="GitHub Repo stars" src="https://img.shields.io/github/stars/Fanghua-Yu/SUPIR?style=social"></a></center></p>
|
| 503 |
"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 504 |
|
|
|
|
| 505 |
|
| 506 |
-
|
| 507 |
-
|
| 508 |
-
The images are not stored but the logs are saved during a month.
|
| 509 |
-
## **How to get SUPIR**
|
| 510 |
-
You can get SUPIR on HuggingFace by [duplicating this space](https://huggingface.co/spaces/Fabrice-TIERCELIN/SUPIR?duplicate=true) and set GPU.
|
| 511 |
-
You can also install SUPIR on your computer following [this tutorial](https://huggingface.co/blog/MonsterMMORPG/supir-sota-image-upscale-better-than-magnific-ai).
|
| 512 |
-
You can install _Pinokio_ on your computer and then install _SUPIR_ into it. It should be quite easy if you have an Nvidia GPU.
|
| 513 |
-
## **Terms of use**
|
| 514 |
-
By using this service, users are required to agree to the following terms: The service is a research preview intended for non-commercial use only. It only provides limited safety measures and may generate offensive content. It must not be used for any illegal, harmful, violent, racist, or sexual purposes. The service may collect user dialogue data for future research. Please submit a feedback to us if you get any inappropriate answer! We will collect those to keep improving our models. For an optimal experience, please use desktop computers for this demo, as mobile devices may compromise its quality.
|
| 515 |
-
## **License**
|
| 516 |
-
The service is a research preview intended for non-commercial use only, subject to the model [License](https://github.com/Fanghua-Yu/SUPIR) of SUPIR.
|
| 517 |
-
"""
|
| 518 |
-
|
| 519 |
-
js = """
|
| 520 |
-
function createGradioAnimation() {
|
| 521 |
-
window.addEventListener("beforeunload", function(e) {
|
| 522 |
-
if (document.getElementById('dummy_button_id') && !document.getElementById('dummy_button_id').disabled) {
|
| 523 |
-
var confirmationMessage = 'A process is still running. '
|
| 524 |
-
+ 'If you leave before saving, your changes will be lost.';
|
| 525 |
|
| 526 |
-
|
| 527 |
-
|
| 528 |
-
return confirmationMessage;
|
| 529 |
-
});
|
| 530 |
-
return 'Animation created';
|
| 531 |
-
}
|
| 532 |
-
"""
|
| 533 |
-
|
| 534 |
-
# Gradio interface
|
| 535 |
-
with gr.Blocks(js=js) as interface:
|
| 536 |
-
if torch.cuda.device_count() == 0:
|
| 537 |
-
with gr.Row():
|
| 538 |
-
gr.HTML("""
|
| 539 |
-
<p style="background-color: red;"><big><big><big><b>⚠️To use SUPIR, <a href="https://huggingface.co/spaces/Fabrice-TIERCELIN/SUPIR?duplicate=true">duplicate this space</a> and set a GPU with 30 GB VRAM.</b>
|
| 540 |
|
| 541 |
-
|
| 542 |
-
|
| 543 |
-
|
| 544 |
-
|
| 545 |
-
|
| 546 |
-
|
| 547 |
-
|
| 548 |
-
|
| 549 |
-
|
| 550 |
-
|
| 551 |
-
|
| 552 |
-
|
| 553 |
-
|
| 554 |
-
|
| 555 |
-
|
| 556 |
-
|
| 557 |
-
|
| 558 |
-
|
| 559 |
-
|
| 560 |
-
|
| 561 |
-
|
| 562 |
-
|
| 563 |
-
|
| 564 |
-
|
| 565 |
-
|
| 566 |
-
|
| 567 |
-
|
| 568 |
-
|
| 569 |
-
|
| 570 |
-
|
| 571 |
-
|
| 572 |
-
|
| 573 |
-
|
| 574 |
-
|
| 575 |
-
|
| 576 |
-
|
| 577 |
-
|
| 578 |
-
|
| 579 |
-
|
| 580 |
-
|
| 581 |
-
with gr.
|
| 582 |
-
with gr.
|
| 583 |
-
|
| 584 |
-
|
| 585 |
-
|
| 586 |
-
|
| 587 |
-
|
| 588 |
-
|
| 589 |
-
|
| 590 |
-
|
| 591 |
-
|
| 592 |
-
|
| 593 |
-
|
| 594 |
-
|
| 595 |
-
|
| 596 |
-
|
| 597 |
-
|
| 598 |
-
|
| 599 |
-
|
| 600 |
-
|
| 601 |
-
|
| 602 |
-
|
| 603 |
-
|
| 604 |
-
|
| 605 |
-
|
| 606 |
-
|
| 607 |
-
|
| 608 |
-
|
| 609 |
-
|
| 610 |
-
|
| 611 |
-
|
| 612 |
-
|
| 613 |
-
|
| 614 |
-
|
| 615 |
-
|
| 616 |
-
|
| 617 |
-
|
| 618 |
-
|
| 619 |
-
|
| 620 |
-
|
| 621 |
-
warning = gr.HTML(elem_id="warning", value = "<center><big>Your computer must <u>not</u> enter into standby mode.</big><br/>On Chrome, you can force to keep a tab alive in <code>chrome://discards/</code></center>", visible = False)
|
| 622 |
-
restore_information = gr.HTML(value = "Restart the process to get another result.", visible = False)
|
| 623 |
-
result_slider = ImageSlider(label = 'Comparator', show_label = False, interactive = False, elem_id = "slider1", show_download_button = False, visible = False)
|
| 624 |
-
result_gallery = gr.Gallery(label = 'Downloadable results', show_label = True, interactive = False, elem_id = "gallery1")
|
| 625 |
-
result_example = gr.HTML(elem_id="result_example", visible = False)
|
| 626 |
-
result_image_example = gr.Image(label="Example Image", visible = False)
|
| 627 |
-
|
| 628 |
-
with gr.Row(elem_id="examples", visible = False):
|
| 629 |
-
gr.Examples(
|
| 630 |
-
label = "Examples for cache",
|
| 631 |
-
examples = [
|
| 632 |
-
[
|
| 633 |
-
"./Examples/Example2.jpeg",
|
| 634 |
-
0,
|
| 635 |
-
"./Examples/Example2.jpeg",
|
| 636 |
-
"La cabeza de un gato atigrado, en una casa, fotorrealista, 8k, extremadamente detallada",
|
| 637 |
-
"Cinematic, High Contrast, highly detailed, taken using a Canon EOS R camera, hyper detailed photo - realistic maximum detail, 32k, Color Grading, ultra HD, extreme meticulous detailing, skin pore detailing, hyper sharpness, perfect without deformations.",
|
| 638 |
-
