Spaces:
Paused
Paused
Update app-backup1.py
Browse files- app-backup1.py +567 -483
app-backup1.py
CHANGED
|
@@ -22,7 +22,7 @@ import os
|
|
| 22 |
import gc
|
| 23 |
from openai import OpenAI
|
| 24 |
import re
|
| 25 |
-
|
| 26 |
# Load system prompts
|
| 27 |
system_prompt_t2v = """λΉμ μ λΉλμ€ μμ±μ μν ν둬ννΈ μ λ¬Έκ°μ
λλ€.
|
| 28 |
μ£Όμ΄μ§ ν둬ννΈλ₯Ό λ€μ ꡬ쑰μ λ§κ² κ°μ ν΄μ£ΌμΈμ:
|
|
@@ -239,14 +239,7 @@ pipeline = XoraVideoPipeline(
|
|
| 239 |
vae=vae,
|
| 240 |
).to(device)
|
| 241 |
|
| 242 |
-
# State λ³μλ€μ μ΄κΈ°ν μμ
|
| 243 |
-
txt2vid_current_height = gr.State(value=320)
|
| 244 |
-
txt2vid_current_width = gr.State(value=512)
|
| 245 |
-
txt2vid_current_num_frames = gr.State(value=257)
|
| 246 |
|
| 247 |
-
img2vid_current_height = gr.State(value=320)
|
| 248 |
-
img2vid_current_width = gr.State(value=512)
|
| 249 |
-
img2vid_current_num_frames = gr.State(value=257)
|
| 250 |
|
| 251 |
# Preset options for resolution and frame configuration
|
| 252 |
# Convert frames to seconds assuming 25 FPS
|
|
@@ -279,27 +272,29 @@ preset_options = [
|
|
| 279 |
]
|
| 280 |
|
| 281 |
def preset_changed(preset):
|
| 282 |
-
selected = next(item for item in preset_options if item["label"] == preset)
|
|
|
|
|
|
|
| 283 |
return [
|
| 284 |
-
selected["height"],
|
| 285 |
-
selected["width"],
|
| 286 |
-
selected["num_frames"],
|
| 287 |
gr.update(visible=False),
|
| 288 |
gr.update(visible=False),
|
| 289 |
gr.update(visible=False),
|
| 290 |
-
]
|
| 291 |
-
|
| 292 |
def generate_video_from_text(
|
| 293 |
-
prompt
|
| 294 |
-
enhance_prompt_toggle
|
| 295 |
-
negative_prompt
|
| 296 |
-
frame_rate
|
| 297 |
-
seed
|
| 298 |
-
num_inference_steps
|
| 299 |
-
guidance_scale
|
| 300 |
-
height
|
| 301 |
-
width
|
| 302 |
-
num_frames
|
| 303 |
progress=gr.Progress(),
|
| 304 |
):
|
| 305 |
if len(prompt.strip()) < 50:
|
|
@@ -308,10 +303,23 @@ def generate_video_from_text(
|
|
| 308 |
duration=5,
|
| 309 |
)
|
| 310 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 311 |
# Translate Korean prompts to English
|
| 312 |
prompt = translate_korean_prompt(prompt)
|
| 313 |
negative_prompt = translate_korean_prompt(negative_prompt)
|
| 314 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 315 |
sample = {
|
| 316 |
"prompt": prompt,
|
| 317 |
"prompt_attention_mask": None,
|
|
@@ -354,7 +362,6 @@ def generate_video_from_text(
|
|
| 354 |
gc.collect()
|
| 355 |
|
| 356 |
output_path = tempfile.mktemp(suffix=".mp4")
|
| 357 |
-
print(images.shape)
|
| 358 |
video_np = images.squeeze(0).permute(1, 2, 3, 0).cpu().float().numpy()
|
| 359 |
video_np = (video_np * 255).astype(np.uint8)
|
| 360 |
height, width = video_np.shape[1:3]
|
|
@@ -371,21 +378,20 @@ def generate_video_from_text(
|
|
| 371 |
|
| 372 |
def generate_video_from_image(
|
| 373 |
image_path,
|
| 374 |
-
prompt
|
| 375 |
-
enhance_prompt_toggle
|
| 376 |
-
negative_prompt
|
| 377 |
-
frame_rate
|
| 378 |
-
seed
|
| 379 |
-
num_inference_steps
|
| 380 |
-
guidance_scale
|
| 381 |
-
height
|
| 382 |
-
width
|
| 383 |
-
num_frames
|
| 384 |
progress=gr.Progress(),
|
| 385 |
):
|
| 386 |
-
|
| 387 |
-
|
| 388 |
-
print("Num Frames: ", num_frames)
|
| 389 |
|
| 390 |
if len(prompt.strip()) < 50:
|
| 391 |
raise gr.Error(
|
|
@@ -393,13 +399,24 @@ def generate_video_from_image(
|
|
| 393 |
duration=5,
|
| 394 |
)
|
| 395 |
|
| 396 |
-
|
| 397 |
-
|
|
|
|
| 398 |
|
| 399 |
# Translate Korean prompts to English
|
| 400 |
prompt = translate_korean_prompt(prompt)
|
| 401 |
negative_prompt = translate_korean_prompt(negative_prompt)
|
| 402 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 403 |
media_items = (
|
| 404 |
load_image_to_tensor_with_resize(image_path, height, width).to(device).detach()
|
| 405 |
)
|
|
@@ -447,6 +464,7 @@ def generate_video_from_image(
|
|
| 447 |
for frame in video_np[..., ::-1]:
|
| 448 |
out.write(frame)
|
| 449 |
out.release()
|
|
|
|
| 450 |
except Exception as e:
|
| 451 |
raise gr.Error(
|
| 452 |
f"λΉλμ€ μμ± μ€ μ€λ₯κ° λ°μνμ΅λλ€. λ€μ μλν΄μ£ΌμΈμ. μ€λ₯: {e}",
|
|
@@ -456,6 +474,12 @@ def generate_video_from_image(
|
|
| 456 |
finally:
|
| 457 |
torch.cuda.empty_cache()
|
| 458 |
gc.collect()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 459 |
|
| 460 |
return output_path
|
| 461 |
|
|
@@ -551,99 +575,95 @@ system_prompt_scenario = """λΉμ μ μμ μ€ν¬λ¦½νΈμ λ§λ λ°°κ²½ μ
|
|
| 551 |
|
| 552 |
|
| 553 |
def analyze_scenario(scenario):
|
| 554 |
-
"""μλ리μ€λ₯Ό λΆμνμ¬ λ°°κ²½ μμμ© ν둬ννΈ μμ±"""
|
| 555 |
-
|
| 556 |
-
|
| 557 |
-
|
| 558 |
-
|
| 559 |
-
|
| 560 |
-
|
| 561 |
-
|
| 562 |
-
|
| 563 |
-
|
| 564 |
-
|
| 565 |
-
|
| 566 |
-
|
| 567 |
-
|
| 568 |
-
|
| 569 |
-
|
| 570 |
-
|
| 571 |
-
|
| 572 |
-
|
| 573 |
-
|
| 574 |
-
λͺ¨λ μΉμ
μ΄ μΌκ΄λ μ€νμΌκ³Ό ν€μ μ μ§νλ©΄μλ μμ°μ€λ½κ² μ΄μ΄μ§λλ‘ κ΅¬μ±νμΈμ.
|
| 575 |
-
|
| 576 |
-
κ° μΉμ
μ ν둬ννΈ μμ±μ λ°λμ λ€μ ꡬ쑰μ λ§κ² κ°μ ν΄μ£ΌμΈμ:
|
| 577 |
-
1. μ£Όμ λμμ λͺ
νν ν λ¬Έμ₯μΌλ‘ μμ
|
| 578 |
-
2. ꡬ체μ μΈ λμκ³Ό μ μ€μ²λ₯Ό μκ° μμλλ‘ μ€λͺ
|
| 579 |
-
3. μΊλ¦ν°/κ°μ²΄μ μΈλͺ¨λ₯Ό μμΈν λ¬μ¬
|
| 580 |
-
4. λ°°κ²½κ³Ό νκ²½ μΈλΆ μ¬νμ ꡬ체μ μΌλ‘ ν¬ν¨
|
| 581 |
-
5. μΉ΄λ©λΌ κ°λμ μμ§μμ λͺ
μ
|
| 582 |
-
6. μ‘°λͺ
κ³Ό μμμ μμΈν μ€λͺ
|
| 583 |
-
7. λ³νλ κ°μμ€λ¬μ΄ μ¬κ±΄μ μμ°μ€λ½κ² ν¬ν¨
|
| 584 |
-
λͺ¨λ μ€λͺ
μ νλμ μμ°μ€λ¬μ΄ λ¬Έλ¨μΌλ‘ μμ±νκ³ ,
|
| 585 |
-
촬μ κ°λ
μ΄ μ΄¬μ λͺ©λ‘μ μ€λͺ
νλ κ²μ²λΌ ꡬ체μ μ΄κ³ μκ°μ μΌλ‘ μμ±νμΈμ.
|
| 586 |
-
200λ¨μ΄λ₯Ό λμ§ μλλ‘ νλ, μ΅λν μμΈνκ² μμ±νμΈμ.
|
| 587 |
|
|
|
|
| 588 |
{scenario}
|
| 589 |
|
| 590 |
-
|
| 591 |
-
|
| 592 |
-
|
| 593 |
-
|
| 594 |
-
|
| 595 |
-
|
| 596 |
-
|
| 597 |
-
|
| 598 |
-
|
| 599 |
-
|
| 600 |
-
|
| 601 |
-
|
| 602 |
-
|
| 603 |
-
|
| 604 |
-
|
| 605 |
-
|
| 606 |
-
|
| 607 |
-
|
| 608 |
-
current_section = line
|
| 609 |
-
else:
|
| 610 |
-
current_section += "\n" + line
|
| 611 |
-
|
| 612 |
-
if current_section:
|
| 613 |
-
sections.append(current_section.strip())
|
| 614 |
|
| 615 |
-
|
| 616 |
-
|
| 617 |
-
sections.append("μΆκ° μΉμ
μ΄ νμν©λλ€.")
