Update app.py
Browse files
app.py
CHANGED
|
@@ -11,17 +11,14 @@ except ImportError:
|
|
| 11 |
InferenceClient = None
|
| 12 |
|
| 13 |
|
| 14 |
-
# =============================================================================
|
| 15 |
-
# CONFIGURATION
|
| 16 |
-
# =============================================================================
|
| 17 |
-
|
| 18 |
APP_TITLE = "Free Image Prompt Generator"
|
| 19 |
|
| 20 |
-
|
| 21 |
"PROMPT_MODEL",
|
| 22 |
"Qwen/Qwen2.5-7B-Instruct",
|
| 23 |
)
|
| 24 |
|
|
|
|
| 25 |
STYLES = [
|
| 26 |
"Cinematic",
|
| 27 |
"Photorealistic",
|
|
@@ -78,7 +75,8 @@ CAMERAS = [
|
|
| 78 |
"Dynamic action perspective",
|
| 79 |
]
|
| 80 |
|
| 81 |
-
|
|
|
|
| 82 |
"text, captions, subtitles, logos, watermarks, signatures, "
|
| 83 |
"malformed anatomy, extra limbs, duplicated objects, distorted faces, "
|
| 84 |
"bad hands, inconsistent perspective, floating objects, visual artifacts, "
|
|
@@ -86,33 +84,25 @@ DEFAULT_NEGATIVE = (
|
|
| 86 |
)
|
| 87 |
|
| 88 |
|
| 89 |
-
# =============================================================================
|
| 90 |
-
# TEXT HELPERS
|
| 91 |
-
# =============================================================================
|
| 92 |
-
|
| 93 |
def clean_text(text: str) -> str:
|
| 94 |
-
"""Normalize whitespace while preserving paragraph breaks."""
|
| 95 |
if not text:
|
| 96 |
return ""
|
| 97 |
|
| 98 |
text = str(text)
|
| 99 |
text = text.replace("\r\n", "\n")
|
| 100 |
text = text.replace("\r", "\n")
|
| 101 |
-
|
| 102 |
text = re.sub(r"[ \t]+", " ", text)
|
| 103 |
text = re.sub(r"\n{3,}", "\n\n", text)
|
| 104 |
|
| 105 |
return text.strip()
|
| 106 |
|
| 107 |
|
| 108 |
-
def
|
| 109 |
-
"""Clean common formatting accidentally returned by language models."""
|
| 110 |
if not text:
|
| 111 |
return ""
|
| 112 |
|
| 113 |
text = str(text).strip()
|
| 114 |
|
| 115 |
-
# Remove Markdown code fences.
|
| 116 |
text = re.sub(
|
| 117 |
r"^```(?:text|prompt|markdown)?\s*",
|
| 118 |
"",
|
|
@@ -122,25 +112,17 @@ def extract_prompt_text(text: str) -> str:
|
|
| 122 |
|
| 123 |
text = re.sub(r"\s*```$", "", text)
|
| 124 |
|
| 125 |
-
# Remove accidental surrounding quotation marks.
|
| 126 |
if len(text) >= 2:
|
| 127 |
-
if
|
| 128 |
-
|
| 129 |
-
or
|
| 130 |
-
(text.startswith("'") and text.endswith("'"))
|
| 131 |
-
):
|
| 132 |
-
text = text[1:-1].strip()
|
| 133 |
|
| 134 |
-
|
|
|
|
| 135 |
|
|
|
|
| 136 |
|
| 137 |
-
# =============================================================================
|
| 138 |
-
# HUGGING FACE CLIENT
|
| 139 |
-
# =============================================================================
|
| 140 |
|
| 141 |
def get_client() -> Optional["InferenceClient"]:
|
| 142 |
-
"""Return a configured Hugging Face client if available."""
