update
Browse files
app.py
CHANGED
|
@@ -5,6 +5,32 @@ import random
|
|
| 5 |
import gradio as gr
|
| 6 |
import spaces
|
| 7 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
|
| 9 |
@dataclass(frozen=True)
|
| 10 |
class SliderConfig:
|
|
@@ -29,47 +55,55 @@ GAMMA_SLIDER = SliderConfig(
|
|
| 29 |
maximum=10.0,
|
| 30 |
step=0.5,
|
| 31 |
value=5.0,
|
| 32 |
-
info="Scheduler adjustment parameter."
|
| 33 |
)
|
| 34 |
|
| 35 |
STEP_SLIDER = SliderConfig(
|
| 36 |
label="Inference Steps",
|
| 37 |
-
minimum=
|
| 38 |
-
maximum=
|
| 39 |
step=1,
|
| 40 |
-
value=
|
| 41 |
-
info="More steps improve quality but take longer."
|
| 42 |
)
|
| 43 |
|
| 44 |
GUIDANCE_SLIDER = SliderConfig(
|
| 45 |
label="Guidance Scale (CFG)",
|
| 46 |
minimum=1.0,
|
| 47 |
-
maximum=
|
| 48 |
step=0.5,
|
| 49 |
-
value=
|
| 50 |
-
info="Controls adherence to the text prompt."
|
| 51 |
)
|
| 52 |
|
| 53 |
|
| 54 |
STYLE_CHOICES = [
|
| 55 |
-
"
|
| 56 |
-
"
|
| 57 |
-
"
|
| 58 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 59 |
]
|
| 60 |
|
| 61 |
|
| 62 |
EXAMPLE_INPUTS = [
|
| 63 |
-
["assets/video_00000000.mp4",
|
| 64 |
-
["assets/video_00000007.mp4",
|
| 65 |
-
["assets/
|
|
|
|
| 66 |
]
|
| 67 |
|
| 68 |
|
| 69 |
PRESET_MODES = {
|
| 70 |
-
"Fast": PresetConfig(shift_gamma=5.0, steps=
|
| 71 |
-
"Balanced": PresetConfig(shift_gamma=5.0, steps=10, guidance_scale=
|
| 72 |
-
"Quality": PresetConfig(shift_gamma=5.0, steps=20, guidance_scale=
|
| 73 |
}
|
| 74 |
|
| 75 |
|
|
@@ -112,7 +146,30 @@ def run_stylization(
|
|
| 112 |
f"Seed={resolved_seed}"
|
| 113 |
)
|
| 114 |
|
| 115 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 116 |
|
| 117 |
|
| 118 |
def _resolve_seed(seed_value, randomize):
|
|
@@ -175,26 +232,31 @@ def build_demo() -> gr.Blocks:
|
|
| 175 |
|
| 176 |
with gr.Row():
|
| 177 |
with gr.Column(scale=1) as control_col:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 178 |
with gr.Tabs():
|
| 179 |
with gr.Tab("Quick Generate"):
|
| 180 |
-
prompt_quick = gr.Dropdown(
|
| 181 |
-
label="Style Instruction",
|
| 182 |
-
choices=STYLE_CHOICES,
|
| 183 |
-
value=STYLE_CHOICES[0],
|
| 184 |
-
allow_custom_value=False,
|
| 185 |
-
)
|
| 186 |
-
|
| 187 |
with gr.Row():
|
| 188 |
-
fast_btn = gr.Button(
|
| 189 |
-
|
| 190 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 191 |
|
| 192 |
_bind_preset_button(
|
| 193 |
button=fast_btn,
|
| 194 |
preset_key="Fast",
|
| 195 |
inputs=[
|
| 196 |
input_video,
|
| 197 |
-
|
| 198 |
],
|
| 199 |
output=output_video,
|
| 200 |
extra_kwargs={"seed": None, "randomize_seed": True},
|
|
@@ -204,7 +266,7 @@ def build_demo() -> gr.Blocks:
|