|
| 618 |
|
| 619 |
-
|
|
|
|
|
|
|
|
|
|
| 620 |
|
| 621 |
except Exception as e:
|
| 622 |
print(f"Error during scenario analysis: {e}")
|
| 623 |
return ["Error occurred during analysis"] * 5
|
| 624 |
|
| 625 |
def generate_section_video(prompt, preset, section_number=1, base_seed=171198, progress=gr.Progress()):
|
| 626 |
-
"""κ° μΉμ
μ λΉλμ€ μμ±
|
| 627 |
try:
|
| 628 |
if not prompt or len(prompt.strip()) < 50:
|
| 629 |
raise gr.Error("ν둬ννΈλ μ΅μ 50μ μ΄μμ΄μ΄μΌ ν©λλ€.")
|
| 630 |
|
| 631 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 632 |
section_seed = base_seed + section_number
|
| 633 |
|
| 634 |
return generate_video_from_text(
|
| 635 |
prompt=prompt,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 636 |
height=selected["height"],
|
| 637 |
width=selected["width"],
|
| 638 |
num_frames=selected["num_frames"],
|
| 639 |
-
seed=section_seed,
|
| 640 |
progress=progress
|
| 641 |
)
|
| 642 |
except Exception as e:
|
| 643 |
print(f"Error in section {section_number}: {e}")
|
| 644 |
raise gr.Error(f"μΉμ
{section_number} μμ± μ€ μ€λ₯: {str(e)}")
|
|
|
|
|
|
|
|
|
|
| 645 |
|
| 646 |
-
# κ°λ³ μΉμ
ν둬ννΈ μμ± ν¨μ μΆκ°
|
| 647 |
def generate_single_section_prompt(scenario, section_number):
|
| 648 |
"""κ°λ³ μΉμ
μ λν ν둬ννΈ μμ±"""
|
| 649 |
section_descriptions = {
|
|
@@ -658,23 +678,35 @@ def generate_single_section_prompt(scenario, section_number):
|
|
| 658 |
{"role": "system", "content": system_prompt_scenario},
|
| 659 |
{"role": "user", "content": f"""
|
| 660 |
λ€μ μ€ν¬λ¦½νΈμ {section_number}λ²μ§Έ μΉμ
({section_descriptions[section_number]})μ λν
|
| 661 |
-
λ°°κ²½ μμ
|
| 662 |
|
|
|
|
| 663 |
{scenario}
|
| 664 |
|
| 665 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 666 |
]
|
| 667 |
|
| 668 |
try:
|
| 669 |
response = client.chat.completions.create(
|
| 670 |
model="gpt-4-1106-preview",
|
| 671 |
messages=messages,
|
| 672 |
-
max_tokens=
|
|
|
|
| 673 |
)
|
| 674 |
-
|
|
|
|
| 675 |
except Exception as e:
|
| 676 |
-
print(f"Error during prompt generation: {e}")
|
| 677 |
-
return "Error occurred during prompt generation"
|
| 678 |
|
| 679 |
|
| 680 |
# λΉλμ€ κ²°ν© ν¨μ μΆκ°
|
|
@@ -716,11 +748,23 @@ def merge_section_videos(section1, section2, section3, section4, section5):
|
|
| 716 |
videos = []
|
| 717 |
|
| 718 |
# κ° μΉμ
λΉλμ€ νμΈ λ° μ²λ¦¬
|
| 719 |
-
for i,
|
| 720 |
-
if
|
| 721 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 722 |
else:
|
| 723 |
-
raise gr.Error(f"μΉμ
{i}μ μμμ΄
|
| 724 |
|
| 725 |
if not videos:
|
| 726 |
raise gr.Error("κ²°ν©ν μμμ΄ μμ΅λλ€.")
|
|
@@ -746,388 +790,428 @@ def merge_section_videos(section1, section2, section3, section4, section5):
|
|
| 746 |
ret, frame = cap.read()
|
| 747 |
if not ret:
|
| 748 |
break
|
| 749 |
-
#
|
| 750 |
if frame.shape[:2] != (height, width):
|
| 751 |
frame = cv2.resize(frame, (width, height))
|
| 752 |
out.write(frame)
|
| 753 |
cap.release()
|
| 754 |
|
| 755 |
out.release()
|
|
|
|
| 756 |
return output_path
|
| 757 |
|
| 758 |
except Exception as e:
|
| 759 |
raise gr.Error(f"λΉλμ€ κ²°ν© μ€ μ€λ₯ λ°μ: {e}")
|
| 760 |
|
| 761 |
-
|
| 762 |
-
|
| 763 |
-
|
| 764 |
-
|
| 765 |
-
# Text to Video Tab
|
| 766 |
-
with gr.TabItem("ν
μ€νΈλ‘ λΉλμ€ λ§λ€κΈ°"):
|
| 767 |
-
with gr.Row():
|
| 768 |
-
with gr.Column():
|
| 769 |
-
txt2vid_prompt = gr.Textbox(
|
| 770 |
-
label="Step 1: ν둬ννΈ μ
λ ₯",
|
| 771 |
-
placeholder="μμ±νκ³ μΆμ λΉλμ€λ₯Ό μ€λͺ
νμΈμ (μ΅μ 50μ)...",
|
| 772 |
-
value="κ·μ¬μ΄ κ³ μμ΄",
|
| 773 |
-
lines=5,
|
| 774 |
-
)
|
| 775 |
-
txt2vid_enhance_toggle = Toggle(
|
| 776 |
-
label="ν둬ννΈ κ°μ ",
|
| 777 |
-
value=False,
|
| 778 |
-
interactive=True,
|
| 779 |
-
)
|
| 780 |
-
|
| 781 |
-
txt2vid_negative_prompt = gr.Textbox(
|
| 782 |
-
label="Step 2: λ€κ±°ν°λΈ ν둬ννΈ μ
λ ₯",
|
| 783 |
-
placeholder="λΉλμ€μμ μνμ§ μλ μμλ₯Ό μ€λͺ
νμΈμ...",
|
| 784 |
-
value="low quality, worst quality, deformed, distorted, warped, motion smear, motion artifacts, fused fingers, incorrect anatomy, strange hands, unattractive",
|
| 785 |
-
lines=2,
|
| 786 |
-
visible=False
|
| 787 |
-
)
|
| 788 |
-
|
| 789 |
-
txt2vid_preset = gr.Dropdown(
|
| 790 |
-
choices=[p["label"] for p in preset_options],
|
| 791 |
-
value="[16:9] 512x320, 10.3μ΄",
|
| 792 |
-
label="Step 2: ν΄μλ ν리μ
μ ν",
|
| 793 |
-
)
|
| 794 |
-
|
| 795 |
-
txt2vid_frame_rate = gr.Slider(
|
| 796 |
-
label="Step 3: νλ μ λ μ΄νΈ",
|
| 797 |
-
minimum=21,
|
| 798 |
-
maximum=30,
|
| 799 |
-
step=1,
|
| 800 |
-
value=25,
|
| 801 |
-
visible=False
|
| 802 |
-
)
|
| 803 |
-
|
| 804 |
-
txt2vid_advanced = create_advanced_options()
|
| 805 |
-
txt2vid_generate = gr.Button(
|
| 806 |
-
"Step 3: λΉλμ€ μμ±",
|
| 807 |
-
variant="primary",
|
| 808 |
-
size="lg",
|
| 809 |
-
)
|
| 810 |
-
|
| 811 |
-
with gr.Column():
|
| 812 |
-
txt2vid_output = gr.Video(label="μμ±λ λΉλμ€")
|
| 813 |
-
|
| 814 |
-
# Image to Video Tab
|
| 815 |
-
with gr.TabItem("μ΄λ―Έμ§λ‘ λΉλμ€ λ§λ€κΈ°"):
|
| 816 |
-
with gr.Row():
|
| 817 |
-
with gr.Column():
|
| 818 |
-
img2vid_image = gr.Image(
|
| 819 |
-
type="filepath",
|
| 820 |
-
label="Step 1: μ
λ ₯ μ΄λ―Έμ§ μ
λ‘λ",
|
| 821 |
-
elem_id="image_upload",
|
| 822 |
-
)
|
| 823 |
-
img2vid_prompt = gr.Textbox(
|
| 824 |
-
label="Step 2: ν둬ννΈ μ
λ ₯",
|
| 825 |
-
placeholder="μ΄λ―Έμ§λ₯Ό μ΄λ»κ² μ λλ©μ΄μ
νν μ§ μ€λͺ
νμΈμ (μ΅μ 50μ)...",
|
| 826 |
-
value="κ·μ¬μ΄ κ³ μμ΄",
|
| 827 |
-
lines=5,
|
| 828 |
-
)
|
| 829 |
-
img2vid_enhance_toggle = Toggle(
|
| 830 |
-
label="ν둬ννΈ μ¦κ°",
|
| 831 |
-
value=False,
|
| 832 |
-
interactive=True,
|
| 833 |
-
)
|
| 834 |
-
img2vid_negative_prompt = gr.Textbox(
|
| 835 |
-
label="Step 3: λ€κ±°ν°λΈ ν둬ννΈ μ
λ ₯",
|
| 836 |
-
placeholder="λΉλμ€μμ μνμ§ μλ μμλ₯Ό μ€λͺ
νμΈμ...",
|
| 837 |
-
value="low quality, worst quality, deformed, distorted, warped, motion smear, motion artifacts, fused fingers, incorrect anatomy, strange hands, unattractive",
|
| 838 |
-
lines=2,
|
| 839 |
-
visible=False
|
| 840 |
-
)
|
