Spaces:
Running
on
Zero
Running
on
Zero
update demo
Browse files
app.py
CHANGED
@@ -33,171 +33,194 @@ css = """
|
|
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 |
negative_prompt_textbox,
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
height_slider,
|
88 |
txt_cfg_scale_slider,
|
89 |
img_cfg_scale_slider,
|
90 |
-
center_crop,
|
91 |
frame_stride,
|
92 |
use_frameinit,
|
93 |
frame_init_noise_level,
|
94 |
-
|
95 |
-
):
|
96 |
-
if self.pipeline is None:
|
97 |
-
raise gr.Error(f"Please select a pretrained pipeline path.")
|
98 |
-
if input_image_path == "":
|
99 |
-
raise gr.Error(f"Please upload an input image.")
|
100 |
-
if (not center_crop) and (width_slider % 8 != 0 or height_slider % 8 != 0):
|
101 |
-
raise gr.Error(f"`height` and `width` have to be divisible by 8 but are {height_slider} and {width_slider}.")
|
102 |
-
if center_crop and (width_slider % 8 != 0 or height_slider % 8 != 0):
|
103 |
-
raise gr.Error(f"`height` and `width` (after cropping) have to be divisible by 8 but are {height_slider} and {width_slider}.")
|
104 |
-
|
105 |
-
if is_xformers_available() and int(torch.__version__.split(".")[0]) < 2: self.pipeline.unet.enable_xformers_memory_efficient_attention()
|
106 |
-
|
107 |
-
if seed_textbox != -1 and seed_textbox != "": torch.manual_seed(int(seed_textbox))
|
108 |
-
else: torch.seed()
|
109 |
-
seed = torch.initial_seed()
|
110 |
-
|
111 |
-
if input_image_path.startswith("http://") or input_image_path.startswith("https://"):
|
112 |
-
first_frame = Image.open(requests.get(input_image_path, stream=True).raw).convert('RGB')
|
113 |
-
else:
|
114 |
-
first_frame = Image.open(input_image_path).convert('RGB')
|
115 |
-
|
116 |
-
original_width, original_height = first_frame.size
|
117 |
-
|
118 |
-
if not center_crop:
|
119 |
-
img_transform = T.Compose([
|
120 |
-
T.ToTensor(),
|
121 |
-
T.Resize((height_slider, width_slider), antialias=None),
|
122 |
-
T.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True),
|
123 |
-
])
|
124 |
-
else:
|
125 |
-
aspect_ratio = original_width / original_height
|
126 |
-
crop_aspect_ratio = width_slider / height_slider
|
127 |
-
if aspect_ratio > crop_aspect_ratio:
|
128 |
-
center_crop_width = int(crop_aspect_ratio * original_height)
|
129 |
-
center_crop_height = original_height
|
130 |
-
elif aspect_ratio < crop_aspect_ratio:
|
131 |
-
center_crop_width = original_width
|
132 |
-
center_crop_height = int(original_width / crop_aspect_ratio)
|
133 |
-
else:
|
134 |
-
center_crop_width = original_width
|
135 |
-
center_crop_height = original_height
|
136 |
-
img_transform = T.Compose([
|
137 |
-
T.ToTensor(),
|
138 |
-
T.CenterCrop((center_crop_height, center_crop_width)),
|
139 |
-
T.Resize((height_slider, width_slider), antialias=None),
|
140 |
-
T.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True),
|
141 |
-
])
|
142 |
-
|
143 |
-
first_frame = img_transform(first_frame).unsqueeze(0)
|
144 |
-
first_frame = first_frame.to("cuda")
|
145 |
-
print("first_frame", first_frame.device)
|
146 |
-
|
147 |
-
if use_frameinit:
|
148 |
-
self.pipeline.init_filter(
|
149 |
-
width = width_slider,
|
150 |
-
height = height_slider,
|
151 |
-
video_length = 16,
|
152 |
-
filter_params = OmegaConf.create({'method': 'gaussian', 'd_s': 0.25, 'd_t': 0.25,})
|
153 |
-
)
|
154 |
|
|
|
|
|
|
|
|
|
155 |
|
156 |
-
|
157 |
-
|
158 |
-
|
159 |
-
|
160 |
-
|
161 |
-
|
162 |
-
|
163 |
-
|
164 |
-
|
165 |
-
|
166 |
-
|
167 |
-
|
168 |
-
|
169 |
-
|
170 |
-
|
171 |
-
|
172 |
-
)
|
173 |
-
|
174 |
-
global sample_idx
|
175 |
-
sample_idx += 1
|
176 |
-
save_sample_path = os.path.join(self.savedir_sample, f"{sample_idx}.mp4")
|
177 |
-
save_videos_grid(sample, save_sample_path, format="mp4")
|
178 |
-
|
179 |
-
sample_config = {
|
180 |
-
"prompt": prompt_textbox,
|
181 |
-
"n_prompt": negative_prompt_textbox,
|
182 |
-
"first_frame_path": input_image_path,
|
183 |
-
"sampler": sampler_dropdown,
|
184 |
-