|
| 143 |
-
|
| 144 |
token = os.getenv("HF_TOKEN")
|
| 145 |
|
| 146 |
if not token:
|
|
@@ -151,20 +133,18 @@ def get_client() -> Optional["InferenceClient"]:
|
|
| 151 |
|
| 152 |
try:
|
| 153 |
return InferenceClient(
|
| 154 |
-
model=
|
| 155 |
token=token,
|
| 156 |
)
|
| 157 |
except Exception as exc:
|
| 158 |
print(
|
| 159 |
-
|
|
|
|
|
|
|
| 160 |
)
|
| 161 |
return None
|
| 162 |
|
| 163 |
|
| 164 |
-
# =============================================================================
|
| 165 |
-
# LOCAL PROMPT ENGINE
|
| 166 |
-
# =============================================================================
|
| 167 |
-
|
| 168 |
def build_local_prompt(
|
| 169 |
idea: str,
|
| 170 |
style: str,
|
|
@@ -172,7 +152,7 @@ def build_local_prompt(
|
|
| 172 |
mood: str,
|
| 173 |
detail: str,
|
| 174 |
camera: str,
|
| 175 |
-
) ->
|
| 176 |
|
| 177 |
idea = clean_text(idea)
|
| 178 |
|
|
@@ -184,77 +164,54 @@ def build_local_prompt(
|
|
| 184 |
|
| 185 |
prompt = (
|
| 186 |
f"{style} image of {idea}. "
|
| 187 |
-
f"Use a {camera.lower()} with intentional framing,
|
| 188 |
-
f"hierarchy, convincing depth, natural perspective,
|
| 189 |
-
f"negative space. "
|
| 190 |
f"Create a coherent environment with meaningful foreground, "
|
| 191 |
-
f"midground, and background elements
|
| 192 |
-
f"Use {mood.lower()} lighting and atmosphere with believable
|
| 193 |
-
f"falloff, dimensional shadows, subtle highlights,
|
| 194 |
-
f"depth. "
|
| 195 |
f"Apply a consistent {style.lower()} aesthetic with "
|
| 196 |
f"{detail.lower()}, refined textures, coherent materials, "
|
| 197 |
f"realistic surface response, and controlled visual complexity. "
|
| 198 |
-
f"Use a deliberate color palette appropriate to the
|
| 199 |
-
f"
|
| 200 |
-
f"
|
| 201 |
-
f"
|
| 202 |
-
f"finish, and cinematic visual storytelling. "
|
| 203 |
f"Aspect ratio {aspect}."
|
| 204 |
)
|
| 205 |
|
| 206 |
-
return clean_text(prompt)
|
| 207 |
-
|
| 208 |
|
| 209 |
-
# =============================================================================
|
| 210 |
-
# AI PROMPT ENGINE
|
| 211 |
-
# =============================================================================
|
| 212 |
|
| 213 |
-
|
| 214 |
You are an expert image-prompt architect.
|
| 215 |
|
| 216 |
-
Transform
|
| 217 |
-
|
| 218 |
-
|
| 219 |
-
|
| 220 |
-
|
| 221 |
-
|
| 222 |
-
|
| 223 |
-
|
| 224 |
-
|
| 225 |
-
|
| 226 |
-
|
| 227 |
-
|
| 228 |
-
|
| 229 |
-
|
| 230 |
-
|
| 231 |
-
|
| 232 |
-
|
| 233 |
-
|
| 234 |
-
|
| 235 |
-
|
| 236 |
-
|
| 237 |
-
|
| 238 |
-
|
| 239 |
-
|
| 240 |
-
17. Do not wrap the prompt in quotation marks.
|
| 241 |
-
|
| 242 |
-
A strong prompt should naturally communicate:
|
| 243 |
-
|
| 244 |
-
- subject
|
| 245 |
-
- environment
|
| 246 |
-
- composition
|
| 247 |
-
- camera language where useful
|
| 248 |
-
- lighting
|
| 249 |
-
- atmosphere
|
| 250 |
-
- materials and textures where useful
|
| 251 |
-
- visual style
|
| 252 |
-
- color
|
| 253 |
-
- depth
|
| 254 |
-
- image quality
|
| 255 |
-
- aspect ratio
|
| 256 |
-
|
| 257 |
-
Use natural prose rather than unnecessary section headings.