| 841 |
-
|
| 842 |
-
img2vid_preset = gr.Dropdown(
|
| 843 |
-
choices=[p["label"] for p in preset_options],
|
| 844 |
-
value="[16:9] 512x320, 10.3μ΄",
|
| 845 |
-
label="Step 3: ν΄μλ ν리μ
μ ν",
|
| 846 |
-
)
|
| 847 |
-
|
| 848 |
-
img2vid_frame_rate = gr.Slider(
|
| 849 |
-
label="Step 4: νλ μ λ μ΄νΈ",
|
| 850 |
-
minimum=21,
|
| 851 |
-
maximum=30,
|
| 852 |
-
step=1,
|
| 853 |
-
value=25,
|
| 854 |
-
visible=False
|
| 855 |
-
)
|
| 856 |
-
|
| 857 |
-
img2vid_advanced = create_advanced_options()
|
| 858 |
-
img2vid_generate = gr.Button(
|
| 859 |
-
"Step 4: λΉλμ€ μμ±",
|
| 860 |
-
variant="primary",
|
| 861 |
-
size="lg",
|
| 862 |
-
)
|
| 863 |
-
|
| 864 |
-
with gr.Column():
|
| 865 |
-
img2vid_output = gr.Video(label="μμ±λ λΉλμ€")
|
| 866 |
-
|
| 867 |
-
|
| 868 |
-
# Scenario to Video Tab (Modified)
|
| 869 |
-
with gr.TabItem("μλ리μ€λ‘ λΉλμ€ λ§λ€κΈ°(μνΌ)"):
|
| 870 |
-
with gr.Row():
|
| 871 |
-
with gr.Column(scale=1):
|
| 872 |
-
scenario_input = gr.Textbox(
|
| 873 |
-
label="μμ μ€ν¬λ¦½νΈ μ
λ ₯",
|
| 874 |
-
placeholder="μ 체 μλ리μ€λ₯Ό μ
λ ₯νμΈμ...",
|
| 875 |
-
lines=10
|
| 876 |
-
)
|
| 877 |
-
scenario_preset = gr.Dropdown(
|
| 878 |
-
choices=[p["label"] for p in preset_options],
|
| 879 |
-
value="[16:9] 512x320, 10.3μ΄",
|
| 880 |
-
label="νλ©΄ ν¬κΈ° μ ν"
|
| 881 |
-
)
|
| 882 |
-
analyze_btn = gr.Button("μλλ¦¬μ€ λΆμ λ° ν둬ννΈ μμ±", variant="primary")
|
| 883 |
-
|
| 884 |
-
with gr.Column(scale=2):
|
| 885 |
-
with gr.Row():
|
| 886 |
-
# μΉμ
1
|
| 887 |
-
with gr.Column():
|
| 888 |
-
section1_prompt = gr.Textbox(
|
| 889 |
-
label="1. λ°°κ²½ λ° νμμ±",
|
| 890 |
-
lines=4
|
| 891 |
-
)
|
| 892 |
-
with gr.Row():
|
| 893 |
-
section1_regenerate = gr.Button("π ν둬ννΈ μμ±")
|
| 894 |
-
section1_generate = gr.Button("π μμ μμ±")
|
| 895 |
-
section1_video = gr.Video(label="μΉμ
1 μμ")
|
| 896 |
-
|
| 897 |
-
# μΉμ
2
|
| 898 |
-
with gr.Column():
|
| 899 |
-
section2_prompt = gr.Textbox(
|
| 900 |
-
label="2. ν₯λ―Έ μ λ°",
|
| 901 |
-
lines=4
|
| 902 |
-
)
|
| 903 |
-
with gr.Row():
|
| 904 |
-
section2_regenerate = gr.Button("π ν둬ννΈ μμ±")
|
| 905 |
-
section2_generate = gr.Button("π μμ μμ±")
|
| 906 |
-
section2_video = gr.Video(label="μΉμ
2 μμ")
|
| 907 |
-
|
| 908 |
-
with gr.Row():
|
| 909 |
-
# μΉμ
3
|
| 910 |
-
with gr.Column():
|
| 911 |
-
section3_prompt = gr.Textbox(
|
| 912 |
-
label="3. ν΄κ²°μ±
μ μ",
|
| 913 |
-
lines=4
|
| 914 |
-
)
|
| 915 |
-
with gr.Row():
|
| 916 |
-
section3_regenerate = gr.Button("π ν둬ννΈ μμ±")
|
| 917 |
-
section3_generate = gr.Button("π μμ μμ±")
|
| 918 |
-
section3_video = gr.Video(label="μΉμ
3 μμ")
|
| 919 |
-
|
| 920 |
-
# μΉμ
4
|
| 921 |
-
with gr.Column():
|
| 922 |
-
section4_prompt = gr.Textbox(
|
| 923 |
-
label="4. λ³Έλ‘ ",
|
| 924 |
-
lines=4
|
| 925 |
-
)
|
| 926 |
-
with gr.Row():
|
| 927 |
-
section4_regenerate = gr.Button("π ν둬ννΈ μμ±")
|
| 928 |
-
section4_generate = gr.Button("π μμ μμ±")
|
| 929 |
-
section4_video = gr.Video(label="μΉμ
4 μμ")
|
| 930 |
-
|
| 931 |
-
with gr.Row():
|
| 932 |
-
# μΉμ
5
|
| 933 |
-
with gr.Column():
|
| 934 |
-
section5_prompt = gr.Textbox(
|
| 935 |
-
label="5. κ²°λ‘ λ° κ°μ‘°",
|
| 936 |
-
lines=4
|
| 937 |
-
)
|
| 938 |
-
with gr.Row():
|
| 939 |
-
section5_regenerate = gr.Button("π ν둬ννΈ μμ±")
|
| 940 |
-
section5_generate = gr.Button("π μμ μμ±")
|
| 941 |
-
section5_video = gr.Video(label="μΉμ
5 μμ")
|
| 942 |
-
|
| 943 |
-
|
| 944 |
-
|
| 945 |
-
# ν΅ν© μμ μΉμ
μΆκ°
|
| 946 |
-
with gr.Row():
|
| 947 |
-
with gr.Column(scale=1):
|
| 948 |
-
# κΈ°μ‘΄μ scenario_inputκ³Ό analyze_btn μ μ§
|
| 949 |
-
merge_videos_btn = gr.Button("ν΅ν© μμ μμ±", variant="primary", size="lg")
|
| 950 |
|
| 951 |
-
|
| 952 |
-
|
| 953 |
-
|
| 954 |
-
|
| 955 |
-
|
| 956 |
-
|
| 957 |
-
|
| 958 |
-
|
| 959 |
-
|
| 960 |
-
|
| 961 |
-
|
| 962 |
-
|
| 963 |
-
|
| 964 |
-
|
| 965 |
-
|
| 966 |
-
txt2vid_current_height,
|
| 967 |
-
txt2vid_current_width,
|
| 968 |
-
txt2vid_current_num_frames,
|
| 969 |
-
*txt2vid_advanced[3:]
|
| 970 |
-
]
|
| 971 |
-
)
|
| 972 |
-
|
| 973 |
-
txt2vid_enhance_toggle.change(
|
| 974 |
-
fn=update_prompt_t2v,
|
| 975 |
-
inputs=[txt2vid_prompt, txt2vid_enhance_toggle],
|
| 976 |
-
outputs=txt2vid_prompt
|
| 977 |
-
)
|
| 978 |
-
|
| 979 |
-
txt2vid_generate.click(
|
| 980 |
-
fn=generate_video_from_text,
|
| 981 |
-
inputs=[
|
| 982 |
-
txt2vid_prompt,
|
| 983 |
-
txt2vid_enhance_toggle,
|
| 984 |
-
txt2vid_negative_prompt,
|
| 985 |
-
txt2vid_frame_rate,
|
| 986 |
-
*txt2vid_advanced[:3],
|
| 987 |
-
txt2vid_current_height,
|
| 988 |
-
txt2vid_current_width,
|
| 989 |
-
txt2vid_current_num_frames,
|
| 990 |
-
],
|
| 991 |
-
outputs=txt2vid_output,
|
| 992 |
-
concurrency_limit=1,
|
| 993 |
-
concurrency_id="generate_video",
|
| 994 |
-
queue=True,
|
| 995 |
-
)
|
| 996 |
-
|
| 997 |
-
img2vid_preset.change(
|
| 998 |
-
fn=preset_changed,
|
| 999 |
-
inputs=[img2vid_preset],
|
| 1000 |
-
outputs=[
|
| 1001 |
-
img2vid_current_height,
|
| 1002 |
-
img2vid_current_width,
|
| 1003 |
-
img2vid_current_num_frames,
|
| 1004 |
-
*img2vid_advanced[3:]
|
| 1005 |
-
]
|
| 1006 |
-
)
|
| 1007 |
-
|
| 1008 |
-
img2vid_enhance_toggle.change(
|
| 1009 |
-
fn=update_prompt_i2v,
|
| 1010 |
-
inputs=[img2vid_prompt, img2vid_enhance_toggle],
|
| 1011 |
-
outputs=img2vid_prompt
|
| 1012 |
-
)
|
| 1013 |
-
|
| 1014 |
-
img2vid_generate.click(
|
| 1015 |
-
fn=generate_video_from_image,
|
| 1016 |
-
inputs=[
|
| 1017 |
-
img2vid_image,
|
| 1018 |
-
img2vid_prompt,
|
| 1019 |
-
img2vid_enhance_toggle,
|
| 1020 |
-
img2vid_negative_prompt,
|
| 1021 |
-
img2vid_frame_rate,
|
| 1022 |
-
*img2vid_advanced[:3],
|
| 1023 |
-
img2vid_current_height,
|
| 1024 |
-
img2vid_current_width,
|
| 1025 |
-
img2vid_current_num_frames,
|
| 1026 |
-
],
|
| 1027 |
-
outputs=img2vid_output,
|
| 1028 |
-
concurrency_limit=1,
|
| 1029 |
-
concurrency_id="generate_video",
|
| 1030 |
-
queue=True,
|
| 1031 |
-
)
|
| 1032 |
-
|
| 1033 |