|
| 204 |
preset_key="Balanced",
|
| 205 |
inputs=[
|
| 206 |
input_video,
|
| 207 |
-
|
| 208 |
],
|
| 209 |
output=output_video,
|
| 210 |
extra_kwargs={"seed": None, "randomize_seed": True},
|
|
@@ -214,19 +276,13 @@ def build_demo() -> gr.Blocks:
|
|
| 214 |
preset_key="Quality",
|
| 215 |
inputs=[
|
| 216 |
input_video,
|
| 217 |
-
|
| 218 |
],
|
| 219 |
output=output_video,
|
| 220 |
extra_kwargs={"seed": None, "randomize_seed": True},
|
| 221 |
)
|
| 222 |
|
| 223 |
with gr.Tab("Advanced Settings"):
|
| 224 |
-
prompt_adv = gr.Dropdown(
|
| 225 |
-
label="Style Instruction",
|
| 226 |
-
choices=STYLE_CHOICES,
|
| 227 |
-
value=STYLE_CHOICES[0],
|
| 228 |
-
allow_custom_value=True,
|
| 229 |
-
)
|
| 230 |
with gr.Row():
|
| 231 |
shift_gamma = _create_slider(GAMMA_SLIDER)
|
| 232 |
guidance_scale = _create_slider(GUIDANCE_SLIDER)
|
|
@@ -235,7 +291,7 @@ def build_demo() -> gr.Blocks:
|
|
| 235 |
num_steps = _create_slider(STEP_SLIDER)
|
| 236 |
randomize_seed_adv = gr.Checkbox(
|
| 237 |
label="Randomize Seed",
|
| 238 |
-
value=
|
| 239 |
info="Checked = new random seed each run. Uncheck to provide your own seed.",
|
| 240 |
)
|
| 241 |
|
|
@@ -251,7 +307,7 @@ def build_demo() -> gr.Blocks:
|
|
| 251 |
fn=run_stylization,
|
| 252 |
inputs=[
|
| 253 |
input_video,
|
| 254 |
-
|
| 255 |
shift_gamma,
|
| 256 |
num_steps,
|
| 257 |
guidance_scale,
|
|
@@ -264,7 +320,7 @@ def build_demo() -> gr.Blocks:
|
|
| 264 |
with gr.Column(scale=1):
|
| 265 |
gr.Examples(
|
| 266 |
examples=EXAMPLE_INPUTS,
|
| 267 |
-
inputs=[input_video,
|
| 268 |
label="Example inputs",
|
| 269 |
)
|
| 270 |
|
|
|
|
| 5 |
import gradio as gr
|
| 6 |
import spaces
|
| 7 |
|
| 8 |
+
import torch
|
| 9 |
+
from diffusers import WanPipeline
|
| 10 |
+
from diffusers.utils import export_to_video, load_video
|
| 11 |
+
from vibt.wan import load_vibt_weight, encode_video
|
| 12 |
+
from vibt.scheduler import ViBTScheduler
|
| 13 |
+
import tempfile
|
| 14 |
+
import os
|
| 15 |
+
import cv2
|
| 16 |
+
|
| 17 |
+
|
| 18 |
+
def get_fps(path):
|
| 19 |
+
cap = cv2.VideoCapture(path)
|
| 20 |
+
fps = cap.get(cv2.CAP_PROP_FPS)
|
| 21 |
+
cap.release()
|
| 22 |
+
return fps
|
| 23 |
+
|
| 24 |
+
|
| 25 |
+
base_model_id = "Wan-AI/Wan2.1-T2V-1.3B-Diffusers"
|
| 26 |
+
pipe = WanPipeline.from_pretrained(base_model_id, torch_dtype=torch.bfloat16).to("cuda")
|
| 27 |
+
load_vibt_weight(
|
| 28 |
+
pipe.transformer,
|
| 29 |
+
"Yuanshi/ViBT",
|
| 30 |
+
"video/video_stylization.safetensors",
|
| 31 |
+
)
|
| 32 |
+
pipe.scheduler = ViBTScheduler.from_scheduler(pipe.scheduler)
|
| 33 |
+
|
| 34 |
|
| 35 |
@dataclass(frozen=True)
|