|
| 258 |
"""
|
| 259 |
|
| 260 |
|
|
@@ -272,10 +229,10 @@ def build_ai_prompt(
|
|
| 272 |
if client is None:
|
| 273 |
return (
|
| 274 |
"",
|
| 275 |
-
"
|
| 276 |
)
|
| 277 |
|
| 278 |
-
|
| 279 |
USER IDEA:
|
| 280 |
{idea}
|
| 281 |
|
|
@@ -285,7 +242,7 @@ STYLE:
|
|
| 285 |
ASPECT RATIO:
|
| 286 |
{aspect}
|
| 287 |
|
| 288 |
-
MOOD
|
| 289 |
{mood}
|
| 290 |
|
| 291 |
DETAIL LEVEL:
|
|
@@ -300,11 +257,11 @@ CAMERA / COMPOSITION:
|
|
| 300 |
messages=[
|
| 301 |
{
|
| 302 |
"role": "system",
|
| 303 |
-
"content":
|
| 304 |
},
|
| 305 |
{
|
| 306 |
"role": "user",
|
| 307 |
-
"content":
|
| 308 |
},
|
| 309 |
],
|
| 310 |
max_tokens=1200,
|
|
@@ -312,43 +269,42 @@ CAMERA / COMPOSITION:
|
|
| 312 |
)
|
| 313 |
|
| 314 |
if not response:
|
| 315 |
-
return "", "
|
| 316 |
|
| 317 |
if not response.choices:
|
| 318 |
-
return "", "
|
| 319 |
|
| 320 |
message = response.choices[0].message
|
| 321 |
|
| 322 |
if not message:
|
| 323 |
-
return "", "
|
| 324 |
|
| 325 |
content = message.content
|
| 326 |
|
| 327 |
if not content:
|
| 328 |
-
return "", "
|
| 329 |
|
| 330 |
-
content =
|
| 331 |
|
| 332 |
if not content:
|
| 333 |
-
return "", "
|
| 334 |
|
| 335 |
return content, None
|
| 336 |
|
| 337 |
except Exception as exc:
|
|
|
|
| 338 |
print(
|
| 339 |
-
|
|
|
|
|
|
|
| 340 |
)
|
| 341 |
|
| 342 |
return (
|
| 343 |
"",
|
| 344 |
-
"
|
| 345 |
)
|
| 346 |
|
| 347 |
|
| 348 |
-
# =============================================================================
|
| 349 |
-
# MAIN GENERATION
|
| 350 |
-
# =============================================================================
|
| 351 |
-
|
| 352 |
def generate_prompt(
|
| 353 |
idea: str,
|
| 354 |
style: str,
|
|
@@ -357,187 +313,113 @@ def generate_prompt(
|
|
| 357 |
detail: str,
|
| 358 |
camera: str,
|
| 359 |
use_ai: bool,
|
| 360 |
-
)
|
| 361 |
|
| 362 |
if not idea or not idea.strip():
|
|
|
|
| 363 |
return (
|
| 364 |
"",
|
| 365 |
-
|
| 366 |
"⚠️ Enter an image idea first.",
|
| 367 |
)
|
| 368 |
|
| 369 |
idea = clean_text(idea)
|
| 370 |
|
| 371 |
if use_ai:
|
|
|
|
| 372 |
ai_prompt, error = build_ai_prompt(
|
| 373 |
-
idea
|
| 374 |
-
style
|
| 375 |
-
aspect
|
| 376 |
-
mood
|
| 377 |
-
detail
|
| 378 |
-
camera
|
| 379 |
)
|
| 380 |
|
| 381 |
if ai_prompt:
|
|
|
|
| 382 |
return (
|
| 383 |
ai_prompt,
|
| 384 |
-
|
| 385 |
-
"✓
|
| 386 |
)
|
| 387 |
|
| 388 |
-
local_prompt
|
| 389 |
-
idea
|
| 390 |
-
style
|
| 391 |
-
aspect
|
| 392 |
-
mood
|
| 393 |
-
detail
|
| 394 |
-
camera
|
| 395 |
)
|
| 396 |
|
| 397 |
return (
|
| 398 |
local_prompt,
|
| 399 |
-
|
| 400 |
-
f"✓
|
| 401 |
)
|
| 402 |
|
| 403 |
-
local_prompt
|
| 404 |
-
idea
|
| 405 |
-
style
|
| 406 |
-
aspect
|
| 407 |
-
mood
|
| 408 |
-
detail
|
| 409 |
-
camera
|
| 410 |
)
|
| 411 |
|
| 412 |
return (
|
| 413 |
local_prompt,
|
| 414 |
-
|
| 415 |
-
"✓
|
| 416 |
)
|
| 417 |
|
| 418 |
|
| 419 |
-
|
| 420 |
-
# PROMPT TRANSFORMATION
|
| 421 |
-
# =============================================================================
|
| 422 |
|
| 423 |
-
TRANSFORMATION_INSTRUCTIONS = {
|
| 424 |
"Improve": """
|
| 425 |
-
Improve this image prompt.