-
# Scenario tab event handlers
|
| 1034 |
-
analyze_btn.click(
|
| 1035 |
-
fn=analyze_scenario,
|
| 1036 |
-
inputs=[scenario_input],
|
| 1037 |
-
outputs=[
|
| 1038 |
-
section1_prompt, section2_prompt, section3_prompt,
|
| 1039 |
-
section4_prompt, section5_prompt
|
| 1040 |
-
]
|
| 1041 |
-
)
|
| 1042 |
-
|
| 1043 |
-
|
| 1044 |
-
# κ° μΉμ
μ ν둬ννΈ μ¬μμ± μ΄λ²€νΈ νΈλ€λ¬ μΆκ°
|
| 1045 |
-
section1_regenerate.click(
|
| 1046 |
-
fn=lambda x: generate_single_section_prompt(x, 1),
|
| 1047 |
-
inputs=[scenario_input],
|
| 1048 |
-
outputs=section1_prompt
|
| 1049 |
-
)
|
| 1050 |
-
|
| 1051 |
-
section2_regenerate.click(
|
| 1052 |
-
fn=lambda x: generate_single_section_prompt(x, 2),
|
| 1053 |
-
inputs=[scenario_input],
|
| 1054 |
-
outputs=section2_prompt
|
| 1055 |
-
)
|
| 1056 |
-
|
| 1057 |
-
section3_regenerate.click(
|
| 1058 |
-
fn=lambda x: generate_single_section_prompt(x, 3),
|
| 1059 |
-
inputs=[scenario_input],
|
| 1060 |
-
outputs=section3_prompt
|
| 1061 |
-
)
|
| 1062 |
-
|
| 1063 |
-
section4_regenerate.click(
|
| 1064 |
-
fn=lambda x: generate_single_section_prompt(x, 4),
|
| 1065 |
-
inputs=[scenario_input],
|
| 1066 |
-
outputs=section4_prompt
|
| 1067 |
-
)
|
| 1068 |
|
| 1069 |
-
|
| 1070 |
-
|
| 1071 |
-
|
| 1072 |
-
|
| 1073 |
-
|
| 1074 |
-
|
| 1075 |
-
|
| 1076 |
-
|
| 1077 |
-
|
| 1078 |
-
|
| 1079 |
-
|
| 1080 |
-
outputs=section1_video,
|
| 1081 |
-
api_name=f"generate_section1"
|
| 1082 |
-
)
|
| 1083 |
-
|
| 1084 |
-
section2_generate.click(
|
| 1085 |
-
fn=lambda p, pr: generate_section_video(p, pr, 2),
|
| 1086 |
-
inputs=[section2_prompt, scenario_preset],
|
| 1087 |
-
outputs=section2_video,
|
| 1088 |
-
api_name=f"generate_section2"
|
| 1089 |
-
)
|
| 1090 |
-
|
| 1091 |
-
section3_generate.click(
|
| 1092 |
-
fn=lambda p, pr: generate_section_video(p, pr, 3),
|
| 1093 |
-
inputs=[section3_prompt, scenario_preset],
|
| 1094 |
-
outputs=section3_video,
|
| 1095 |
-
api_name=f"generate_section3"
|
| 1096 |
-
)
|
| 1097 |
-
|
| 1098 |
-
section4_generate.click(
|
| 1099 |
-
fn=lambda p, pr: generate_section_video(p, pr, 4),
|
| 1100 |
-
inputs=[section4_prompt, scenario_preset],
|
| 1101 |
-
outputs=section4_video,
|
| 1102 |
-
api_name=f"generate_section4"
|
| 1103 |
-
)
|
| 1104 |
-
|
| 1105 |
-
section5_generate.click(
|
| 1106 |
-
fn=lambda p, pr: generate_section_video(p, pr, 5),
|
| 1107 |
-
inputs=[section5_prompt, scenario_preset],
|
| 1108 |
-
outputs=section5_video,
|
| 1109 |
-
api_name=f"generate_section5"
|
| 1110 |
-
)
|
| 1111 |
-
|
| 1112 |
-
|
| 1113 |
-
|
| 1114 |
|
| 1115 |
|
| 1116 |
-
|
| 1117 |
-
|
| 1118 |
-
|
| 1119 |
-
|
| 1120 |
-
|
| 1121 |
-
|
| 1122 |
-
|
| 1123 |
-
|
| 1124 |
-
|
| 1125 |
-
|
| 1126 |
-
|
| 1127 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1128 |
|
| 1129 |
-
|
| 1130 |
if __name__ == "__main__":
|
| 1131 |
-
|
| 1132 |
-
|
| 1133 |
-
|
|
|
|
|
|
| 22 |
import gc
|
| 23 |
from openai import OpenAI
|
| 24 |
import re
|
| 25 |
+
import time
|
| 26 |
# Load system prompts
|
| 27 |
system_prompt_t2v = """λΉμ μ λΉλμ€ μμ±μ μν ν둬ννΈ μ λ¬Έκ°μ
λλ€.
|
| 28 |
μ£Όμ΄μ§ ν둬ννΈλ₯Ό λ€μ ꡬ쑰μ λ§κ² κ°μ ν΄μ£ΌμΈμ:
|
|
|
|
| 239 |
vae=vae,
|
| 240 |
).to(device)
|
| 241 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 242 |
|
|
|
|
|
|
|
|
|
|
| 243 |
|
| 244 |
# Preset options for resolution and frame configuration
|
| 245 |
# Convert frames to seconds assuming 25 FPS
|
|
|
|
| 272 |
]
|
| 273 |
|
| 274 |
def preset_changed(preset):
|
| 275 |
+
selected = next((item for item in preset_options if item["label"] == preset), None)
|
| 276 |
+
if selected is None:
|
| 277 |
+
raise gr.Error("Invalid preset selected")
|
| 278 |
return [
|
| 279 |
+
gr.State(value=selected["height"]),
|
| 280 |
+
gr.State(value=selected["width"]),
|
| 281 |
+
gr.State(value=selected["num_frames"]),
|
| 282 |
gr.update(visible=False),
|
| 283 |
gr.update(visible=False),
|
| 284 |
gr.update(visible=False),
|
| 285 |
+
]
|
| 286 |
+
|
| 287 |
def generate_video_from_text(
|
| 288 |
+
prompt,
|
| 289 |
+
enhance_prompt_toggle,
|
| 290 |
+
negative_prompt,
|
| 291 |
+
frame_rate,
|
| 292 |
+
seed,
|
| 293 |
+
num_inference_steps,
|
| 294 |
+
guidance_scale,
|
| 295 |
+
height,
|
| 296 |
+
width,
|
| 297 |
+
num_frames,
|
| 298 |
progress=gr.Progress(),
|
| 299 |
):
|
| 300 |
if len(prompt.strip()) < 50:
|
|
|
|
| 303 |
duration=5,
|
| 304 |
)
|
| 305 |
|
| 306 |
+
# ν둬ννΈ κ°μ μ΄ νμ±νλ κ²½μ°
|
| 307 |
+
if enhance_prompt_toggle:
|
| 308 |
+
prompt = enhance_prompt(prompt, "t2v")
|
| 309 |
+
|
| 310 |
# Translate Korean prompts to English
|
| 311 |
prompt = translate_korean_prompt(prompt)
|
| 312 |
negative_prompt = translate_korean_prompt(negative_prompt)
|
| 313 |
|
| 314 |
+
# κΈ°λ³Έκ° μ€μ
|
| 315 |
+
height = height or 320
|
| 316 |
+
width = width or 512
|
| 317 |
+
num_frames = num_frames or 257
|
| 318 |
+
frame_rate = frame_rate or 25
|
| 319 |
+
seed = seed or 171198
|
| 320 |
+
num_inference_steps = num_inference_steps or 41
|
| 321 |
+
guidance_scale = guidance_scale or 4.0
|
| 322 |
+
|
| 323 |
sample = {
|
| 324 |
"prompt": prompt,
|
| 325 |
"prompt_attention_mask": None,
|
|
|
|
| 362 |
gc.collect()
|
| 363 |
|
| 364 |
output_path = tempfile.mktemp(suffix=".mp4")
|
|
|
|
| 365 |
video_np = images.squeeze(0).permute(1, 2, 3, 0).cpu().float().numpy()
|
| 366 |
video_np = (video_np * 255).astype(np.uint8)
|
| 367 |
height, width = video_np.shape[1:3]
|
|
|
|
| 378 |
|
| 379 |
def generate_video_from_image(
|
| 380 |
image_path,
|
| 381 |
+
prompt,
|
| 382 |
+
enhance_prompt_toggle,
|
| 383 |
+
negative_prompt,
|
| 384 |
+
frame_rate,
|
| 385 |
+
seed,
|
| 386 |
+
num_inference_steps,
|
| 387 |
+
guidance_scale,
|
| 388 |
+
height,
|
| 389 |
+
width,
|
| 390 |
+
num_frames,
|
| 391 |
progress=gr.Progress(),
|
| 392 |
):
|
| 393 |
+
if not image_path:
|
| 394 |
+
raise gr.Error("μ
λ ₯ μ΄λ―Έμ§λ₯Ό μ 곡ν΄μ£ΌμΈμ.", duration=5)
|
|
|
|
| 395 |
|