"num_inference_steps": sample_step_slider,
|
185 |
-
"guidance_scale_text": txt_cfg_scale_slider,
|
186 |
-
"guidance_scale_image": img_cfg_scale_slider,
|
187 |
-
"width": width_slider,
|
188 |
-
"height": height_slider,
|
189 |
-
"video_length": 8,
|
190 |
-
"seed": seed
|
191 |
-
}
|
192 |
-
json_str = json.dumps(sample_config, indent=4)
|
193 |
-
with open(os.path.join(self.savedir, "logs.json"), "a") as f:
|
194 |
-
f.write(json_str)
|
195 |
-
f.write("\n\n")
|
196 |
-
|
197 |
-
return gr.Video(value=save_sample_path)
|
198 |
|
|
|
|
|
199 |
|
200 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
201 |
|
202 |
|
203 |
def ui():
|
@@ -257,7 +280,7 @@ def ui():
|
|
257 |
|
258 |
with gr.Row():
|
259 |
input_image = gr.Image(label="Input Image", interactive=True)
|
260 |
-
input_image.upload(fn=
|
261 |
result_video = gr.Video(label="Generated Animation", interactive=False, autoplay=True)
|
262 |
|
263 |
def update_and_resize_image(input_image_path, height_slider, width_slider, center_crop):
|
@@ -265,7 +288,6 @@ def ui():
|
|
265 |
pil_image = Image.open(requests.get(input_image_path, stream=True).raw).convert('RGB')
|
266 |
else:
|
267 |
pil_image = Image.open(input_image_path).convert('RGB')
|
268 |
-
controller.image_resolution = pil_image.size
|
269 |
original_width, original_height = pil_image.size
|
270 |
|
271 |
if center_crop:
|
@@ -293,7 +315,7 @@ def ui():
|
|
293 |
input_image_path.submit(fn=update_and_resize_image, inputs=[input_image_path, height_slider, width_slider, center_crop], outputs=[input_image])
|
294 |
|
295 |
generate_button.click(
|
296 |
-
fn=
|
297 |
inputs=[
|
298 |
prompt_textbox,
|
299 |
negative_prompt_textbox,
|
|
|
33 |
}
|
34 |
"""
|
35 |
|
36 |
+
|
37 |
+
basedir = os.getcwd()
|
38 |
+
savedir = os.path.join(basedir, "samples/Gradio", datetime.now().strftime("%Y-%m-%dT%H-%M-%S"))
|
39 |
+
savedir_sample = os.path.join(savedir, "sample")
|
40 |
+
os.makedirs(savedir, exist_ok=True)
|
41 |
+
|
42 |
+
# config models
|
43 |
+
pipeline = ConditionalAnimationPipeline.from_pretrained("TIGER-Lab/ConsistI2V", torch_dtype=torch.float16,)
|
44 |
+
pipeline.to("cuda")
|
45 |
+
# pipeline.to("cuda")
|
46 |
+
|
47 |
+
def update_textbox_and_save_image(input_image, height_slider, width_slider, center_crop):
|
48 |
+
pil_image = Image.fromarray(input_image.astype(np.uint8)).convert("RGB")
|
49 |
+
img_path = os.path.join(savedir, "input_image.png")
|
50 |
+
pil_image.save(img_path)
|
51 |
+
|
52 |
+
original_width, original_height = pil_image.size
|
53 |
+
if center_crop:
|
54 |
+
crop_aspect_ratio = width_slider / height_slider
|
55 |
+
aspect_ratio = original_width / original_height
|
56 |
+
if aspect_ratio > crop_aspect_ratio:
|
57 |
+
new_width = int(crop_aspect_ratio * original_height)
|
58 |
+
left = (original_width - new_width) / 2
|
59 |
+
top = 0
|
60 |
+
right = left + new_width
|
61 |
+
bottom = original_height
|
62 |
+
pil_image = pil_image.crop((left, top, right, bottom))
|
63 |
+
elif aspect_ratio < crop_aspect_ratio:
|
64 |
+
new_height = int(original_width / crop_aspect_ratio)
|
65 |
+
top = (original_height - new_height) / 2
|
66 |
+
left = 0
|
67 |
+
right = original_width
|
68 |
+
bottom = top + new_height
|
69 |
+
pil_image = pil_image.crop((left, top, right, bottom))
|
70 |
+
|
71 |
+
pil_image = pil_image.resize((width_slider, height_slider))
|
72 |
+
return gr.Textbox(value=img_path), gr.Image(value=np.array(pil_image))
|
73 |
+
|
74 |
+
|
75 |
+
def animate(
|
76 |
+
prompt_textbox,
|
77 |
+
negative_prompt_textbox,
|
78 |
+
input_image_path,
|
79 |
+
sampler_dropdown,
|
80 |
+
sample_step_slider,
|
81 |
+
width_slider,
|
82 |
+
height_slider,
|
83 |
+
txt_cfg_scale_slider,
|
84 |
+
img_cfg_scale_slider,
|
85 |
+
center_crop,
|
86 |
+
frame_stride,
|
87 |
+
use_frameinit,
|
88 |
+
frame_init_noise_level,
|
89 |
+
seed_textbox
|
90 |
+
):
|
91 |
+
if pipeline is None:
|
92 |
+
raise gr.Error(f"Please select a pretrained pipeline path.")