|
| 426 |
|
| 427 |
Preserve the exact core subject and concept.
|
| 428 |
|
| 429 |
-
Increase useful visual specificity
|
| 430 |
-
Strengthen composition, lighting, depth,
|
| 431 |
-
|
| 432 |
|
| 433 |
-
Remove
|
| 434 |
|
| 435 |
Do not radically change the concept.
|
| 436 |
|
| 437 |
-
Return ONLY the improved
|
| 438 |
""",
|
| 439 |
|
| 440 |
"Remix": """
|
| 441 |
-
Remix this image prompt.
|
| 442 |
|
| 443 |
Preserve the core subject and concept.
|
| 444 |
|
| 445 |
-
Create a fresh
|
| 446 |
-
|
| 447 |
-
or visual treatment.
|
| 448 |
|
| 449 |
Do not replace the main subject.
|
| 450 |
|
| 451 |
-
|
| 452 |
-
|
| 453 |
-
Return ONLY the remixed image prompt.
|
| 454 |
""",
|
| 455 |
|
| 456 |
"Simplify": """
|
| 457 |
-
Simplify this image prompt.
|
| 458 |
|
| 459 |
-
Preserve the important subject, environment, style, mood
|
| 460 |
|
| 461 |
-
Remove unnecessary adjectives, repetition
|
| 462 |
-
over-engineered descriptions.
|
| 463 |
|
| 464 |
-
Make the prompt concise
|
| 465 |
|
| 466 |
-
Return ONLY the simplified
|
| 467 |
""",
|
| 468 |
}
|
| 469 |
|
| 470 |
|
| 471 |
-
def transform_prompt(
|
| 472 |
-
prompt: str,
|
| 473 |
-
action: str,
|
| 474 |
-
) -> Tuple[str, str]:
|
| 475 |
-
|
| 476 |
-
if not prompt or not prompt.strip():
|
| 477 |
-
return (
|
| 478 |
-
"",
|
| 479 |
-
"⚠️ Generate a prompt first.",
|
| 480 |
-
)
|
| 481 |
-
|
| 482 |
-
prompt = clean_text(prompt)
|
| 483 |
-
|
| 484 |
-
instruction = TRANSFORMATION_INSTRUCTIONS.get(action)
|
| 485 |
-
|
| 486 |
-
if not instruction:
|
| 487 |
-
return (
|
| 488 |
-
prompt,
|
| 489 |
-
"✓ Prompt ready.",
|
| 490 |
-
)
|
| 491 |
-
|
| 492 |
-
client = get_client()
|
| 493 |
-
|
| 494 |
-
if client is not None:
|
| 495 |
-
try:
|
| 496 |
-
response = client.chat_completion(
|
| 497 |
-
messages=[
|
| 498 |
-
{
|
| 499 |
-
"role": "system",
|
| 500 |
-
"content": (
|
| 501 |
-
"You are an expert image prompt editor. "
|
| 502 |
-
"Return only the requested image prompt."