| 396 |
if len(prompt.strip()) < 50:
|
| 397 |
raise gr.Error(
|
|
|
|
| 399 |
duration=5,
|
| 400 |
)
|
| 401 |
|
| 402 |
+
# ν둬ννΈ κ°μ μ΄ νμ±νλ κ²½μ°
|
| 403 |
+
if enhance_prompt_toggle:
|
| 404 |
+
prompt = enhance_prompt(prompt, "i2v")
|
| 405 |
|
| 406 |
# Translate Korean prompts to English
|
| 407 |
prompt = translate_korean_prompt(prompt)
|
| 408 |
negative_prompt = translate_korean_prompt(negative_prompt)
|
| 409 |
|
| 410 |
+
# κΈ°λ³Έκ° μ€μ
|
| 411 |
+
height = height or 320
|
| 412 |
+
width = width or 512
|
| 413 |
+
num_frames = num_frames or 257
|
| 414 |
+
frame_rate = frame_rate or 25
|
| 415 |
+
seed = seed or 171198
|
| 416 |
+
num_inference_steps = num_inference_steps or 41
|
| 417 |
+
guidance_scale = guidance_scale or 4.0
|
| 418 |
+
|
| 419 |
+
# μ΄λ―Έμ§ λ‘λ λ° μ μ²λ¦¬
|
| 420 |
media_items = (
|
| 421 |
load_image_to_tensor_with_resize(image_path, height, width).to(device).detach()
|
| 422 |
)
|
|
|
|
| 464 |
for frame in video_np[..., ::-1]:
|
| 465 |
out.write(frame)
|
| 466 |
out.release()
|
| 467 |
+
|
| 468 |
except Exception as e:
|
| 469 |
raise gr.Error(
|
| 470 |
f"λΉλμ€ μμ± μ€ μ€λ₯κ° λ°μνμ΅λλ€. λ€μ μλν΄μ£ΌμΈμ. μ€λ₯: {e}",
|
|
|
|
| 474 |
finally:
|
| 475 |
torch.cuda.empty_cache()
|
| 476 |
gc.collect()
|
| 477 |
+
if 'images' in locals():
|
| 478 |
+
del images
|
| 479 |
+
if 'video_np' in locals():
|
| 480 |
+
del video_np
|
| 481 |
+
if 'media_items' in locals():
|
| 482 |
+
del media_items
|
| 483 |
|
| 484 |
return output_path
|
| 485 |
|
|
|
|
| 575 |
|
| 576 |
|
| 577 |
def analyze_scenario(scenario):
|
| 578 |
+
"""μλ리μ€λ₯Ό λΆμνμ¬ κ° μΉμ
λ³ λ°°κ²½ μμμ© ν둬ννΈ μμ±"""
|
| 579 |
+
try:
|
| 580 |
+
# κ° μΉμ
λ³ ν둬ννΈ μμ±μ μν λ©μμ§ κ΅¬μ±
|
| 581 |
+
section_prompts = []
|
| 582 |
+
|
| 583 |
+
for section_num in range(1, 6):
|
| 584 |
+
section_descriptions = {
|
| 585 |
+
1: "λ°°κ²½ λ° νμμ±: μ£Όμ μ μ λ°μ μΈ λΆμκΈ°λ₯Ό νννλ λ°°κ²½ μ¬",
|
| 586 |
+
2: "ν₯λ―Έ μ λ°: κΈ΄μ₯κ°μ΄λ κ°λ±μ μμνλ λΆμκΈ° μλ λ°°κ²½",
|
| 587 |
+
3: "ν΄κ²°μ±
μ μ: ν¬λ§μ μ΄κ³ λ°μ ν€μ λ°°κ²½ μ ν",
|
| 588 |
+
4: "λ³Έλ‘ : μμ κ° μκ³ μ λ’°λλ₯Ό λμ΄λ λ°°κ²½",
|
| 589 |
+
5: "κ²°λ‘ : μν©νΈ μλ λ§λ¬΄λ¦¬λ₯Ό μν μλμ μΈ λ°°κ²½"
|
| 590 |
+
}
|
| 591 |
+
|
| 592 |
+
messages = [
|
| 593 |
+
{"role": "system", "content": system_prompt_scenario},
|
| 594 |
+
{"role": "user", "content": f"""
|
| 595 |
+
λ€μ μ€ν¬λ¦½νΈμ {section_num}λ²μ§Έ μΉμ
({section_descriptions[section_num]})μ λν
|
| 596 |
+
λ°°κ²½ μμ ν둬ννΈλ₯Ό μμ±ν΄μ£ΌμΈμ.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 597 |
|
| 598 |
+
μ€ν¬λ¦½νΈ:
|
| 599 |
{scenario}
|
| 600 |
|
| 601 |
+
μ£Όμμ¬ν:
|
| 602 |
+
1. ν΄λΉ μΉμ
μ νΉμ±({section_descriptions[section_num]})μ λ§λ λΆμκΈ°μ ν€μ λ°μνμΈμ.
|
| 603 |
+
2. μ§μ μ μΈ μ ν/μλΉμ€ λ¬μ¬λ νΌνκ³ , κ°μ±μ μ΄κ³ μμ μ μΈ λ°°κ²½ μμμ μ§μ€νμΈμ.
|
| 604 |
+
3. λ€μ ꡬ쑰λ₯Ό λ°λμ ν¬ν¨νμΈμ:
|
| 605 |
+
- μ£Όμ λμμ λͺ
νν ν λ¬Έμ₯μΌλ‘ μμ
|
| 606 |
+
- ꡬ체μ μΈ λμκ³Ό οΏ½οΏ½οΏ½μ€μ²λ₯Ό μκ° μμλλ‘ μ€λͺ
|
| 607 |
+
- λ°°κ²½κ³Ό νκ²½ μΈλΆ μ¬νμ ꡬ체μ μΌλ‘ ν¬ν¨
|
| 608 |
+
- μΉ΄λ©λΌ κ°λμ μμ§μμ λͺ
μ
|
| 609 |
+
- μ‘°λͺ
κ³Ό μμμ μμΈν μ€λͺ
|
| 610 |
+
- λ³νλ κ°μμ€λ¬μ΄ μ¬κ±΄μ μμ°μ€λ½κ² ν¬ν¨"""}
|
| 611 |
+
]
|
| 612 |
+
|
| 613 |
+
response = client.chat.completions.create(
|
| 614 |
+
model="gpt-4-1106-preview",
|
| 615 |
+
messages=messages,
|
| 616 |
+
max_tokens=1000,
|
| 617 |
+
temperature=0.7
|
| 618 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 619 |
|
| 620 |
+
section_prompt = response.choices[0].message.content.strip()
|
| 621 |
+
section_prompts.append(f"{section_num}. {section_prompt}")
|
|
|
|
| 622 |
|
| 623 |
+
# API μμ² μ¬μ΄μ μ§§μ λλ μ΄ μΆκ°
|
| 624 |
+
time.sleep(1)
|
| 625 |
+
|
| 626 |
+
return section_prompts
|
| 627 |
|
| 628 |
except Exception as e:
|
| 629 |
print(f"Error during scenario analysis: {e}")
|
| 630 |
return ["Error occurred during analysis"] * 5
|
| 631 |
|
| 632 |
def generate_section_video(prompt, preset, section_number=1, base_seed=171198, progress=gr.Progress()):
|
| 633 |
+
"""κ° μΉμ
μ λΉλμ€ μμ±"""
|
| 634 |
try:
|
| 635 |
if not prompt or len(prompt.strip()) < 50:
|
| 636 |
raise gr.Error("ν둬ννΈλ μ΅μ 50μ μ΄μμ΄μ΄μΌ ν©λλ€.")
|
| 637 |
|
| 638 |
+
if not preset:
|
| 639 |
+
raise gr.Error("ν΄μλ ν리μ
μ μ νν΄μ£ΌμΈμ.")
|
| 640 |
+
|
| 641 |
+
selected = next((item for item in preset_options if item["label"] == preset), None)
|
| 642 |
+
if not selected:
|
| 643 |
+
raise gr.Error("μ¬λ°λ₯΄μ§ μμ ν리μ
μ
λλ€.")
|
| 644 |
+
|
| 645 |
section_seed = base_seed + section_number
|
| 646 |
|
| 647 |
return generate_video_from_text(
|
| 648 |
prompt=prompt,
|
| 649 |
+
enhance_prompt_toggle=False, # μΉμ
μμ±μλ ν둬ννΈ μ¦κ° λΉνμ±ν
|
| 650 |
+
negative_prompt="low quality, worst quality, deformed, distorted, warped",
|
| 651 |
+
frame_rate=25,
|
| 652 |
+
seed=section_seed,
|
| 653 |
+
num_inference_steps=41,
|
| 654 |
+
guidance_scale=4.0,
|
| 655 |
height=selected["height"],
|
| 656 |
width=selected["width"],
|
| 657 |
num_frames=selected["num_frames"],
|
|
|
|
| 658 |
progress=progress
|
| 659 |
)
|
| 660 |
except Exception as e:
|
| 661 |
print(f"Error in section {section_number}: {e}")
|
| 662 |
raise gr.Error(f"μΉμ
{section_number} μμ± μ€ μ€λ₯: {str(e)}")
|
| 663 |
+
finally:
|
| 664 |
+
torch.cuda.empty_cache()
|
| 665 |
+
gc.collect()
|
| 666 |
|
|
|
|
| 667 |
def generate_single_section_prompt(scenario, section_number):
|
| 668 |
"""κ°λ³ μΉμ
μ λν ν둬ννΈ μμ±"""
|
| 669 |
section_descriptions = {
|
|
|
|
| 678 |
{"role": "system", "content": system_prompt_scenario},
|
| 679 |
{"role": "user", "content": f"""
|
| 680 |
λ€μ μ€ν¬λ¦½νΈμ {section_number}λ²μ§Έ μΉμ
({section_descriptions[section_number]})μ λν
|
| 681 |
+
λ°°κ²½ μμ ν둬ννΈλ₯Ό μμ±ν΄μ£ΌμΈμ.
|
| 682 |
|
| 683 |
+
μ€ν¬λ¦½νΈ:
|
| 684 |
{scenario}
|
| 685 |
|
| 686 |
+
μ£Όμμ¬ν:
|
| 687 |
+
1. ν΄λΉ μΉμ
μ νΉμ±({section_descriptions[section_number]})μ λ§λ λΆμκΈ°μ ν€μ λ°μνμΈμ.
|
| 688 |
+
2. μ§μ μ μΈ μ ν/μλΉμ€ λ¬μ¬λ νΌνκ³ , κ°μ±μ μ΄κ³ μμ μ μΈ λ°°κ²½ μμμ μ§μ€νμΈμ.