"painting, oil painting, illustration, drawing, art, sketch, anime, cartoon, CG Style, 3D render, unreal engine, blurring, aliasing, pixel, unsharp, weird textures, ugly, dirty, messy, worst quality, low quality, frames, watermark, signature, jpeg artifacts, deformed, lowres, over-smooth",
|
| 639 |
-
1, # num_samples
|
| 640 |
-
32, # min_size
|
| 641 |
-
1, # downscale
|
| 642 |
-
1, # upscale
|
| 643 |
-
100, # edm_steps
|
| 644 |
-
-1, # s_stage1
|
| 645 |
-
1, # s_stage2
|
| 646 |
-
7.5, # s_cfg
|
| 647 |
-
True, # randomize_seed
|
| 648 |
-
42, # seed
|
| 649 |
-
5, # s_churn
|
| 650 |
-
1.003, # s_noise
|
| 651 |
-
"Wavelet", # color_fix_type
|
| 652 |
-
"fp16", # diff_dtype
|
| 653 |
-
"bf16", # ae_dtype
|
| 654 |
-
1.0, # gamma_correction
|
| 655 |
-
True, # linear_CFG
|
| 656 |
-
4, # spt_linear_CFG
|
| 657 |
-
False, # linear_s_stage2
|
| 658 |
-
0., # spt_linear_s_stage2
|
| 659 |
-
"v0-Q", # model_select
|
| 660 |
-
"input", # output_format
|
| 661 |
-
60 # allocation
|
| 662 |
-
],
|
| 663 |
-
[
|
| 664 |
-
"./Examples/Example2.jpeg",
|
| 665 |
-
0,
|
| 666 |
-
"./Examples/Example2.jpeg",
|
| 667 |
-
"La cabeza de un gato atigrado, en una casa, fotorrealista, 8k, extremadamente detallada",
|
| 668 |
-
"Cinematic, High Contrast, highly detailed, taken using a Canon EOS R camera, hyper detailed photo - realistic maximum detail, 32k, Color Grading, ultra HD, extreme meticulous detailing, skin pore detailing, hyper sharpness, perfect without deformations.",
|
| 669 |
-
"painting, oil painting, illustration, drawing, art, sketch, anime, cartoon, CG Style, 3D render, unreal engine, blurring, aliasing, pixel, unsharp, weird textures, ugly, dirty, messy, worst quality, low quality, frames, watermark, signature, jpeg artifacts, deformed, lowres, over-smooth",
|
| 670 |
-
4, # num_samples
|
| 671 |
-
32, # min_size
|
| 672 |
-
1, # downscale
|
| 673 |
-
1, # upscale
|
| 674 |
-
100, # edm_steps
|
| 675 |
-
-1, # s_stage1
|
| 676 |
-
1, # s_stage2
|
| 677 |
-
7.5, # s_cfg
|
| 678 |
-
True, # randomize_seed
|
| 679 |
-
42, # seed
|
| 680 |
-
5, # s_churn
|
| 681 |
-
1.003, # s_noise
|
| 682 |
-
"Wavelet", # color_fix_type
|
| 683 |
-
"fp16", # diff_dtype
|
| 684 |
-
"bf16", # ae_dtype
|
| 685 |
-
1.0, # gamma_correction
|
| 686 |
-
True, # linear_CFG
|
| 687 |
-
4, # spt_linear_CFG
|
| 688 |
-
False, # linear_s_stage2
|
| 689 |
-
0., # spt_linear_s_stage2
|
| 690 |
-
"v0-Q", # model_select
|
| 691 |
-
"input", # output_format
|
| 692 |
-
60 # allocation
|
| 693 |
-
]
|
| 694 |
-
],
|
| 695 |
-
run_on_click = True,
|
| 696 |
-
fn = stage2_process_example,
|
| 697 |
-
inputs = [
|
| 698 |
-
input_image,
|
| 699 |
-
rotation,
|
| 700 |
-
denoise_image,
|
| 701 |
-
prompt,
|
| 702 |
-
a_prompt,
|
| 703 |
-
n_prompt,
|
| 704 |
-
num_samples,
|
| 705 |
-
min_size,
|
| 706 |
-
downscale,
|
| 707 |
-
upscale,
|
| 708 |
-
edm_steps,
|
| 709 |
-
s_stage1,
|
| 710 |
-
s_stage2,
|
| 711 |
-
s_cfg,
|
| 712 |
-
randomize_seed,
|
| 713 |
-
seed,
|
| 714 |
-
s_churn,
|
| 715 |
-
s_noise,
|
| 716 |
-
color_fix_type,
|
| 717 |
-
diff_dtype,
|
| 718 |
-
ae_dtype,
|
| 719 |
-
gamma_correction,
|
| 720 |
-
linear_CFG,
|
| 721 |
-
spt_linear_CFG,
|
| 722 |
-
linear_s_stage2,
|
| 723 |
-
spt_linear_s_stage2,
|
| 724 |
-
model_select,
|
| 725 |
-
output_format,
|
| 726 |
-
allocation
|
| 727 |
-
],
|
| 728 |
-
outputs = [
|
| 729 |
-
result_example,
|
| 730 |
-
warning,
|
| 731 |
-
dummy_button,
|
| 732 |
-
prompt_hint
|
| 733 |
-
],
|
| 734 |
-
cache_examples = True,
|
| 735 |
-
)
|
| 736 |
-
|
| 737 |
-
gr.Examples(
|
| 738 |
-
label = "Examples for demo",
|
| 739 |
-
examples = [
|
| 740 |
-
[
|
| 741 |
-
"./Examples/Example1.png",
|
| 742 |
-
0,
|
| 743 |
-
"./Examples/Example1.png",
|
| 744 |
-
"Group of people, walking, happy, in the street, photorealistic, 8k, extremely detailled",
|
| 745 |
-
"Cinematic, High Contrast, highly detailed, taken using a Canon EOS R camera, hyper detailed photo - realistic maximum detail, 32k, Color Grading, ultra HD, extreme meticulous detailing, skin pore detailing, hyper sharpness, perfect without deformations.",
|
| 746 |
-
"painting, oil painting, illustration, drawing, art, sketch, anime, cartoon, CG Style, 3D render, unreal engine, blurring, aliasing, pixel, unsharp, weird textures, ugly, dirty, messy, worst quality, low quality, frames, watermark, signature, jpeg artifacts, deformed, lowres, over-smooth",
|
| 747 |
-
2, # num_samples
|
| 748 |
-
1024, # min_size
|
| 749 |
-
1, # downscale
|
| 750 |
-
8, # upscale
|
| 751 |
-
100, # edm_steps
|
| 752 |
-
-1, # s_stage1
|
| 753 |
-
1, # s_stage2
|
| 754 |
-
7.5, # s_cfg
|
| 755 |
-
False, # randomize_seed
|
| 756 |
-
42, # seed
|
| 757 |
-
5, # s_churn
|
| 758 |
-
1.003, # s_noise
|
| 759 |
-
"AdaIn", # color_fix_type
|
| 760 |
-
"fp16", # diff_dtype
|
| 761 |
-
"bf16", # ae_dtype
|
| 762 |
-
1.0, # gamma_correction
|
| 763 |
-
True, # linear_CFG
|
| 764 |
-
4, # spt_linear_CFG
|
| 765 |
-
False, # linear_s_stage2
|
| 766 |
-
0., # spt_linear_s_stage2
|
| 767 |
-
"v0-Q", # model_select
|
| 768 |
-
"input", # output_format
|
| 769 |
-
180 # allocation
|
| 770 |
-
],
|
| 771 |
-
[
|
| 772 |
-
"./Examples/Example2.jpeg",
|
| 773 |
-
0,
|
| 774 |
-
"./Examples/Example2.jpeg",
|
| 775 |
-
"La cabeza de un gato atigrado, en una casa, fotorrealista, 8k, extremadamente detallada",
|
| 776 |
-