|
| 503 |
-
),
|
| 504 |
-
},
|
| 505 |
-
{
|
| 506 |
-
"role": "user",
|
| 507 |
-
"content": (
|
| 508 |
-
f"{instruction}\n\n"
|
| 509 |
-
f"CURRENT PROMPT:\n{prompt}"
|
| 510 |
-
),
|
| 511 |
-
},
|
| 512 |
-
],
|
| 513 |
-
max_tokens=1200,
|
| 514 |
-
temperature=(
|
| 515 |
-
0.8
|
| 516 |
-
if action == "Remix"
|
| 517 |
-
else 0.5
|
| 518 |
-
),
|
| 519 |
-
)
|
| 520 |
-
|
| 521 |
-
if response and response.choices:
|
| 522 |
-
content = response.choices[0].message.content
|
| 523 |
-
|
| 524 |
-
if content:
|
| 525 |
-
content = extract_prompt_text(content)
|
| 526 |
-
|
| 527 |
-
if content:
|
| 528 |
-
return (
|
| 529 |
-
content,
|
| 530 |
-
f"✓ Prompt {action.lower()}ed with Smart AI",
|
| 531 |
-
)
|
| 532 |
-
|
| 533 |
-
except Exception as exc:
|
| 534 |
-
print(
|
| 535 |
-
f"[HF TRANSFORM ERROR] {type(exc).__name__}: {exc}"
|
| 536 |
-
)
|
| 537 |
-
|
| 538 |
-
return local_transform(prompt, action)
|
| 539 |
-
|
| 540 |
-
|
| 541 |
def local_transform(
|
| 542 |
prompt: str,
|
| 543 |
action: str,
|
|
@@ -547,17 +429,15 @@ def local_transform(
|
|
| 547 |
|
| 548 |
if action == "Improve":
|
| 549 |
|
| 550 |
-
prompt = prompt.rstrip(". ")
|
| 551 |
-
|
| 552 |
addition = (
|
| 553 |
-
" Enhance
|
| 554 |
"natural lighting, refined textures, believable materials, "
|
| 555 |
"and strong focal detail."
|
| 556 |
)
|
| 557 |
|
| 558 |
return (
|
| 559 |
-
|
| 560 |
-
"✓
|
| 561 |
)
|
| 562 |
|
| 563 |
if action == "Simplify":
|
|
@@ -572,225 +452,123 @@ def local_transform(
|
|
| 572 |
|
| 573 |
return (
|
| 574 |
clean_text(prompt),
|
| 575 |
-
"✓
|
| 576 |
)
|
| 577 |
|
| 578 |
if action == "Remix":
|
| 579 |
|
| 580 |
-
|
| 581 |
"with dramatic atmospheric depth",
|
| 582 |
"under cinematic environmental lighting",
|
| 583 |
"with a striking editorial composition",
|
| 584 |
"with subtle surreal visual elements",
|
| 585 |
"during a visually dramatic transitional moment",
|
| 586 |
-
"with strong foreground-to-background depth",
|
| 587 |
"using an unconventional cinematic perspective",
|
| 588 |
]
|
| 589 |
|
| 590 |
-
addition = random.choice(
|
| 591 |
-
|
| 592 |
-
prompt = prompt.rstrip(". ")
|
| 593 |
|
| 594 |
return (
|
| 595 |
-
|
| 596 |
-
"✓
|
| 597 |
)
|
| 598 |
|
| 599 |
return (
|
| 600 |
prompt,
|
| 601 |
-
"✓ Prompt ready
|
| 602 |
-
)
|
| 603 |
-
|
| 604 |
-
|
| 605 |
-
def improve_prompt(prompt: str):
|
| 606 |
-
return transform_prompt(prompt, "Improve")
|
| 607 |
-
|
| 608 |
-
|
| 609 |
-
def remix_prompt(prompt: str):
|
| 610 |
-
return transform_prompt(prompt, "Remix")
|
| 611 |
-
|
| 612 |
-
|
| 613 |
-
def simplify_prompt(prompt: str):
|
| 614 |
-
return transform_prompt(prompt, "Simplify")
|
| 615 |
-
|
| 616 |
-
|
| 617 |
-
# =============================================================================
|
| 618 |
-
# SURPRISE ME
|
| 619 |
-
# =============================================================================
|
| 620 |
-
|
| 621 |
-
SURPRISE_IDEAS = [
|
| 622 |
-
"a tiny astronaut exploring an enormous abandoned greenhouse",
|
| 623 |
-
"a fox sitting inside an ancient observatory beneath a meteor shower",