|
| 689 |
+
3. λ€μ ꡬ쑰λ₯Ό λ°λμ ν¬ν¨νμΈμ:
|
| 690 |
+
- μ£Όμ λμμ λͺ
νν ν λ¬Έμ₯μΌλ‘ μμ
|
| 691 |
+
- ꡬ체μ μΈ λμκ³Ό μ μ€μ²λ₯Ό μκ° μμλλ‘ μ€λͺ
|
| 692 |
+
- λ°°κ²½κ³Ό νκ²½ μΈλΆ μ¬νμ ꡬ체μ μΌλ‘ ν¬ν¨
|
| 693 |
+
- μΉ΄λ©λΌ κ°λμ μμ§μμ λͺ
μ
|
| 694 |
+
- μ‘°λͺ
κ³Ό μμμ μμΈν μ€λͺ
|
| 695 |
+
- λ³νλ κ°μμ€λ¬μ΄ μ¬κ±΄μ μμ°μ€λ½κ² ν¬ν¨"""}
|
| 696 |
]
|
| 697 |
|
| 698 |
try:
|
| 699 |
response = client.chat.completions.create(
|
| 700 |
model="gpt-4-1106-preview",
|
| 701 |
messages=messages,
|
| 702 |
+
max_tokens=1000, # ν ν° μ μ¦κ°
|
| 703 |
+
temperature=0.7
|
| 704 |
)
|
| 705 |
+
generated_prompt = response.choices[0].message.content.strip()
|
| 706 |
+
return f"{section_number}. {generated_prompt}"
|
| 707 |
except Exception as e:
|
| 708 |
+
print(f"Error during prompt generation for section {section_number}: {e}")
|
| 709 |
+
return f"Error occurred during prompt generation for section {section_number}"
|
| 710 |
|
| 711 |
|
| 712 |
# λΉλμ€ κ²°ν© ν¨μ μΆκ°
|
|
|
|
| 748 |
videos = []
|
| 749 |
|
| 750 |
# κ° μΉμ
λΉλμ€ νμΈ λ° μ²λ¦¬
|
| 751 |
+
for i, video_path in enumerate([section1, section2, section3, section4, section5], 1):
|
| 752 |
+
if video_path:
|
| 753 |
+
if os.path.exists(video_path):
|
| 754 |
+
try:
|
| 755 |
+
# λΉλμ€ νμΌ κ²μ¦
|
| 756 |
+
cap = cv2.VideoCapture(video_path)
|
| 757 |
+
if cap.isOpened():
|
| 758 |
+
videos.append(video_path)
|
| 759 |
+
cap.release()
|
| 760 |
+
else:
|
| 761 |
+
raise gr.Error(f"μΉμ
{i}μ μμ νμΌμ΄ μμλμκ±°λ μ½μ μ μμ΅λλ€.")
|
| 762 |
+
except Exception as e:
|
| 763 |
+
raise gr.Error(f"μΉμ
{i} μμ μ²λ¦¬ μ€ μ€λ₯: {str(e)}")
|
| 764 |
+
else:
|
| 765 |
+
raise gr.Error(f"μΉμ
{i}μ μμ νμΌμ μ°Ύμ μ μμ΅λλ€.")
|
| 766 |
else:
|
| 767 |
+
raise gr.Error(f"μΉμ
{i}μ μμμ΄ μμ΅λλ€.")
|
| 768 |
|
| 769 |
if not videos:
|
| 770 |
raise gr.Error("κ²°ν©ν μμμ΄ μμ΅λλ€.")
|
|
|
|
| 790 |
ret, frame = cap.read()
|
| 791 |
if not ret:
|
| 792 |
break
|
| 793 |
+
# νλ μ ν¬κΈ°κ° λ€λ₯Έ κ²½μ° λ¦¬μ¬μ΄μ¦
|
| 794 |
if frame.shape[:2] != (height, width):
|
| 795 |
frame = cv2.resize(frame, (width, height))
|
| 796 |
out.write(frame)
|
| 797 |
cap.release()
|
| 798 |
|
| 799 |
out.release()
|
| 800 |
+
print(f"Successfully merged {len(videos)} videos")
|
| 801 |
return output_path
|
| 802 |
|
| 803 |
except Exception as e:
|
| 804 |
raise gr.Error(f"λΉλμ€ κ²°ν© μ€ μ€λ₯ λ°μ: {e}")
|
| 805 |
|
| 806 |
+
def generate_script(topic):
|
| 807 |
+
"""μ£Όμ μ λ§λ μ€ν¬λ¦½νΈ μμ±"""
|
| 808 |
+
if not topic:
|
| 809 |
+
return "μ£Όμ λ₯Ό μ
λ ₯ν΄μ£ΌμΈμ."
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 810 |
|
| 811 |
+
messages = [
|
| 812 |
+
{"role": "system", "content": """λΉμ μ μμ μ€ν¬λ¦½νΈ μμ± μ λ¬Έκ°μ
λλ€.
|
| 813 |
+
μ£Όμ΄μ§ μ£Όμ λ‘ λ€μ ꡬ쑰μ λ§λ 5κ° μΉμ
μ μ€ν¬λ¦½νΈλ₯Ό μμ±ν΄μ£ΌμΈμ:
|
| 814 |
+
|
| 815 |
+
1. λ°°κ²½ λ° νμμ±: μ£Όμ μκ°μ μμ²μμ ν₯λ―Έ μ λ°
|
| 816 |
+
2. ν₯λ―Έ μ λ°: ꡬ체μ μΈ λ΄μ© μ κ°μ νΈκΈ°μ¬ μκ·Ή
|
| 817 |
+
3. ν΄κ²°μ±
μ μ: ν΅μ¬ λ΄μ©κ³Ό ν΄κ²°λ°©μ μ μ
|
| 818 |
+
4. λ³Έλ‘ : μμΈν μ€λͺ
κ³Ό μ₯μ λΆκ°
|
| 819 |
+
5. κ²°λ‘ : ν΅μ¬ λ©μμ§ κ°μ‘°μ νλ μ λ
|
| 820 |
+
|
| 821 |
+
κ° μΉμ
μ μμ°μ€λ½κ² μ°κ²°λμ΄μΌ νλ©°,
|
| 822 |
+
μ 체μ μΌλ‘ μΌκ΄λ ν€κ³Ό λΆμκΈ°λ₯Ό μ μ§νλ©΄μλ
|
| 823 |
+
μμ²μμ κ΄μ¬μ λκΉμ§ μ μ§ν μ μλλ‘ μμ±ν΄μ£ΌμΈμ."""},
|
| 824 |
+
{"role": "user", "content": f"λ€μ μ£Όμ λ‘ μμ μ€ν¬λ¦½νΈλ₯Ό μμ±ν΄μ£ΌμΈμ: {topic}"}
|
| 825 |
+
]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 826 |
|
| 827 |
+
try:
|
| 828 |
+
response = client.chat.completions.create(
|
| 829 |
+
model="gpt-4-1106-preview",
|
| 830 |
+
messages=messages,
|
| 831 |
+
max_tokens=2000,
|
| 832 |
+
temperature=0.7
|
| 833 |
+
)
|
| 834 |
+
return response.choices[0].message.content.strip()
|
| 835 |
+
except Exception as e:
|
| 836 |
+
print(f"Error during script generation: {e}")
|
| 837 |
+
return "μ€ν¬λ¦½νΈ μμ± μ€ μ€λ₯κ° λ°μνμ΅λλ€."