"Cinematic, High Contrast, highly detailed, taken using a Canon EOS R camera, hyper detailed photo - realistic maximum detail, 32k, Color Grading, ultra HD, extreme meticulous detailing, skin pore detailing, hyper sharpness, perfect without deformations.",
|
| 777 |
-
"painting, oil painting, illustration, drawing, art, sketch, anime, cartoon, CG Style, 3D render, unreal engine, blurring, aliasing, pixel, unsharp, weird textures, ugly, dirty, messy, worst quality, low quality, frames, watermark, signature, jpeg artifacts, deformed, lowres, over-smooth",
|
| 778 |
-
1, # num_samples
|
| 779 |
-
1024, # min_size
|
| 780 |
-
1, # downscale
|
| 781 |
-
1, # upscale
|
| 782 |
-
100, # edm_steps
|
| 783 |
-
-1, # s_stage1
|
| 784 |
-
1, # s_stage2
|
| 785 |
-
7.5, # s_cfg
|
| 786 |
-
False, # randomize_seed
|
| 787 |
-
42, # seed
|
| 788 |
-
5, # s_churn
|
| 789 |
-
1.003, # s_noise
|
| 790 |
-
"Wavelet", # color_fix_type
|
| 791 |
-
"fp16", # diff_dtype
|
| 792 |
-
"bf16", # ae_dtype
|
| 793 |
-
1.0, # gamma_correction
|
| 794 |
-
True, # linear_CFG
|
| 795 |
-
4, # spt_linear_CFG
|
| 796 |
-
False, # linear_s_stage2
|
| 797 |
-
0., # spt_linear_s_stage2
|
| 798 |
-
"v0-Q", # model_select
|
| 799 |
-
"input", # output_format
|
| 800 |
-
60 # allocation
|
| 801 |
-
],
|
| 802 |
-
[
|
| 803 |
-
"./Examples/Example3.webp",
|
| 804 |
-
0,
|
| 805 |
-
"./Examples/Example3.webp",
|
| 806 |
-
"A red apple",
|
| 807 |
-
"Cinematic, High Contrast, highly detailed, taken using a Canon EOS R camera, hyper detailed photo - realistic maximum detail, 32k, Color Grading, ultra HD, extreme meticulous detailing, skin pore detailing, hyper sharpness, perfect without deformations.",
|
| 808 |
-
"painting, oil painting, illustration, drawing, art, sketch, anime, cartoon, CG Style, 3D render, unreal engine, blurring, aliasing, pixel, unsharp, weird textures, ugly, dirty, messy, worst quality, low quality, frames, watermark, signature, jpeg artifacts, deformed, lowres, over-smooth",
|
| 809 |
-
1, # num_samples
|
| 810 |
-
1024, # min_size
|
| 811 |
-
1, # downscale
|
| 812 |
-
1, # upscale
|
| 813 |
-
200, # edm_steps
|
| 814 |
-
-1, # s_stage1
|
| 815 |
-
1, # s_stage2
|
| 816 |
-
7.5, # s_cfg
|
| 817 |
-
False, # randomize_seed
|
| 818 |
-
42, # seed
|
| 819 |
-
5, # s_churn
|
| 820 |
-
1.003, # s_noise
|
| 821 |
-
"Wavelet", # color_fix_type
|
| 822 |
-
"fp16", # diff_dtype
|
| 823 |
-
"bf16", # ae_dtype
|
| 824 |
-
1.0, # gamma_correction
|
| 825 |
-
True, # linear_CFG
|
| 826 |
-
4, # spt_linear_CFG
|
| 827 |
-
False, # linear_s_stage2
|
| 828 |
-
0., # spt_linear_s_stage2
|
| 829 |
-
"v0-Q", # model_select
|
| 830 |
-
"input", # output_format
|
| 831 |
-
180 # allocation
|
| 832 |
-
],
|
| 833 |
-
[
|
| 834 |
-
"./Examples/Example3.webp",
|
| 835 |
-
0,
|
| 836 |
-
"./Examples/Example3.webp",
|
| 837 |
-
"A red marble",
|
| 838 |
-
"Cinematic, High Contrast, highly detailed, taken using a Canon EOS R camera, hyper detailed photo - realistic maximum detail, 32k, Color Grading, ultra HD, extreme meticulous detailing, skin pore detailing, hyper sharpness, perfect without deformations.",
|
| 839 |
-
"painting, oil painting, illustration, drawing, art, sketch, anime, cartoon, CG Style, 3D render, unreal engine, blurring, aliasing, pixel, unsharp, weird textures, ugly, dirty, messy, worst quality, low quality, frames, watermark, signature, jpeg artifacts, deformed, lowres, over-smooth",
|
| 840 |
-
1, # num_samples
|
| 841 |
-
1024, # min_size
|
| 842 |
-
1, # downscale
|
| 843 |
-
1, # upscale
|
| 844 |
-
200, # edm_steps
|
| 845 |
-
-1, # s_stage1
|
| 846 |
-
1, # s_stage2
|
| 847 |
-
7.5, # s_cfg
|
| 848 |
-
False, # randomize_seed
|
| 849 |
-
42, # seed
|
| 850 |
-
5, # s_churn
|
| 851 |
-
1.003, # s_noise
|
| 852 |
-
"Wavelet", # color_fix_type
|
| 853 |
-
"fp16", # diff_dtype
|
| 854 |
-
"bf16", # ae_dtype
|
| 855 |
-
1.0, # gamma_correction
|
| 856 |
-
True, # linear_CFG
|
| 857 |
-
4, # spt_linear_CFG
|
| 858 |
-
False, # linear_s_stage2
|
| 859 |
-
0., # spt_linear_s_stage2
|
| 860 |
-
"v0-Q", # model_select
|
| 861 |
-
"input", # output_format
|
| 862 |
-
180 # allocation
|
| 863 |
-
],
|
| 864 |
-
],
|
| 865 |
-
run_on_click = True,
|
| 866 |
-
fn = stage2_process,
|
| 867 |
-
inputs = [
|
| 868 |
-
input_image,
|
| 869 |
-
rotation,
|
| 870 |
-
denoise_image,
|
| 871 |
-
prompt,
|
| 872 |
-
a_prompt,
|
| 873 |
-
n_prompt,
|
| 874 |
-
num_samples,
|
| 875 |
-
min_size,
|
| 876 |
-
downscale,
|
| 877 |
-
upscale,
|
| 878 |
-
edm_steps,
|
| 879 |
-
s_stage1,
|
| 880 |
-
s_stage2,
|
| 881 |
-
s_cfg,
|
| 882 |
-
randomize_seed,
|
| 883 |
-
seed,
|
| 884 |
-
s_churn,
|
| 885 |
-
s_noise,
|
| 886 |
-
color_fix_type,
|
| 887 |
-
diff_dtype,
|
| 888 |
-
ae_dtype,
|
| 889 |
-
gamma_correction,
|
| 890 |
-
linear_CFG,
|
| 891 |
-
spt_linear_CFG,
|
| 892 |
-
linear_s_stage2,
|
| 893 |
-
spt_linear_s_stage2,
|
| 894 |
-
model_select,
|
| 895 |
-
output_format,
|
| 896 |
-
allocation
|
| 897 |
-
],
|
| 898 |
-
outputs = [
|
| 899 |
-
result_slider,
|
| 900 |
-
result_gallery,
|
| 901 |
-
restore_information,
|
| 902 |
-
reset_btn,
|
| 903 |
-
warning,
|
| 904 |
-
dummy_button
|
| 905 |
-
],
|
| 906 |
-
cache_examples = False,
|
| 907 |
-
)
|
| 908 |
-
|
| 909 |
-
with gr.Row():