|
| 624 |
-
"a futuristic Brisbane street transformed by bioluminescent plants",
|
| 625 |
-
"a mechanical whale swimming through a flooded cathedral",
|
| 626 |
-
"a lone explorer discovering a hidden city beneath the ocean",
|
| 627 |
-
"a tiny robot repairing a gigantic antique clock",
|
| 628 |
-
"a medieval library built inside the hollow trunk of an enormous tree",
|
| 629 |
-
"an abandoned space station slowly being reclaimed by vegetation",
|
| 630 |
-
"a cyberpunk night market during a tropical storm",
|
| 631 |
-
"a mysterious train crossing an impossible landscape at sunrise",
|
| 632 |
-
"an ancient temple hidden beneath a frozen ocean",
|
| 633 |
-
"a giant botanical garden growing inside an abandoned spacecraft",
|
| 634 |
-
]
|
| 635 |
-
|
| 636 |
-
|
| 637 |
-
def surprise_me():
|
| 638 |
-
idea = random.choice(SURPRISE_IDEAS)
|
| 639 |
-
style = random.choice(STYLES)
|
| 640 |
-
aspect = random.choice(ASPECT_RATIOS)
|
| 641 |
-
mood = random.choice(MOODS)
|
| 642 |
-
detail = random.choice(DETAIL_LEVELS)
|
| 643 |
-
camera = random.choice(CAMERAS)
|
| 644 |
-
|
| 645 |
-
return (
|
| 646 |
-
idea,
|
| 647 |
-
style,
|
| 648 |
-
aspect,
|
| 649 |
-
mood,
|
| 650 |
-
detail,
|
| 651 |
-
camera,
|
| 652 |
-
"🎲 Surprise idea created — hit Generate Prompt.",
|
| 653 |
)
|
| 654 |
|
| 655 |
|
| 656 |
-
|
| 657 |
-
|
| 658 |
-
|
| 659 |
-
|
| 660 |
-
EXAMPLES = [
|
| 661 |
-
[
|
| 662 |
-
"A futuristic Brisbane skyline after a tropical storm",
|
| 663 |
-
"Cinematic",
|
| 664 |
-
"16:9",
|
| 665 |
-
"Moody",
|
| 666 |
-
"Highly detailed",
|
| 667 |
-
"Wide establishing shot",
|
| 668 |
-
],
|
| 669 |
-
[
|
| 670 |
-
"A tiny robot repairing an antique mechanical clock",
|
| 671 |
-
"3D Render",
|
| 672 |
-
"1:1",
|
| 673 |
-
"Whimsical",
|
| 674 |
-
"Extremely detailed",
|
| 675 |
-
"Macro composition",
|
| 676 |
-
],
|
| 677 |
-
[
|
| 678 |
-
"A medieval library hidden inside a giant tree",
|
| 679 |
-
"Fantasy",
|
| 680 |
-
"4:5",
|
| 681 |
-
"Serene",
|
| 682 |
-
"Highly detailed",
|
| 683 |
-
"Symmetrical composition",
|
| 684 |
-
],
|
| 685 |
-
[
|
| 686 |
-
"A lone astronaut exploring an abandoned underwater city",
|
| 687 |
-
"Concept Art",
|
| 688 |
-
"21:9",
|
| 689 |
-
"Mysterious",
|
| 690 |
-
"Extremely detailed",
|
| 691 |
-
"Wide establishing shot",
|
| 692 |
-
],
|
| 693 |
-
]
|
| 694 |
-
|
| 695 |
-
|
| 696 |
-
# =============================================================================
|
| 697 |
-
# CUSTOM CSS
|
| 698 |
-
# =============================================================================
|
| 699 |
-
|
| 700 |
-
CUSTOM_CSS = """
|
| 701 |
-
.gradio-container {
|
| 702 |
-
max-width: 1100px !important;
|
| 703 |
-
margin: auto !important;
|
| 704 |
-
}
|
| 705 |
|
| 706 |
-
|
| 707 |
-
text-align: center;
|
| 708 |
-
padding: 18px 10px 8px 10px;
|
| 709 |
-
}
|
| 710 |
|
| 711 |
-
|
| 712 |
-
|
| 713 |
-
|
| 714 |
-
|
| 715 |
|
| 716 |
-
|
| 717 |
-
font-size: 1.05rem;
|
| 718 |
-
opacity: 0.8;
|
| 719 |
-
}
|
| 720 |
|
| 721 |
-
|
| 722 |
-
|
| 723 |
-
}
|
| 724 |
|
| 725 |
-
|
| 726 |
-
text-align: center;