|
93 |
+
if input_image_path == "":
|
94 |
+
raise gr.Error(f"Please upload an input image.")
|
95 |
+
if (not center_crop) and (width_slider % 8 != 0 or height_slider % 8 != 0):
|
96 |
+
raise gr.Error(f"`height` and `width` have to be divisible by 8 but are {height_slider} and {width_slider}.")
|
97 |
+
if center_crop and (width_slider % 8 != 0 or height_slider % 8 != 0):
|
98 |
+
raise gr.Error(f"`height` and `width` (after cropping) have to be divisible by 8 but are {height_slider} and {width_slider}.")
|
99 |
+
|
100 |
+
if is_xformers_available() and int(torch.__version__.split(".")[0]) < 2: pipeline.unet.enable_xformers_memory_efficient_attention()
|
101 |
+
|
102 |
+
if seed_textbox != -1 and seed_textbox != "": torch.manual_seed(int(seed_textbox))
|
103 |
+
else: torch.seed()
|
104 |
+
seed = torch.initial_seed()
|
105 |
+
|
106 |
+
if input_image_path.startswith("http://") or input_image_path.startswith("https://"):
|
107 |
+
first_frame = Image.open(requests.get(input_image_path, stream=True).raw).convert('RGB')
|
108 |
+
else:
|
109 |
+
first_frame = Image.open(input_image_path).convert('RGB')
|
110 |
+
|
111 |
+
original_width, original_height = first_frame.size
|
112 |
+
|
113 |
+
if not center_crop:
|
114 |
+
img_transform = T.Compose([
|
115 |
+
T.ToTensor(),
|
116 |
+
T.Resize((height_slider, width_slider), antialias=None),
|
117 |
+
T.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True),
|
118 |
+
])
|
119 |
+
else:
|
120 |
+
aspect_ratio = original_width / original_height
|
121 |
+
crop_aspect_ratio = width_slider / height_slider
|
122 |
+
if aspect_ratio > crop_aspect_ratio:
|
123 |
+
center_crop_width = int(crop_aspect_ratio * original_height)
|
124 |
+
center_crop_height = original_height
|
125 |
+
elif aspect_ratio < crop_aspect_ratio:
|
126 |
+
center_crop_width = original_width
|
127 |
+
center_crop_height = int(original_width / crop_aspect_ratio)
|
128 |
+
else:
|
129 |
+
center_crop_width = original_width
|
130 |
+
center_crop_height = original_height
|
131 |
+
img_transform = T.Compose([
|
132 |
+
T.ToTensor(),
|
133 |
+
T.CenterCrop((center_crop_height, center_crop_width)),
|
134 |
+
T.Resize((height_slider, width_slider), antialias=None),
|
135 |
+
T.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True),
|
136 |
+
])
|
137 |
+
|
138 |
+
first_frame = img_transform(first_frame).unsqueeze(0)
|
139 |
+
|
140 |
+
if use_frameinit:
|
141 |
+
pipeline.init_filter(
|
142 |
+
width = width_slider,
|
143 |
+
height = height_slider,
|
144 |
+
video_length = 16,
|
145 |
+
filter_params = OmegaConf.create({'method': 'gaussian', 'd_s': 0.25, 'd_t': 0.25,})
|
146 |
+
)
|
147 |
+
|
148 |
+
sample = run_pipeline(
|
149 |
+
pipeline,
|
150 |
+
prompt_textbox,
|
151 |
negative_prompt_textbox,
|
152 |
+
first_frame,
|
153 |
+
sample_step_slider,
|
154 |
+
width_slider,
|
155 |
+
height_slider,
|
|
|
156 |
txt_cfg_scale_slider,
|
157 |
img_cfg_scale_slider,
|
|
|
158 |
frame_stride,
|
159 |
use_frameinit,
|
160 |
frame_init_noise_level,
|
161 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
162 |
|
163 |
+
global sample_idx
|
164 |
+
sample_idx += 1
|
165 |
+