|
| 910 |
-
gr.Markdown(claim_md)
|
| 911 |
-
|
| 912 |
-
input_image.upload(fn = check_upload, inputs = [
|
| 913 |
-
input_image
|
| 914 |
-
], outputs = [
|
| 915 |
-
rotation
|
| 916 |
-
], queue = False, show_progress = False)
|
| 917 |
-
|
| 918 |
-
denoise_button.click(fn = check_and_update, inputs = [
|
| 919 |
-
input_image
|
| 920 |
-
], outputs = [warning, dummy_button], queue = False, show_progress = False).success(fn = stage1_process, inputs = [
|
| 921 |
-
input_image,
|
| 922 |
-
gamma_correction,
|
| 923 |
-
diff_dtype,
|
| 924 |
-
ae_dtype
|
| 925 |
-
], outputs=[
|
| 926 |
-
denoise_image,
|
| 927 |
-
denoise_information,
|
| 928 |
-
dummy_button
|
| 929 |
-
])
|
| 930 |
-
|
| 931 |
-
diffusion_button.click(fn = update_seed, inputs = [
|
| 932 |
-
randomize_seed,
|
| 933 |
-
seed
|
| 934 |
-
], outputs = [
|
| 935 |
-
seed
|
| 936 |
-
], queue = False, show_progress = False).then(fn = check_and_update, inputs = [
|
| 937 |
-
input_image
|
| 938 |
-
], outputs = [warning, dummy_button], queue = False, show_progress = False).success(fn=stage2_process, inputs = [
|
| 939 |
-
input_image,
|
| 940 |
-
rotation,
|
| 941 |
-
denoise_image,
|
| 942 |
prompt,
|
| 943 |
-
|
| 944 |
-
|
| 945 |
-
|
| 946 |
-
|
| 947 |
-
|
| 948 |
-
|
| 949 |
-
|
| 950 |
-
|
| 951 |
-
|
| 952 |
-
s_cfg,
|
| 953 |
-
randomize_seed,
|
| 954 |
-
seed,
|
| 955 |
-
s_churn,
|
| 956 |
-
s_noise,
|
| 957 |
-
color_fix_type,
|
| 958 |
-
diff_dtype,
|
| 959 |
-
ae_dtype,
|
| 960 |
-
gamma_correction,
|
| 961 |
-
linear_CFG,
|
| 962 |
-
spt_linear_CFG,
|
| 963 |
-
linear_s_stage2,
|
| 964 |
-
spt_linear_s_stage2,
|
| 965 |
-
model_select,
|
| 966 |
-
output_format,
|
| 967 |
-
allocation
|
| 968 |
-
], outputs = [
|
| 969 |
-
result_slider,
|
| 970 |
-
result_gallery,
|
| 971 |
-
restore_information,
|
| 972 |
-
reset_btn,
|
| 973 |
-
warning,
|
| 974 |
-
dummy_button
|
| 975 |
-
]).success(fn = log_information, inputs = [
|
| 976 |
-
result_gallery
|
| 977 |
-
], outputs = [], queue = False, show_progress = False)
|
| 978 |
-
|
| 979 |
-
result_gallery.change(on_select_result, [result_slider, result_gallery], result_slider)
|
| 980 |
-
result_gallery.select(on_select_result, [result_slider, result_gallery], result_slider)
|
| 981 |
-
result_example.change(on_render_image_example, result_example, result_image_example)
|
| 982 |
|
| 983 |
-
|
| 984 |
-
|
| 985 |
-
|
| 986 |
-
|
| 987 |
-
|
| 988 |
-
s_stage2,
|
| 989 |
-
s_stage1,
|
| 990 |
-
s_churn,
|
| 991 |
-
s_noise,
|
| 992 |
-
a_prompt,
|
| 993 |
-
n_prompt,
|
| 994 |
-
color_fix_type,
|
| 995 |
-
linear_CFG,
|
| 996 |
-
spt_linear_CFG,
|
| 997 |
-
linear_s_stage2,
|
| 998 |
-
spt_linear_s_stage2,
|
| 999 |
-
model_select
|
| 1000 |
-
])
|
| 1001 |
|
| 1002 |
-
|
| 1003 |
-
|
| 1004 |
-
|
| 1005 |
-
|
| 1006 |
-
|
| 1007 |
-
|
| 1008 |
-
|
| 1009 |
-
|
| 1010 |
-
|
| 1011 |
-
|
| 1012 |
-
|
| 1013 |
-
|
| 1014 |
-
|
| 1015 |
-
s_stage2,
|
| 1016 |
-
s_cfg,
|
| 1017 |
-
randomize_seed,
|
| 1018 |
-
seed,
|
| 1019 |
-
s_churn,
|
| 1020 |
-
s_noise,
|
| 1021 |
-
color_fix_type,
|
| 1022 |
-
diff_dtype,
|
| 1023 |
-
ae_dtype,
|
| 1024 |
-
gamma_correction,
|
| 1025 |
-
linear_CFG,
|
| 1026 |
-
spt_linear_CFG,
|
| 1027 |
-
linear_s_stage2,
|
| 1028 |
-
spt_linear_s_stage2,
|
| 1029 |
-
model_select,
|
| 1030 |
-
output_format,
|
| 1031 |
-
allocation
|
| 1032 |
-
], queue = False, show_progress = False)
|
| 1033 |
|
| 1034 |
-
def handle_field_debug_change(input_image_debug_data, prompt_debug_data,
|
|
|
|
| 1035 |
input_image_debug_value[0] = input_image_debug_data
|
|
|
|
|
|
|
| 1036 |
prompt_debug_value[0] = prompt_debug_data
|
| 1037 |
-
|
| 1038 |
return []
|
| 1039 |
|
| 1040 |
input_image_debug.upload(
|
| 1041 |
fn=handle_field_debug_change,
|
| 1042 |
-
inputs=[input_image_debug, prompt_debug,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1043 |
outputs=[]
|
| 1044 |
)
|
| 1045 |
|
| 1046 |
prompt_debug.change(
|
| 1047 |
fn=handle_field_debug_change,
|
| 1048 |
-
inputs=[input_image_debug, prompt_debug,
|
| 1049 |
outputs=[]
|
| 1050 |
)
|
| 1051 |
|
| 1052 |
-
|
| 1053 |
fn=handle_field_debug_change,
|
| 1054 |
-
inputs=[input_image_debug, prompt_debug,
|
| 1055 |
outputs=[]
|
| 1056 |
)
|
| 1057 |
-
|
| 1058 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
import os
|
| 2 |
+
# PyTorch 2.8 (temporary hack)
|
| 3 |
+
os.system('pip install --upgrade --pre --extra-index-url https://download.pytorch.org/whl/nightly/cu126 "torch<2.9" spaces')
|
| 4 |
+
|
| 5 |
+
# --- 1. Model Download and Setup (Diffusers Backend) ---
|
| 6 |
+
import spaces
|
| 7 |
+
import torch
|
| 8 |
+
from diffusers import FlowMatchEulerDiscreteScheduler
|
| 9 |
+
from diffusers.pipelines.wan.pipeline_wan_i2v import WanImageToVideoPipeline
|
| 10 |
+
from diffusers.models.transformers.transformer_wan import WanTransformer3DModel
|
| 11 |
+
from diffusers.utils.export_utils import export_to_video
|
| 12 |
import gradio as gr
|
| 13 |
+
import tempfile
|
| 14 |
import numpy as np
|
| 15 |
+
from PIL import Image
|
|
|
|
|
|
|
|
|
|
|
|
|
| 16 |
import random
|
| 17 |
+
import gc
|
| 18 |
+
from gradio_client import Client, handle_file # Import for API call
|
| 19 |
|
| 20 |
+
# Import the optimization function from the separate file
|