|
| 727 |
-
font-size: 0.9rem;
|
| 728 |
-
}
|
| 729 |
|
| 730 |
-
|
| 731 |
-
min-height: 52px;
|
| 732 |
-
font-size: 1.05rem !important;
|
| 733 |
-
font-weight: 600 !important;
|
| 734 |
-
}
|
| 735 |
|
| 736 |
-
|
| 737 |
-
|
| 738 |
-
|
| 739 |
-
"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 740 |
|
|
|
|
| 741 |
|
| 742 |
-
|
| 743 |
-
# GRADIO APPLICATION
|
| 744 |
-
# =============================================================================
|
| 745 |
|
| 746 |
-
|
| 747 |
-
title=APP_TITLE,
|
| 748 |
-
css=CUSTOM_CSS,
|
| 749 |
-
) as demo:
|
| 750 |
|
| 751 |
-
|
| 752 |
-
# HERO
|
| 753 |
-
# -------------------------------------------------------------------------
|
| 754 |
|
| 755 |
-
|
| 756 |
-
"""
|
| 757 |
-
<div class="hero">
|
| 758 |
|
| 759 |
-
|
|
|
|
|
|
|
|
|
|
| 760 |
|
| 761 |
-
|
| 762 |
|
| 763 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 764 |
|
| 765 |
-
|
| 766 |
-
|
|
|
|
| 767 |
)
|
| 768 |
|
| 769 |
-
# -------------------------------------------------------------------------
|
| 770 |
-
# MAIN WORKSPACE
|
| 771 |
-
# -------------------------------------------------------------------------
|
| 772 |
|
| 773 |
-
|
|
|
|
| 774 |
|
| 775 |
-
# =====================================================================
|
| 776 |
-
# INPUT PANEL
|
| 777 |
-
# =====================================================================
|
| 778 |
|
| 779 |
-
|
|
|
|
| 780 |
|
| 781 |
-
idea = gr.Textbox(
|
| 782 |
-
label="What do you want to create?",
|
| 783 |
-
placeholder=(
|
| 784 |
-
"Example: A lone astronaut exploring an abandoned "
|
| 785 |
-
"underwater city"
|
| 786 |
-
),
|
| 787 |
-
lines=5,
|
| 788 |
-
max_lines=10,
|
| 789 |
-
)
|
| 790 |
|
| 791 |
-
|
| 792 |
-
|
| 793 |
-
variant="secondary",
|
| 794 |
-
)
|
| 795 |
|
| 796 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
InferenceClient = None
|
| 12 |
|
| 13 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 14 |
APP_TITLE = "Free Image Prompt Generator"
|
| 15 |
|
| 16 |
+
MODEL = os.getenv(
|
| 17 |
"PROMPT_MODEL",
|
| 18 |
"Qwen/Qwen2.5-7B-Instruct",
|
| 19 |
)
|
| 20 |
|
| 21 |
+
|
| 22 |
STYLES = [
|
| 23 |
"Cinematic",
|
| 24 |
"Photorealistic",
|
|
|
|
| 75 |
"Dynamic action perspective",
|
| 76 |
]
|
| 77 |
|
| 78 |
+
|
| 79 |
+
NEGATIVE_PROMPT = (
|
| 80 |
"text, captions, subtitles, logos, watermarks, signatures, "
|
| 81 |
"malformed anatomy, extra limbs, duplicated objects, distorted faces, "
|
| 82 |
"bad hands, inconsistent perspective, floating objects, visual artifacts, "
|
|
|
|
| 84 |
)
|
| 85 |
|
| 86 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 87 |
def clean_text(text: str) -> str:
|
|
|
|
| 88 |
if not text:
|
| 89 |
return ""
|
| 90 |
|
| 91 |
text = str(text)
|
| 92 |
text = text.replace("\r\n", "\n")
|
| 93 |
text = text.replace("\r", "\n")
|
|
|
|
| 94 |
text = re.sub(r"[ \t]+", " ", text)
|
| 95 |
text = re.sub(r"\n{3,}", "\n\n", text)
|
| 96 |
|
| 97 |
return text.strip()
|
| 98 |
|
| 99 |
|
| 100 |
+
def clean_ai_response(text: str) -> str:
|
|
|
|
| 101 |
if not text:
|
| 102 |
return ""
|