save_sample_path = os.path.join(savedir_sample, f"{sample_idx}.mp4")
|
166 |
+
save_videos_grid(sample, save_sample_path, format="mp4")
|
167 |
|
168 |
+
sample_config = {
|
169 |
+
"prompt": prompt_textbox,
|
170 |
+
"n_prompt": negative_prompt_textbox,
|
171 |
+
"first_frame_path": input_image_path,
|
172 |
+
"sampler": sampler_dropdown,
|
173 |
+
"num_inference_steps": sample_step_slider,
|
174 |
+
"guidance_scale_text": txt_cfg_scale_slider,
|
175 |
+
"guidance_scale_image": img_cfg_scale_slider,
|
176 |
+
"width": width_slider,
|
177 |
+
"height": height_slider,
|
178 |
+
"video_length": 8,
|
179 |
+
"seed": seed
|
180 |
+
}
|
181 |
+
json_str = json.dumps(sample_config, indent=4)
|
182 |
+
with open(os.path.join(savedir, "logs.json"), "a") as f:
|
183 |
+
f.write(json_str)
|
184 |
+
f.write("\n\n")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
185 |
|
186 |
+
return gr.Video(value=save_sample_path)
|
187 |
+
|
188 |
|
189 |
+
@spaces.GPU
|
190 |
+
def run_pipeline(
|
191 |
+
pipeline,
|
192 |
+
prompt_textbox,
|
193 |
+
negative_prompt_textbox,
|
194 |
+
first_frame,
|
195 |
+
sample_step_slider,
|
196 |
+
width_slider,
|
197 |
+
height_slider,
|
198 |
+
txt_cfg_scale_slider,
|
199 |
+
img_cfg_scale_slider,
|
200 |
+
frame_stride,
|
201 |
+
use_frameinit,
|
202 |
+
frame_init_noise_level,
|
203 |
+
|
204 |
+
):
|
205 |
+
first_frame = first_frame.to("cuda")
|
206 |
+
sample = pipeline(
|
207 |
+
prompt_textbox,
|
208 |
+
negative_prompt = negative_prompt_textbox,
|
209 |
+
first_frames = first_frame,
|
210 |
+
num_inference_steps = sample_step_slider,
|
211 |
+
guidance_scale_txt = txt_cfg_scale_slider,
|
212 |
+
guidance_scale_img = img_cfg_scale_slider,
|
213 |
+
width = width_slider,
|
214 |
+
height = height_slider,
|
215 |
+
video_length = 16,
|
216 |
+
noise_sampling_method = "pyoco_mixed",
|
217 |
+
noise_alpha = 1.0,
|
218 |
+
frame_stride = frame_stride,
|
219 |
+
use_frameinit = use_frameinit,
|
220 |
+
frameinit_noise_level = frame_init_noise_level,
|
221 |
+
camera_motion = None,
|
222 |
+
).videos
|
223 |
+
return sample
|
224 |
|
225 |
|
226 |
def ui():
|
|
|
280 |
|
281 |
with gr.Row():
|
282 |
input_image = gr.Image(label="Input Image", interactive=True)
|
283 |
+
input_image.upload(fn=update_textbox_and_save_image, inputs=[input_image, height_slider, width_slider, center_crop], outputs=[input_image_path, input_image])
|
284 |
result_video = gr.Video(label="Generated Animation", interactive=False, autoplay=True)
|
285 |
|
286 |
def update_and_resize_image(input_image_path, height_slider, width_slider, center_crop):
|
|
|
288 |
pil_image = Image.open(requests.get(input_image_path, stream=True).raw).convert('RGB')
|
289 |
else:
|
290 |
pil_image = Image.open(input_image_path).convert('RGB')
|
|
|
291 |
original_width, original_height = pil_image.size
|
292 |
|
293 |
if center_crop:
|
|
|
315 |
input_image_path.submit(fn=update_and_resize_image, inputs=[input_image_path, height_slider, width_slider, center_crop], outputs=[input_image])
|
316 |
|
317 |
generate_button.click(
|
318 |
+
fn=animate,
|
319 |
inputs=[
|
320 |
prompt_textbox,
|
321 |
negative_prompt_textbox,
|