| 21 |
+
from optimization import optimize_pipeline_
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
|
| 23 |
+
# --- Constants and Model Loading ---
|
| 24 |
+
MODEL_ID = "Wan-AI/Wan2.2-I2V-A14B-Diffusers"
|
| 25 |
|
| 26 |
+
# --- NEW: Flexible Dimension Constants ---
|
| 27 |
+
MAX_DIMENSION = 832
|
| 28 |
+
MIN_DIMENSION = 480
|
| 29 |
+
DIMENSION_MULTIPLE = 16
|
| 30 |
+
SQUARE_SIZE = 480
|
|
|
|
|
|
|
|
|
|
| 31 |
|
| 32 |
+
MAX_SEED = np.iinfo(np.int32).max
|
| 33 |
|
| 34 |
+
FIXED_FPS = 24
|
| 35 |
+
MIN_FRAMES_MODEL = 8
|
| 36 |
+
MAX_FRAMES_MODEL = 81
|
|
|
|
|
|
|
| 37 |
|
| 38 |
+
MIN_DURATION = round(MIN_FRAMES_MODEL/FIXED_FPS, 1)
|
| 39 |
+
MAX_DURATION = round(MAX_FRAMES_MODEL/FIXED_FPS, 1)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 40 |
|
| 41 |
input_image_debug_value = [None]
|
| 42 |
+
input_video_debug_value = [None]
|
| 43 |
+
end_image_debug_value = [None]
|
| 44 |
prompt_debug_value = [None]
|
| 45 |
+
total_second_length_debug_value = [None]
|
| 46 |
+
|
| 47 |
+
default_negative_prompt = "Vibrant colors, overexposure, static, blurred details, subtitles, error, style, artwork, painting, image, still, overall gray, worst quality, low quality, JPEG compression residue, ugly, mutilated, extra fingers, poorly drawn hands, poorly drawn faces, deformed, disfigured, malformed limbs, fused fingers, still image, cluttered background, three legs, many people in the background, walking backwards, overexposure, jumpcut, crossfader, "
|
| 48 |
+
|
| 49 |
+
print("Loading models into memory. This may take a few minutes...")
|
| 50 |
+
|
| 51 |
+
pipe = WanImageToVideoPipeline.from_pretrained(
|
| 52 |
+
MODEL_ID,
|
| 53 |
+
transformer=WanTransformer3DModel.from_pretrained('cbensimon/Wan2.2-I2V-A14B-bf16-Diffusers',
|
| 54 |
+
subfolder='transformer',
|
| 55 |
+
torch_dtype=torch.bfloat16,
|
| 56 |
+
device_map='cuda',
|
| 57 |
+
),
|
| 58 |
+
transformer_2=WanTransformer3DModel.from_pretrained('cbensimon/Wan2.2-I2V-A14B-bf16-Diffusers',
|
| 59 |
+
subfolder='transformer_2',
|
| 60 |
+
torch_dtype=torch.bfloat16,
|
| 61 |
+
device_map='cuda',
|
| 62 |
+
),
|
| 63 |
+
torch_dtype=torch.bfloat16,
|
| 64 |
+
)
|
| 65 |
+
pipe.scheduler = FlowMatchEulerDiscreteScheduler.from_config(pipe.scheduler.config, shift=8.0)
|
| 66 |
+
pipe.to('cuda')
|
| 67 |
+
|
| 68 |
+
|
| 69 |
+
|
| 70 |
+
print("Optimizing pipeline...")
|
| 71 |
+
for i in range(3):
|
| 72 |
+
gc.collect()
|
| 73 |
+
torch.cuda.synchronize()
|
| 74 |
+
torch.cuda.empty_cache()
|
| 75 |
|
| 76 |
+
optimize_pipeline_(pipe,
|
| 77 |
+
image=Image.new('RGB', (MAX_DIMENSION, MIN_DIMENSION)),
|
| 78 |
+
prompt='prompt',
|
| 79 |
+
height=MIN_DIMENSION,
|
| 80 |
+
width=MAX_DIMENSION,
|
| 81 |
+
num_frames=MAX_FRAMES_MODEL,
|
| 82 |
+
)
|
| 83 |
+
print("All models loaded and optimized. Gradio app is ready.")
|
| 84 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 85 |
|
| 86 |
+
# --- 2. Image Processing and Application Logic ---
|
| 87 |
+
def generate_end_frame(start_img, gen_prompt, progress=gr.Progress(track_tqdm=True)):
|
| 88 |
+
"""Calls an external Gradio API to generate an image."""
|
| 89 |
+
if start_img is None:
|
| 90 |
+
raise gr.Error("Please provide a Start Frame first.")
|
| 91 |
|
| 92 |
+
hf_token = os.getenv("HF_TOKEN")
|
| 93 |
+
if not hf_token:
|
| 94 |
+
raise gr.Error("HF_TOKEN not found in environment variables. Please set it in your Space secrets.")
|
|
|
|
| 95 |
|
| 96 |
+
with tempfile.NamedTemporaryFile(suffix=".png", delete=False) as tmpfile:
|
| 97 |
+
start_img.save(tmpfile.name)
|
| 98 |
+
tmp_path = tmpfile.name
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 99 |
|
| 100 |
+
progress(0.1, desc="Connecting to image generation API...")
|
| 101 |
+
client = Client("multimodalart/nano-banana-private")
|
| 102 |
+
|
| 103 |
+
progress(0.5, desc=f"Generating with prompt: '{gen_prompt}'...")
|
| 104 |
+
try:
|
| 105 |
+
result = client.predict(
|
| 106 |
+
prompt=gen_prompt,
|
| 107 |
+
images=[
|
| 108 |
+
{"image": handle_file(tmp_path)}
|
| 109 |
+
],
|
| 110 |
+
manual_token=hf_token,
|
| 111 |
+
api_name="/unified_image_generator"
|
| 112 |
+
)
|
| 113 |
+
finally:
|
| 114 |
+
os.remove(tmp_path)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 115 |
|
| 116 |
+
progress(1.0, desc="Done!")
|
| 117 |
+
print(result)
|
| 118 |
+
return result
|
| 119 |
|
| 120 |
+
def switch_to_upload_tab():
|
| 121 |
+
"""Returns a gr.Tabs update to switch to the first tab."""
|
| 122 |
+
return gr.Tabs(selected="upload_tab")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 123 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 124 |
|
| 125 |
+
def process_image_for_video(image: Image.Image) -> Image.Image:
|
| 126 |
+
"""
|
| 127 |
+
Resizes an image based on the following rules for video generation:
|
| 128 |
+
1. The longest side will be scaled down to MAX_DIMENSION if it's larger.
|
| 129 |
+
2. The shortest side will be scaled up to MIN_DIMENSION if it's smaller.
|
| 130 |
+
3. The final dimensions will be rounded to the nearest multiple of DIMENSION_MULTIPLE.
|
| 131 |
+
4. Square images are resized to a fixed SQUARE_SIZE.
|
| 132 |
+
The aspect ratio is preserved as closely as possible.