| 36 |
class SliderConfig:
|
|
|
|
| 55 |
maximum=10.0,
|
| 56 |
step=0.5,
|
| 57 |
value=5.0,
|
| 58 |
+
info="Scheduler adjustment parameter.",
|
| 59 |
)
|
| 60 |
|
| 61 |
STEP_SLIDER = SliderConfig(
|
| 62 |
label="Inference Steps",
|
| 63 |
+
minimum=6,
|
| 64 |
+
maximum=28,
|
| 65 |
step=1,
|
| 66 |
+
value=10,
|
| 67 |
+
info="More steps improve quality but take longer.",
|
| 68 |
)
|
| 69 |
|
| 70 |
GUIDANCE_SLIDER = SliderConfig(
|
| 71 |
label="Guidance Scale (CFG)",
|
| 72 |
minimum=1.0,
|
| 73 |
+
maximum=5.0,
|
| 74 |
step=0.5,
|
| 75 |
+
value=2,
|
| 76 |
+
info="Controls adherence to the text prompt.",
|
| 77 |
)
|
| 78 |
|
| 79 |
|
| 80 |
STYLE_CHOICES = [
|
| 81 |
+
"Make it Illustration style.",
|
| 82 |
+
"Make it a drawing by Van Gogh.",
|
| 83 |
+
"Make it a pencil sketch style.",
|
| 84 |
+
"Make it watercolor drawing style.",
|
| 85 |
+
"Make it a Pixel Art.",
|
| 86 |
+
"Make it a Japanese anime style, cel shading.",
|
| 87 |
+
"Make it the style of Neon Light Art.",
|
| 88 |
+
"Make it papercut style.",
|
| 89 |
+
"Make it a blueprint.",
|
| 90 |
+
"Make it Comic Book Style.",
|
| 91 |
+
"Render the subject as a classical sculpture carved from a single block of pristine white marble.",
|
| 92 |
]
|
| 93 |
|
| 94 |
|
| 95 |
EXAMPLE_INPUTS = [
|
| 96 |
+
["assets/video_00000000.mp4", STYLE_CHOICES[0]],
|
| 97 |
+
["assets/video_00000007.mp4", STYLE_CHOICES[1]],
|
| 98 |
+
["assets/video_00000019.mp4", STYLE_CHOICES[2]],
|
| 99 |
+
["assets/video_00000071.mp4", STYLE_CHOICES[3]],
|
| 100 |
]
|
| 101 |
|
| 102 |
|
| 103 |
PRESET_MODES = {
|
| 104 |
+
"Fast": PresetConfig(shift_gamma=5.0, steps=6, guidance_scale=2),
|
| 105 |
+
"Balanced": PresetConfig(shift_gamma=5.0, steps=10, guidance_scale=2),
|
| 106 |
+
"Quality": PresetConfig(shift_gamma=5.0, steps=20, guidance_scale=2),
|
| 107 |
}
|
| 108 |
|
| 109 |
|
|
|
|
| 146 |
f"Seed={resolved_seed}"
|
| 147 |
)
|
| 148 |
|
| 149 |
+
source_video = load_video(input_video_path)
|
| 150 |
+
source_video = [each.resize((832, 480)) for each in source_video][:81]
|
| 151 |
+
if len(source_video) < 81:
|
| 152 |
+
source_video += [source_video[-1]] * (81 - len(source_video))
|
| 153 |
+
source_fps = get_fps(input_video_path)
|
| 154 |
+
|
| 155 |
+
source_latents = encode_video(pipe, source_video)
|
| 156 |
+
|
| 157 |
+
pipe.scheduler.set_parameters(
|
| 158 |
+
noise_scale=1.0, shift_gamma=shift_gamma, seed=resolved_seed
|
| 159 |
+
)
|
| 160 |
+
|
| 161 |
+
output = pipe(
|
| 162 |
+
prompt=prompt,
|
| 163 |
+
num_inference_steps=steps,
|
| 164 |
+
guidance_scale=guidance_scale,
|
| 165 |
+
latents=source_latents,
|
| 166 |
+