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 838 |
|
| 839 |
|
| 840 |
+
def cleanup():
|
| 841 |
+
"""λ©λͺ¨λ¦¬ μ 리 ν¨μ"""
|
| 842 |
+
torch.cuda.empty_cache()
|
| 843 |
+
gc.collect()
|
| 844 |
+
|
| 845 |
+
with gr.Blocks(theme="Yntec/HaleyCH_Theme_Orange") as iface:
|
| 846 |
+
# State λ³μλ€μ μ΄κΈ°ν
|
| 847 |
+
txt2vid_current_height = gr.State(value=320)
|
| 848 |
+
txt2vid_current_width = gr.State(value=512)
|
| 849 |
+
txt2vid_current_num_frames = gr.State(value=257)
|
| 850 |
+
|
| 851 |
+
img2vid_current_height = gr.State(value=320)
|
| 852 |
+
img2vid_current_width = gr.State(value=512)
|
| 853 |
+
img2vid_current_num_frames = gr.State(value=257)
|
| 854 |
+
|
| 855 |
+
with gr.Tabs():
|
| 856 |
+
# Text to Video Tab
|
| 857 |
+
with gr.TabItem("ν
μ€νΈλ‘ λΉλμ€ λ§λ€κΈ°"):
|
| 858 |
+
with gr.Row():
|
| 859 |
+
with gr.Column():
|
| 860 |
+
txt2vid_prompt = gr.Textbox(
|
| 861 |
+
label="Step 1: ν둬ννΈ μ
λ ₯",
|
| 862 |
+
placeholder="μμ±νκ³ μΆμ λΉλμ€λ₯Ό μ€λͺ
νμΈμ (μ΅μ 50μ)...",
|
| 863 |
+
value="κ·μ¬μ΄ κ³ μμ΄",
|
| 864 |
+
lines=5,
|
| 865 |
+
)
|
| 866 |
+
txt2vid_enhance_toggle = Toggle(
|
| 867 |
+
label="ν둬ννΈ μ¦κ°",
|
| 868 |
+
value=False,
|
| 869 |
+
interactive=True,
|
| 870 |
+
)
|
| 871 |
+
txt2vid_negative_prompt = gr.Textbox(
|
| 872 |
+
label="Step 2: λ€κ±°ν°λΈ ν둬ννΈ μ
λ ₯",
|
| 873 |
+
placeholder="λΉλμ€μμ μνμ§ μλ μμλ₯Ό μ€λͺ
νμΈμ...",
|
| 874 |
+
value="low quality, worst quality, deformed, distorted, warped, motion smear, motion artifacts, fused fingers, incorrect anatomy, strange hands, unattractive",
|
| 875 |
+
lines=2,
|
| 876 |
+
visible=False
|
| 877 |
+
)
|
| 878 |
+
txt2vid_preset = gr.Dropdown(
|
| 879 |
+
choices=[p["label"] for p in preset_options],
|
| 880 |
+
value="[16:9] 512x320, 10.3μ΄",
|
| 881 |
+
label="Step 2: ν΄μλ ν리μ
μ ν",
|
| 882 |
+
)
|
| 883 |
+
txt2vid_frame_rate = gr.Slider(
|
| 884 |
+
label="Step 3: νλ μ λ μ΄νΈ",
|
| 885 |
+
minimum=21,
|
| 886 |
+
maximum=30,
|
| 887 |
+
step=1,
|
| 888 |
+
value=25,
|
| 889 |
+
visible=False
|
| 890 |
+
)
|
| 891 |
+
txt2vid_advanced = create_advanced_options()
|
| 892 |
+
txt2vid_generate = gr.Button(
|
| 893 |
+
"Step 3: λΉλμ€ μμ±",
|
| 894 |
+
variant="primary",
|
| 895 |
+
size="lg",
|
| 896 |
+
)
|
| 897 |
+
with gr.Column():
|
| 898 |
+
txt2vid_output = gr.Video(label="μμ±λ λΉλμ€")
|
| 899 |
+
|
| 900 |
+
|
| 901 |
+
# Image to Video Tab
|
| 902 |
+
with gr.TabItem("μ΄λ―Έμ§λ‘ λΉλμ€ λ§λ€κΈ°"):
|
| 903 |
+
with gr.Row():
|
| 904 |
+
with gr.Column():
|
| 905 |
+
img2vid_image = gr.Image(
|
| 906 |
+
type="filepath",
|
| 907 |
+
label="Step 1: μ
λ ₯ μ΄λ―Έμ§ μ
λ‘λ",
|
| 908 |
+
elem_id="image_upload",
|
| 909 |
+
)
|
| 910 |
+
img2vid_prompt = gr.Textbox(
|
| 911 |
+
label="Step 2: ν둬ννΈ μ
λ ₯",
|
| 912 |
+
placeholder="μ΄λ―Έμ§λ₯Ό μ΄λ»κ² μ λλ©μ΄μ
νν μ§ μ€λͺ
νμΈμ (μ΅μ 50μ)...",
|
| 913 |
+
value="κ·μ¬μ΄ κ³ μμ΄",
|
| 914 |
+
lines=5,
|
| 915 |
+
)
|
| 916 |
+
img2vid_enhance_toggle = Toggle(
|
| 917 |
+
label="ν둬ννΈ μ¦κ°",
|
| 918 |
+
value=False,
|
| 919 |
+
interactive=True,
|
| 920 |
+
)
|
| 921 |
+
img2vid_negative_prompt = gr.Textbox(
|
| 922 |
+
label="Step 3: λ€κ±°ν°λΈ ν둬ννΈ μ
λ ₯",
|
| 923 |
+
placeholder="λΉλμ€μμ μνμ§ μλ μμλ₯Ό μ€λͺ
νμΈμ...",
|
| 924 |
+
value="low quality, worst quality, deformed, distorted, warped, motion smear, motion artifacts, fused fingers, incorrect anatomy, strange hands, unattractive",
|
| 925 |
+
lines=2,
|
| 926 |
+
visible=False
|
| 927 |
+
)
|
| 928 |
+
img2vid_preset = gr.Dropdown(
|
| 929 |
+
choices=[p["label"] for p in preset_options],
|
| 930 |
+
value="[16:9] 512x320, 10.3μ΄",
|
| 931 |
+
label="Step 3: ν΄μλ ν리μ
μ ν",
|
| 932 |
+
)
|
| 933 |
+
img2vid_frame_rate = gr.Slider(
|
| 934 |
+
label="Step 4: νλ μ λ μ΄νΈ",
|
| 935 |
+
minimum=21,
|
| 936 |
+
maximum=30,
|
| 937 |
+
step=1,
|
| 938 |
+
value=25,
|
| 939 |
+
visible=False
|
| 940 |
+
)
|
| 941 |
+
img2vid_advanced = create_advanced_options()
|
| 942 |
+
img2vid_generate = gr.Button(
|
| 943 |
+
"Step 4: λΉλμ€ μμ±",
|
| 944 |
+
variant="primary",
|
| 945 |
+
size="lg",
|
| 946 |
+
)
|
| 947 |
+
with gr.Column():
|
| 948 |
+
img2vid_output = gr.Video(label="μμ±λ λΉλμ€")
|
| 949 |
+
|
| 950 |
+
|
| 951 |
+
# Scenario Tab
|
| 952 |
+
with gr.TabItem("μλ리μ€λ‘ λΉλμ€ λ§λ€κΈ°(μνΌ)"):
|
| 953 |
+
with gr.Row():
|
| 954 |
+
with gr.Column(scale=1):
|
| 955 |
+
script_topic = gr.Textbox(
|
| 956 |
+
label="μ€ν¬λ¦½νΈ μμ±",
|
| 957 |
+
placeholder="κ²¨μΈ μΌλ³Έ μ¨μ² μ¬νμ μ£Όμ λ‘ λ°μ λλμΌλ‘ μ€ν¬λ¦½νΈ μμ±νλΌ",
|
| 958 |
+
lines=2
|
| 959 |
+
)
|
| 960 |
+
generate_script_btn = gr.Button("μ€ν¬λ¦½νΈ μμ±", variant="primary")
|
| 961 |
+
|
| 962 |
+
scenario_input = gr.Textbox(
|
| 963 |
+
label="μμ μ€ν¬λ¦½νΈ μ
λ ₯",
|
| 964 |
+
placeholder="μ 체 μλ리μ€λ₯Ό μ
λ ₯νμΈμ...",
|
| 965 |
+
lines=10
|
| 966 |
+
)
|
| 967 |
+
scenario_preset = gr.Dropdown(
|
| 968 |
+
choices=[p["label"] for p in preset_options],
|
| 969 |
+
value="[16:9] 512x320, 10.3μ΄",
|
| 970 |
+
label="νλ©΄ ν¬κΈ° μ ν"
|
| 971 |
+
)
|
| 972 |
+
analyze_btn = gr.Button("μλλ¦¬μ€ λΆμ λ° ν둬ννΈ μμ±", variant="primary")
|
| 973 |
+
|
| 974 |
+
with gr.Column(scale=2):
|
| 975 |
+
with gr.Row():
|
| 976 |
+
# μΉμ
1
|
| 977 |
+
with gr.Column():
|
| 978 |
+
section1_prompt = gr.Textbox(
|
| 979 |
+
label="1. λ°°κ²½ λ° νμμ±",
|
| 980 |
+
lines=4
|
| 981 |
+
)
|
| 982 |
+
with gr.Row():
|
| 983 |
+
section1_regenerate = gr.Button("π ν둬ννΈ μμ±")
|
| 984 |
+
section1_generate = gr.Button("π μμ μμ±")
|
| 985 |
+
section1_video = gr.Video(label="μΉμ
1 μμ")
|
| 986 |
+
|
| 987 |
+
# μΉμ
2
|
| 988 |
+
with gr.Column():
|
| 989 |
+
section2_prompt = gr.Textbox(
|
| 990 |
+
label="2. ν₯λ―Έ μ λ°",
|
| 991 |
+
lines=4
|
| 992 |
+
)
|
| 993 |
+
with gr.Row():
|
| 994 |
+
section2_regenerate = gr.Button("π ν둬ννΈ μμ±")
|
| 995 |
+
section2_generate = gr.Button("π μμ μμ±")
|
| 996 |
+
section2_video = gr.Video(label="μΉμ
2 μμ")
|
| 997 |
+
|
| 998 |
+
|
| 999 |
+
|
| 1000 |
+
with gr.Row():
|
| 1001 |
+
# μΉμ
3
|
| 1002 |
+
with gr.Column():
|
| 1003 |
+
section3_prompt = gr.Textbox(
|
| 1004 |
+
label="3. ν΄κ²°μ±
μ μ",
|
| 1005 |
+
lines=4
|
| 1006 |
+
)
|
| 1007 |
+
with gr.Row():
|
| 1008 |
+
section3_regenerate = gr.Button("π ν둬ννΈ μμ±")
|
| 1009 |
+
section3_generate = gr.Button("π μμ μμ±")
|
| 1010 |
+
section3_video = gr.Video(label="μΉμ
3 μμ")
|
| 1011 |
+
|
| 1012 |
+
# μΉμ
4
|
| 1013 |
+
with gr.Column():
|
| 1014 |
+
section4_prompt = gr.Textbox(
|
| 1015 |
+
label="4. λ³Έλ‘ ",
|
| 1016 |
+
lines=4
|
| 1017 |
+
)
|
| 1018 |
+
with gr.Row():
|
| 1019 |
+
section4_regenerate = gr.Button("π ν둬ννΈ μμ±")
|
| 1020 |
+
section4_generate = gr.Button("π μμ μμ±")
|
| 1021 |
+
section4_video = gr.Video(label="μΉμ