|
| 133 |
+
"""
|
| 134 |
+
width, height = image.size
|
| 135 |
+
|
| 136 |
+
# Rule 4: Handle square images
|
| 137 |
+
if width == height:
|
| 138 |
+
return image.resize((SQUARE_SIZE, SQUARE_SIZE), Image.Resampling.LANCZOS)
|
| 139 |
+
|
| 140 |
+
# Determine target dimensions while preserving aspect ratio
|
| 141 |
+
aspect_ratio = width / height
|
| 142 |
+
new_width, new_height = width, height
|
| 143 |
+
|
| 144 |
+
# Rule 1: Scale down if too large
|
| 145 |
+
if new_width > MAX_DIMENSION or new_height > MAX_DIMENSION:
|
| 146 |
+
if aspect_ratio > 1: # Landscape
|
| 147 |
+
scale = MAX_DIMENSION / new_width
|
| 148 |
+
else: # Portrait
|
| 149 |
+
scale = MAX_DIMENSION / new_height
|
| 150 |
+
new_width *= scale
|
| 151 |
+
new_height *= scale
|
| 152 |
+
|
| 153 |
+
# Rule 2: Scale up if too small
|
| 154 |
+
if new_width < MIN_DIMENSION or new_height < MIN_DIMENSION:
|
| 155 |
+
if aspect_ratio > 1: # Landscape
|
| 156 |
+
scale = MIN_DIMENSION / new_height
|
| 157 |
+
else: # Portrait
|
| 158 |
+
scale = MIN_DIMENSION / new_width
|
| 159 |
+
new_width *= scale
|
| 160 |
+
new_height *= scale
|
| 161 |
+
|
| 162 |
+
# Rule 3: Round to the nearest multiple of DIMENSION_MULTIPLE
|
| 163 |
+
final_width = int(round(new_width / DIMENSION_MULTIPLE) * DIMENSION_MULTIPLE)
|
| 164 |
+
final_height = int(round(new_height / DIMENSION_MULTIPLE) * DIMENSION_MULTIPLE)
|
| 165 |
+
|
| 166 |
+
# Ensure final dimensions are at least the minimum
|
| 167 |
+
final_width = max(final_width, MIN_DIMENSION if aspect_ratio < 1 else SQUARE_SIZE)
|
| 168 |
+
final_height = max(final_height, MIN_DIMENSION if aspect_ratio > 1 else SQUARE_SIZE)
|
| 169 |
+
|
| 170 |
+
|
| 171 |
+
return image.resize((final_width, final_height), Image.Resampling.LANCZOS)
|
| 172 |
+
|
| 173 |
+
def resize_and_crop_to_match(target_image, reference_image):
|
| 174 |
+
"""Resizes and center-crops the target image to match the reference image's dimensions."""
|
| 175 |
+
ref_width, ref_height = reference_image.size
|
| 176 |
+
target_width, target_height = target_image.size
|
| 177 |
+
scale = max(ref_width / target_width, ref_height / target_height)
|
| 178 |
+
new_width, new_height = int(target_width * scale), int(target_height * scale)
|
| 179 |
+
resized = target_image.resize((new_width, new_height), Image.Resampling.LANCZOS)
|
| 180 |
+
left, top = (new_width - ref_width) // 2, (new_height - ref_height) // 2
|
| 181 |
+
return resized.crop((left, top, left + ref_width, top + ref_height))
|
| 182 |
+
|
| 183 |
+
def generate_video(
|
| 184 |
+
start_image_pil,
|
| 185 |
+
end_image_pil,
|
| 186 |
prompt,
|
| 187 |
+
negative_prompt=default_negative_prompt,
|
| 188 |
+
duration_seconds=2.1,
|
| 189 |
+
steps=8,
|
| 190 |
+
guidance_scale=1,
|
| 191 |
+
guidance_scale_2=1,
|
| 192 |
+
seed=42,
|
| 193 |
+
randomize_seed=True,
|
| 194 |
+
progress=gr.Progress(track_tqdm=True)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 195 |
):
|
| 196 |
+
allocation_time = 120
|
| 197 |
+
|
| 198 |
+
if input_image_debug_value[0] is not None or end_image_debug_value[0] is not None or prompt_debug_value[0] is not None or total_second_length_debug_value[0] is not None:
|
| 199 |
+
start_image_pil = input_image_debug_value[0]
|
| 200 |
+
end_image_pil = end_image_debug_value[0]
|
| 201 |
+
prompt = prompt_debug_value[0]
|
| 202 |
+
duration_seconds = total_second_length_debug_value[0]
|
| 203 |
+
allocation_time = min(duration_seconds * 60 * 100, 120)
|
| 204 |
+
|
| 205 |
+
return generate_video_on_gpu(
|
| 206 |
+
start_image_pil,
|
| 207 |
+
end_image_pil,
|
| 208 |
+
prompt,
|
| 209 |
+
negative_prompt,
|
| 210 |
+
duration_seconds,
|
| 211 |
+
steps,
|
| 212 |
+
guidance_scale,
|
| 213 |
+
guidance_scale_2,
|
| 214 |
+
seed,
|
| 215 |
+
randomize_seed,
|
| 216 |
+
progress,
|
| 217 |
+
allocation_time
|
| 218 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 219 |
|
| 220 |
def get_duration(
|
| 221 |
+
start_image_pil,
|
| 222 |
+
end_image_pil,
|
| 223 |
prompt,
|
| 224 |
+
negative_prompt,
|
| 225 |
+
duration_seconds,
|
| 226 |
+
steps,
|
| 227 |
+
guidance_scale,
|
| 228 |
+
guidance_scale_2,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 229 |
seed,
|
| 230 |
+
randomize_seed,
|
| 231 |
+
progress,
|
| 232 |
+
allocation_time
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 233 |
):
|
| 234 |
+
return allocation_time
|
| 235 |
|
| 236 |
@spaces.GPU(duration=get_duration)
|
| 237 |
+
def generate_video_on_gpu(
|
| 238 |
+
start_image_pil,
|
| 239 |
+
end_image_pil,
|
| 240 |
prompt,
|
| 241 |
+
negative_prompt,
|
| 242 |
+
duration_seconds,
|
| 243 |
+
steps,
|
| 244 |
+
guidance_scale,
|
| 245 |
+
guidance_scale_2,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 246 |
seed,
|
| 247 |
+
randomize_seed,
|
| 248 |
+
progress,
|
| 249 |
+
allocation_time
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 250 |
):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 251 |
"""
|
| 252 |
+
Generates a video by interpolating between a start and end image, guided by a text prompt,
|
| 253 |
+
using the diffusers Wan2.2 pipeline.
|
| 254 |
+
"""
|
| 255 |
+
if start_image_pil is None or end_image_pil is None:
|
| 256 |
+
raise gr.Error("Please upload both a start and an end image.")
|
| 257 |
|
| 258 |
+
progress(0.1, desc="Preprocessing images...")