).frames[0]
|
| 167 |
+
|
| 168 |
+
tmp_dir = tempfile.mkdtemp()
|
| 169 |
+
out_path = os.path.join(tmp_dir, f"{random.randint(0, 2**31 - 1)}.mp4")
|
| 170 |
+
export_to_video(output, out_path, fps=source_fps)
|
| 171 |
+
print(out_path)
|
| 172 |
+
return out_path
|
| 173 |
|
| 174 |
|
| 175 |
def _resolve_seed(seed_value, randomize):
|
|
|
|
| 232 |
|
| 233 |
with gr.Row():
|
| 234 |
with gr.Column(scale=1) as control_col:
|
| 235 |
+
prompt = gr.Dropdown(
|
| 236 |
+
label="Style Instruction",
|
| 237 |
+
choices=STYLE_CHOICES,
|
| 238 |
+
value=STYLE_CHOICES[0],
|
| 239 |
+
allow_custom_value=True,
|
| 240 |
+
)
|
| 241 |
with gr.Tabs():
|
| 242 |
with gr.Tab("Quick Generate"):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 243 |
with gr.Row():
|
| 244 |
+
fast_btn = gr.Button(
|
| 245 |
+
"⚡ Fast Generate", variant="primary"
|
| 246 |
+
)
|
| 247 |
+
balanced_btn = gr.Button(
|
| 248 |
+
"🎯 Balanced Generate", variant="primary"
|
| 249 |
+
)
|
| 250 |
+
quality_btn = gr.Button(
|
| 251 |
+
"🌟 Quality Generate", variant="primary"
|
| 252 |
+
)
|
| 253 |
|
| 254 |
_bind_preset_button(
|
| 255 |
button=fast_btn,
|
| 256 |
preset_key="Fast",
|
| 257 |
inputs=[
|
| 258 |
input_video,
|
| 259 |
+
prompt,
|
| 260 |
],
|
| 261 |
output=output_video,
|
| 262 |
extra_kwargs={"seed": None, "randomize_seed": True},
|
|
|
|
| 266 |
preset_key="Balanced",
|
| 267 |
inputs=[
|
| 268 |
input_video,
|
| 269 |
+
prompt,
|
| 270 |
],
|
| 271 |
output=output_video,
|
| 272 |
extra_kwargs={"seed": None, "randomize_seed": True},
|
|
|
|
| 276 |
preset_key="Quality",
|
| 277 |
inputs=[
|
| 278 |
input_video,
|
| 279 |
+
prompt,
|
| 280 |
],
|
| 281 |
output=output_video,
|
| 282 |
extra_kwargs={"seed": None, "randomize_seed": True},
|
| 283 |
)
|
| 284 |
|
| 285 |
with gr.Tab("Advanced Settings"):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 286 |
with gr.Row():
|
| 287 |
shift_gamma = _create_slider(GAMMA_SLIDER)
|
| 288 |
guidance_scale = _create_slider(GUIDANCE_SLIDER)
|
|
|
|
| 291 |
num_steps = _create_slider(STEP_SLIDER)
|
| 292 |
randomize_seed_adv = gr.Checkbox(
|
| 293 |
label="Randomize Seed",
|
| 294 |
+
value=True,
|
| 295 |
info="Checked = new random seed each run. Uncheck to provide your own seed.",
|
| 296 |
)
|
| 297 |
|
|
|
|
| 307 |
fn=run_stylization,
|
| 308 |
inputs=[
|
| 309 |
input_video,
|
| 310 |
+
prompt,
|
| 311 |
shift_gamma,
|
| 312 |
num_steps,
|
| 313 |
guidance_scale,
|
|
|
|
| 320 |
with gr.Column(scale=1):
|
| 321 |
gr.Examples(
|
| 322 |
examples=EXAMPLE_INPUTS,
|
| 323 |
+
inputs=[input_video, prompt],
|
| 324 |
label="Example inputs",
|
| 325 |
)
|
| 326 |
|