4 μμ")
|
| 1022 |
+
|
| 1023 |
+
with gr.Row():
|
| 1024 |
+
# μΉμ
5
|
| 1025 |
+
with gr.Column():
|
| 1026 |
+
section5_prompt = gr.Textbox(
|
| 1027 |
+
label="5. κ²°λ‘ λ° κ°μ‘°",
|
| 1028 |
+
lines=4
|
| 1029 |
+
)
|
| 1030 |
+
with gr.Row():
|
| 1031 |
+
section5_regenerate = gr.Button("π ν둬ννΈ μμ±")
|
| 1032 |
+
section5_generate = gr.Button("π μμ μμ±")
|
| 1033 |
+
section5_video = gr.Video(label="μΉμ
5 μμ")
|
| 1034 |
+
|
| 1035 |
+
# ν΅ν© μμ μΉμ
|
| 1036 |
+
with gr.Row():
|
| 1037 |
+
with gr.Column(scale=1):
|
| 1038 |
+
merge_videos_btn = gr.Button("ν΅ν© μμ μμ±", variant="primary", size="lg")
|
| 1039 |
+
|
| 1040 |
+
with gr.Column(scale=2):
|
| 1041 |
+
with gr.Row():
|
| 1042 |
+
merged_video_output = gr.Video(label="ν΅ν© μμ")
|
| 1043 |
+
|
| 1044 |
+
|
| 1045 |
+
# Text to Video Tab handlers
|
| 1046 |
+
txt2vid_preset.change(
|
| 1047 |
+
fn=preset_changed,
|
| 1048 |
+
inputs=[txt2vid_preset],
|
| 1049 |
+
outputs=[
|
| 1050 |
+
txt2vid_current_height,
|
| 1051 |
+
txt2vid_current_width,
|
| 1052 |
+
txt2vid_current_num_frames,
|
| 1053 |
+
txt2vid_advanced[3], # height_slider
|
| 1054 |
+
txt2vid_advanced[4], # width_slider
|
| 1055 |
+
txt2vid_advanced[5], # num_frames_slider
|
| 1056 |
+
]
|
| 1057 |
+
)
|
| 1058 |
+
|
| 1059 |
+
txt2vid_enhance_toggle.change(
|
| 1060 |
+
fn=update_prompt_t2v,
|
| 1061 |
+
inputs=[txt2vid_prompt, txt2vid_enhance_toggle],
|
| 1062 |
+
outputs=txt2vid_prompt
|
| 1063 |
+
)
|
| 1064 |
+
|
| 1065 |
+
txt2vid_generate.click(
|
| 1066 |
+
fn=generate_video_from_text,
|
| 1067 |
+
inputs=[
|
| 1068 |
+
txt2vid_prompt,
|
| 1069 |
+
txt2vid_enhance_toggle,
|
| 1070 |
+
txt2vid_negative_prompt,
|
| 1071 |
+
txt2vid_frame_rate,
|
| 1072 |
+
txt2vid_advanced[0], # seed
|
| 1073 |
+
txt2vid_advanced[1], # inference_steps
|
| 1074 |
+
txt2vid_advanced[2], # guidance_scale
|
| 1075 |
+
txt2vid_current_height,
|
| 1076 |
+
txt2vid_current_width,
|
| 1077 |
+
txt2vid_current_num_frames,
|
| 1078 |
+
],
|
| 1079 |
+
outputs=txt2vid_output,
|
| 1080 |
+
)
|
| 1081 |
+
|
| 1082 |
+
# Image to Video Tab handlers
|
| 1083 |
+
img2vid_preset.change(
|
| 1084 |
+
fn=preset_changed,
|
| 1085 |
+
inputs=[img2vid_preset],
|
| 1086 |
+
outputs=[
|
| 1087 |
+
img2vid_current_height,
|
| 1088 |
+
img2vid_current_width,
|
| 1089 |
+
img2vid_current_num_frames,
|
| 1090 |
+
img2vid_advanced[3], # height_slider
|
| 1091 |
+
img2vid_advanced[4], # width_slider
|
| 1092 |
+
img2vid_advanced[5], # num_frames_slider
|
| 1093 |
+
]
|
| 1094 |
+
)
|
| 1095 |
+
|
| 1096 |
+
img2vid_enhance_toggle.change(
|
| 1097 |
+
fn=update_prompt_i2v,
|
| 1098 |
+
inputs=[img2vid_prompt, img2vid_enhance_toggle],
|
| 1099 |
+
outputs=img2vid_prompt
|
| 1100 |
+
)
|
| 1101 |
+
|
| 1102 |
+
img2vid_generate.click(
|
| 1103 |
+
fn=generate_video_from_image,
|
| 1104 |
+
inputs=[
|
| 1105 |
+
img2vid_image,
|
| 1106 |
+
img2vid_prompt,
|
| 1107 |
+
img2vid_enhance_toggle,
|
| 1108 |
+
img2vid_negative_prompt,
|
| 1109 |
+
img2vid_frame_rate,
|
| 1110 |
+
img2vid_advanced[0], # seed
|
| 1111 |
+
img2vid_advanced[1], # inference_steps
|
| 1112 |
+
img2vid_advanced[2], # guidance_scale
|
| 1113 |
+
img2vid_current_height,
|
| 1114 |
+
img2vid_current_width,
|
| 1115 |
+
img2vid_current_num_frames,
|
| 1116 |
+
],
|
| 1117 |
+
outputs=img2vid_output,
|
| 1118 |
+
)
|
| 1119 |
+
|
| 1120 |
+
|
| 1121 |
+
|
| 1122 |
+
# Scenario Tab handlers
|
| 1123 |
+
generate_script_btn.click(
|
| 1124 |
+
fn=generate_script,
|
| 1125 |
+
inputs=[script_topic],
|
| 1126 |
+
outputs=[scenario_input]
|
| 1127 |
+
)
|
| 1128 |
+
|
| 1129 |
+
analyze_btn.click(
|
| 1130 |
+
fn=analyze_scenario,
|
| 1131 |
+
inputs=[scenario_input],
|
| 1132 |
+
outputs=[
|
| 1133 |
+
section1_prompt, section2_prompt, section3_prompt,
|
| 1134 |
+
section4_prompt, section5_prompt
|
| 1135 |
+
]
|
| 1136 |
+
)
|
| 1137 |
+
|
| 1138 |
+
# μΉμ
λ³ ν둬ννΈ μ¬μμ± νΈλ€λ¬
|
| 1139 |
+
section1_regenerate.click(
|
| 1140 |
+
fn=lambda x: generate_single_section_prompt(x, 1),
|
| 1141 |
+
inputs=[scenario_input],
|
| 1142 |
+
outputs=section1_prompt
|
| 1143 |
+
)
|
| 1144 |
+
|
| 1145 |
+
section2_regenerate.click(
|
| 1146 |
+
fn=lambda x: generate_single_section_prompt(x, 2),
|
| 1147 |
+
inputs=[scenario_input],
|
| 1148 |
+
outputs=section2_prompt
|
| 1149 |
+
)
|
| 1150 |
+
|
| 1151 |
+
section3_regenerate.click(
|
| 1152 |
+
fn=lambda x: generate_single_section_prompt(x, 3),
|
| 1153 |
+
inputs=[scenario_input],
|
| 1154 |
+
outputs=section3_prompt
|
| 1155 |
+
)
|
| 1156 |
+
|
| 1157 |
+
section4_regenerate.click(
|
| 1158 |
+
fn=lambda x: generate_single_section_prompt(x, 4),
|
| 1159 |
+
inputs=[scenario_input],
|
| 1160 |
+
outputs=section4_prompt
|
| 1161 |
+
)
|
| 1162 |
+
|
| 1163 |
+
section5_regenerate.click(
|
| 1164 |
+
fn=lambda x: generate_single_section_prompt(x, 5),
|
| 1165 |
+
inputs=[scenario_input],
|
| 1166 |
+
outputs=section5_prompt
|
| 1167 |
+
)
|
| 1168 |
+
|
| 1169 |
+
# μΉμ
λ³ λΉλμ€ μμ± νΈλ€λ¬
|
| 1170 |
+
section1_generate.click(
|
| 1171 |
+
fn=lambda p, pr: generate_section_video(p, pr, 1),
|
| 1172 |
+
inputs=[section1_prompt, scenario_preset],
|
| 1173 |
+
outputs=section1_video
|
| 1174 |
+
)
|
| 1175 |
+
|
| 1176 |
+
section2_generate.click(
|
| 1177 |
+
fn=lambda p, pr: generate_section_video(p, pr, 2),
|
| 1178 |
+
inputs=[section2_prompt, scenario_preset],
|
| 1179 |
+
outputs=section2_video
|
| 1180 |
+
)
|
| 1181 |
+
|
| 1182 |
+
section3_generate.click(
|
| 1183 |
+
fn=lambda p, pr: generate_section_video(p, pr, 3),
|
| 1184 |
+
inputs=[section3_prompt, scenario_preset],
|
| 1185 |
+
outputs=section3_video
|
| 1186 |
+
)
|
| 1187 |
+
|
| 1188 |
+
section4_generate.click(
|
| 1189 |
+
fn=lambda p, pr: generate_section_video(p, pr, 4),
|
| 1190 |
+
inputs=[section4_prompt, scenario_preset],
|
| 1191 |
+
outputs=section4_video
|
| 1192 |
+
)
|
| 1193 |
+
|
| 1194 |
+
section5_generate.click(
|
| 1195 |
+
fn=lambda p, pr: generate_section_video(p, pr, 5),
|
| 1196 |
+
inputs=[section5_prompt, scenario_preset],
|
| 1197 |
+
outputs=section5_video
|
| 1198 |
+
)
|
| 1199 |
+
|
| 1200 |
+
# ν΅ν© μμ μμ± νΈλ€λ¬
|
| 1201 |
+
merge_videos_btn.click(
|
| 1202 |
+
fn=merge_section_videos,
|
| 1203 |
+
inputs=[
|
| 1204 |
+
section1_video,
|
| 1205 |
+
section2_video,
|
| 1206 |
+
section3_video,
|
| 1207 |
+
section4_video,
|
| 1208 |
+
section5_video
|
| 1209 |
+
],
|
| 1210 |
+
outputs=merged_video_output
|
| 1211 |
+
)
|
| 1212 |
|
|
|
|
| 1213 |
if __name__ == "__main__":
|
| 1214 |
+
iface.queue(max_size=64, default_concurrency_limit=1, api_open=False).launch(
|
| 1215 |
+
share=True,
|
| 1216 |
+
show_api=False
|
| 1217 |
+
)
|