|
| 259 |
|
| 260 |
+
# Step 1: Process the start image to get our target dimensions based on the new rules.
|
| 261 |
+
processed_start_image = process_image_for_video(start_image_pil)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 262 |
|
| 263 |
+
# Step 2: Make the end image match the *exact* dimensions of the processed start image.
|
| 264 |
+
processed_end_image = resize_and_crop_to_match(end_image_pil, processed_start_image)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 265 |
|
| 266 |
+
target_height, target_width = processed_start_image.height, processed_start_image.width
|
| 267 |
+
|
| 268 |
+
# Handle seed and frame count
|
| 269 |
+
current_seed = random.randint(0, MAX_SEED) if randomize_seed else int(seed)
|
| 270 |
+
num_frames = np.clip(int(round(duration_seconds * FIXED_FPS)), MIN_FRAMES_MODEL, MAX_FRAMES_MODEL)
|
| 271 |
+
|
| 272 |
+
progress(0.2, desc=f"Generating {num_frames} frames at {target_width}x{target_height} (seed: {current_seed})...")
|
| 273 |
+
|
| 274 |
+
output_frames_list = pipe(
|
| 275 |
+
image=processed_start_image,
|
| 276 |
+
last_image=processed_end_image,
|
| 277 |
+
prompt=prompt,
|
| 278 |
+
negative_prompt=negative_prompt,
|
| 279 |
+
height=target_height,
|
| 280 |
+
width=target_width,
|
| 281 |
+
num_frames=num_frames,
|
| 282 |
+
guidance_scale=float(guidance_scale),
|
| 283 |
+
guidance_scale_2=float(guidance_scale_2),
|
| 284 |
+
num_inference_steps=int(steps),
|
| 285 |
+
generator=torch.Generator(device="cuda").manual_seed(current_seed),
|
| 286 |
+
).frames[0]
|
| 287 |
+
|
| 288 |
+
progress(0.9, desc="Encoding and saving video...")
|
| 289 |
+
|
| 290 |
+
with tempfile.NamedTemporaryFile(suffix=".mp4", delete=False) as tmpfile:
|
| 291 |
+
video_path = tmpfile.name
|
| 292 |
+
|
| 293 |
+
export_to_video(output_frames_list, video_path, fps=FIXED_FPS)
|
| 294 |
+
|
| 295 |
+
progress(1.0, desc="Done!")
|
| 296 |
+
return video_path, gr.update(value = video_path, visible = True), current_seed
|
| 297 |
+
|
| 298 |
+
|
| 299 |
+
# --- 3. Gradio User Interface ---
|
| 300 |
+
|
| 301 |
+
with gr.Blocks() as app:
|
| 302 |
+
gr.Markdown("# Wan 2.2 First/Last Frame Video Fast")
|
| 303 |
+
gr.Markdown("Based on the [Wan 2.2 First/Last Frame workflow](https://www.reddit.com/r/StableDiffusion/comments/1me4306/psa_wan_22_does_first_frame_last_frame_out_of_the/), applied to 🧨 Diffusers + [lightx2v/Wan2.2-Lightning](https://huggingface.co/lightx2v/Wan2.2-Lightning) 8-step LoRA")
|
| 304 |
+
|
| 305 |
+
with gr.Row(elem_id="general_items"):
|
| 306 |
+
with gr.Column():
|
| 307 |
+
with gr.Group(elem_id="group_all"):
|
| 308 |
+
with gr.Row():
|
| 309 |
+
start_image = gr.Image(type="pil", label="Start Frame", sources=["upload", "clipboard"])
|
| 310 |
+
# Capture the Tabs component in a variable and assign IDs to tabs
|
| 311 |
+
with gr.Tabs(elem_id="group_tabs") as tabs:
|
| 312 |
+
with gr.TabItem("Upload", id="upload_tab"):
|
| 313 |
+
end_image = gr.Image(type="pil", label="End Frame", sources=["upload", "clipboard"])
|
| 314 |
+
with gr.TabItem("Generate", id="generate_tab"):
|
| 315 |
+
generate_5seconds = gr.Button("Generate scene 5 seconds in the future", elem_id="fivesec")
|
| 316 |
+
gr.Markdown("Generate a custom end-frame with an edit model like [Nano Banana](https://huggingface.co/spaces/multimodalart/nano-banana) or [Qwen Image Edit](https://huggingface.co/spaces/multimodalart/Qwen-Image-Edit-Fast)", elem_id="or_item")
|
| 317 |
+
prompt = gr.Textbox(label="Prompt", info="Describe the transition between the two images")
|
| 318 |
+
|
| 319 |
+
with gr.Accordion("Advanced Settings", open=False):
|
| 320 |
+
duration_seconds_input = gr.Slider(minimum=MIN_DURATION, maximum=MAX_DURATION, step=0.1, value=2.1, label="Video Duration (seconds)", info=f"Clamped to model's {MIN_FRAMES_MODEL}-{MAX_FRAMES_MODEL} frames at {FIXED_FPS}fps.")
|
| 321 |
+
negative_prompt_input = gr.Textbox(label="Negative Prompt", value=default_negative_prompt, lines=3)
|
| 322 |
+
steps_slider = gr.Slider(minimum=1, maximum=30, step=1, value=8, label="Inference Steps")
|
| 323 |
+
guidance_scale_input = gr.Slider(minimum=0.0, maximum=10.0, step=0.5, value=1.0, label="Guidance Scale - high noise")
|
| 324 |
+
guidance_scale_2_input = gr.Slider(minimum=0.0, maximum=10.0, step=0.5, value=1.0, label="Guidance Scale - low noise")
|
| 325 |
+
with gr.Row():
|
| 326 |
+
seed_input = gr.Slider(label="Seed", minimum=0, maximum=MAX_SEED, step=1, value=42)
|
| 327 |
+
randomize_seed_checkbox = gr.Checkbox(label="Randomize seed", value=True)
|
| 328 |
+
|
| 329 |
+
with gr.Accordion("Debug", elem_id="wan_accordion", open=False):
|
| 330 |
+
input_image_debug = gr.Image(type="pil", label="Image Debug", height=320)
|
| 331 |
+
input_video_debug = gr.Video(sources='upload', label="Input Video Debug", height=320, visible = False)
|
| 332 |
+
end_image_debug = gr.Image(type="pil", label="End Image Debug", height=320)
|
| 333 |
+
prompt_debug = gr.Textbox(elem_id="wan_prompt_debug", label="Prompt Debug", value='')
|
| 334 |
+
total_second_length_debug = gr.Slider(label="Additional Video Length to Generate (seconds) Debug", minimum=1, maximum=120, value=10, step=0.1)
|
| 335 |
+
|
| 336 |
+
generate_button = gr.Button("Generate Video", variant="primary")
|
| 337 |
+
|
| 338 |
+
with gr.Column():
|
| 339 |
+
output_video = gr.Video(label="Generated Video", autoplay = True, loop = True)
|
| 340 |
+
download_button = gr.DownloadButton(label="Download", visible = True)
|
| 341 |
+
|
| 342 |
+
# Main video generation button
|
| 343 |
+
ui_inputs = [
|
| 344 |
+
start_image,
|
| 345 |
+
end_image,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 346 |
prompt,
|
| 347 |
+
negative_prompt_input,
|
| 348 |
+
duration_seconds_input,
|
| 349 |
+
steps_slider,
|
| 350 |
+
guidance_scale_input,
|
| 351 |
+
guidance_scale_2_input,
|
| 352 |
+
seed_input,
|
| 353 |
+
randomize_seed_checkbox
|
| 354 |
+
]
|
| 355 |
+
ui_outputs = [output_video, download_button, seed_input]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 356 |
|
| 357 |
+
generate_button.click(
|
| 358 |
+
fn=generate_video,
|
| 359 |
+
inputs=ui_inputs,
|
| 360 |
+
outputs=ui_outputs
|
| 361 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 362 |
|
| 363 |
+
generate_5seconds.click(
|
| 364 |
+
fn=switch_to_upload_tab,
|
| 365 |
+
inputs=None,
|
| 366 |
+
outputs=[tabs]
|
| 367 |
+
).then(
|
| 368 |
+
fn=lambda img: generate_end_frame(img, "this image is a still frame from a movie. generate a new frame with what happens on this scene 5 seconds in the future"),
|
| 369 |
+
inputs=[start_image],
|
| 370 |
+
outputs=[end_image]
|
| 371 |
+
).success(
|
| 372 |
+
fn=generate_video,
|
| 373 |
+
inputs=ui_inputs,
|
| 374 |
+
outputs=ui_outputs
|
| 375 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 376 |
|
| 377 |
+
def handle_field_debug_change(input_image_debug_data, input_video_debug_data, end_image_debug_data, prompt_debug_data, total_second_length_debug_data):
|
| 378 |
+
print("handle_field_debug_change")
|
| 379 |
input_image_debug_value[0] = input_image_debug_data
|
| 380 |
+
input_video_debug_value[0] = input_video_debug_data
|
| 381 |
+
end_image_debug_value[0] = end_image_debug_data
|
| 382 |
prompt_debug_value[0] = prompt_debug_data
|
| 383 |
+
total_second_length_debug_value[0] = total_second_length_debug_data
|
| 384 |
return []
|
| 385 |
|
| 386 |
input_image_debug.upload(
|
| 387 |
fn=handle_field_debug_change,
|
| 388 |
+
inputs=[input_image_debug, input_video_debug, end_image_debug, prompt_debug, total_second_length_debug],
|
| 389 |
+
outputs=[]
|
| 390 |
+
)
|
| 391 |
+
|
| 392 |
+
input_video_debug.upload(
|
| 393 |
+
fn=handle_field_debug_change,
|
| 394 |
+
inputs=[input_image_debug, input_video_debug, end_image_debug, prompt_debug, total_second_length_debug],
|
| 395 |
+
outputs=[]
|
| 396 |
+
)
|
| 397 |
+
|
| 398 |
+
end_image_debug.upload(
|
| 399 |
+
fn=handle_field_debug_change,
|
| 400 |
+
inputs=[input_image_debug, input_video_debug, end_image_debug, prompt_debug, total_second_length_debug],
|
| 401 |
outputs=[]
|
| 402 |
)
|
| 403 |
|
| 404 |
prompt_debug.change(
|
| 405 |
fn=handle_field_debug_change,
|
| 406 |
+
inputs=[input_image_debug, input_video_debug, end_image_debug, prompt_debug, total_second_length_debug],
|
| 407 |
outputs=[]
|
| 408 |
)
|
| 409 |
|
| 410 |
+
total_second_length_debug.change(
|
| 411 |
fn=handle_field_debug_change,
|
| 412 |
+
inputs=[input_image_debug, input_video_debug, end_image_debug, prompt_debug, total_second_length_debug],
|
| 413 |
outputs=[]
|
| 414 |
)
|
| 415 |
+
|
| 416 |
+
with gr.Row(elem_id="wan_image_examples", visible=False):
|
| 417 |
+
gr.Examples(
|
| 418 |
+
label = "Examples from images",
|
| 419 |
+
examples = [
|
| 420 |
+
["ugly_sonic.jpeg", "squatting_sonic.png", "the character dodges the missiles"],
|
| 421 |
+
["capyabara_zoomed.png", "capyabara.webp", "a dramatic dolly zoom"],
|
| 422 |
+
["squatting_sonic.png", "ugly_sonic.jpeg", "the character jumps"],
|
| 423 |
+
["poli_tower.png", "tower_takes_off.png", "the man turns around"],
|
| 424 |
+
["capyabara.webp", "capyabara_zoomed.png", "a straight forward zoom"],
|
| 425 |
+
],
|
| 426 |
+
inputs = [start_image, end_image, prompt],
|
| 427 |
+
outputs = ui_outputs,
|
| 428 |
+
fn = generate_video,
|
| 429 |
+
run_on_click = True,
|
| 430 |
+
cache_examples = True,
|
| 431 |
+
)
|
| 432 |
+
|
| 433 |
+
gr.Examples(
|
| 434 |
+
examples = [
|
| 435 |
+
["poli_tower.png", "tower_takes_off.png", "the man turns around"],
|
| 436 |
+
["ugly_sonic.jpeg", "squatting_sonic.png", "the character dodges the missiles"],
|
| 437 |
+
["capyabara_zoomed.png", "capyabara.webp", "a dramatic dolly zoom"],
|
| 438 |
+
],
|
| 439 |
+
inputs = [start_image, end_image, prompt],
|
| 440 |
+
outputs = ui_outputs,
|
| 441 |
+
fn = generate_video,
|
| 442 |
+
cache_examples = False,
|
| 443 |
+
)
|
| 444 |
+
|
| 445 |
+
if __name__ == "__main__":
|
| 446 |
+
app.launch(share=True)
|
requirements.txt
CHANGED
|
@@ -1,43 +1,11 @@
|
|
| 1 |
-
|
| 2 |
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
wandb==0.20.1
|
| 13 |
-
httpx==0.28.1
|
| 14 |
-
transformers==4.43.0
|
| 15 |
-
accelerate==1.8.0
|
| 16 |
-
scikit-learn==1.7.0
|
| 17 |
-
einops==0.8.1
|
| 18 |
-
einops-exts==0.0.4
|
| 19 |
-
timm==1.0.15
|
| 20 |
-
openai-clip==1.0.1
|
| 21 |
-
fsspec==2025.5.1
|
| 22 |
-
kornia==0.8.1
|
| 23 |
-
matplotlib==3.10.3
|
| 24 |
-
ninja==1.11.1.4
|
| 25 |
-
omegaconf==2.3.0
|
| 26 |
-
opencv-python==4.11.0.86
|
| 27 |
-
pandas==2.3.0
|
| 28 |
-
pillow==11.2.1
|
| 29 |
-
pytorch-lightning==2.5.1.post0
|
| 30 |
-
PyYAML==6.0.2
|
| 31 |
-
scipy==1.15.3
|
| 32 |
-
tqdm==4.67.1
|
| 33 |
-
triton==3.3.0
|
| 34 |
-
urllib3==2.4.0
|
| 35 |
-
webdataset==0.2.111
|
| 36 |
-
xformers==0.0.30
|
| 37 |
-
facexlib==0.3.0
|
| 38 |
-
k-diffusion==0.1.1.post1
|
| 39 |
-
diffusers==0.33.1
|
| 40 |
-
imageio==2.37.0
|
| 41 |
-
pillow-heif==0.22.0
|
| 42 |
-
|
| 43 |
-
open-clip-torch==2.24.0
|
|
|
|
| 1 |
+
git+https://github.com/linoytsaban/diffusers.git@wan22-loras
|
| 2 |
|
| 3 |
+
transformers
|
| 4 |
+
accelerate
|
| 5 |
+
safetensors
|
| 6 |
+
sentencepiece
|
| 7 |
+
peft
|
| 8 |
+
ftfy
|
| 9 |
+
imageio-ffmpeg
|
| 10 |
+
opencv-python
|
| 11 |
+
torchao==0.11.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|