| 103 |
|
| 104 |
text = str(text).strip()
|
| 105 |
|
|
|
|
| 106 |
text = re.sub(
|
| 107 |
r"^```(?:text|prompt|markdown)?\s*",
|
| 108 |
"",
|
|
|
|
| 112 |
|
| 113 |
text = re.sub(r"\s*```$", "", text)
|
| 114 |
|
|
|
|
| 115 |
if len(text) >= 2:
|
| 116 |
+
if text[0] == '"' and text[-1] == '"':
|
| 117 |
+
text = text[1:-1]
|
|
|
|
|
|
|
|
|
|
|
|
|
| 118 |
|
| 119 |
+
elif text[0] == "'" and text[-1] == "'":
|
| 120 |
+
text = text[1:-1]
|
| 121 |
|
| 122 |
+
return clean_text(text)
|
| 123 |
|
|
|
|
|
|
|
|
|
|
| 124 |
|
| 125 |
def get_client() -> Optional["InferenceClient"]:
|
|
|
|
|
|
|
| 126 |
token = os.getenv("HF_TOKEN")
|
| 127 |
|
| 128 |
if not token:
|
|
|
|
| 133 |
|
| 134 |
try:
|
| 135 |
return InferenceClient(
|
| 136 |
+
model=MODEL,
|
| 137 |
token=token,
|
| 138 |
)
|
| 139 |
except Exception as exc:
|
| 140 |
print(
|
| 141 |
+
"[CLIENT ERROR]",
|
| 142 |
+
type(exc).__name__,
|
| 143 |
+
str(exc),
|
| 144 |
)
|
| 145 |
return None
|
| 146 |
|
| 147 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 148 |
def build_local_prompt(
|
| 149 |
idea: str,
|
| 150 |
style: str,
|
|
|
|
| 152 |
mood: str,
|
| 153 |
detail: str,
|
| 154 |
camera: str,
|
| 155 |
+
) -> str:
|
| 156 |
|
| 157 |
idea = clean_text(idea)
|
| 158 |
|
|
|
|
| 164 |
|
| 165 |
prompt = (
|
| 166 |
f"{style} image of {idea}. "
|
| 167 |
+
f"Use a {camera.lower()} with intentional framing, "
|
| 168 |
+
f"strong visual hierarchy, convincing depth, natural perspective, "
|
| 169 |
+
f"and balanced negative space. "
|
| 170 |
f"Create a coherent environment with meaningful foreground, "
|
| 171 |
+
f"midground, and background elements supporting the subject. "
|
| 172 |
+
f"Use {mood.lower()} lighting and atmosphere with believable "
|
| 173 |
+
f"light falloff, dimensional shadows, subtle highlights, "
|
| 174 |
+
f"and atmospheric depth. "
|
| 175 |
f"Apply a consistent {style.lower()} aesthetic with "
|
| 176 |
f"{detail.lower()}, refined textures, coherent materials, "
|
| 177 |
f"realistic surface response, and controlled visual complexity. "
|
| 178 |
+
f"Use a deliberate color palette appropriate to the scene. "
|
| 179 |
+
f"High visual fidelity, crisp focal detail, clean edges, "
|
| 180 |
+
f"nuanced textures, realistic spatial relationships, "
|
| 181 |
+
f"polished professional finish, cinematic visual storytelling. "
|
|
|
|
| 182 |
f"Aspect ratio {aspect}."
|
| 183 |
)
|
| 184 |
|
| 185 |
+
return clean_text(prompt)
|
|
|
|
| 186 |
|
|
|
|
|
|
|
|
|
|
| 187 |
|
| 188 |
+
AI_SYSTEM = """
|
| 189 |
You are an expert image-prompt architect.
|
| 190 |
|
| 191 |
+
Transform the user's idea into a production-ready image-generation prompt.
|
| 192 |
+
|
| 193 |
+
Preserve the user's core concept, subject and intent.
|
| 194 |
+
|
| 195 |
+
Add useful visual specificity involving:
|
| 196 |
+
subject, environment, composition, camera, lighting, atmosphere, materials,
|
| 197 |
+
textures, color, depth and visual style.
|
| 198 |
+
|
| 199 |
+
Resolve reasonable ambiguity intelligently.
|
| 200 |
+
|
| 201 |
+
Do not invent unrelated story elements, people, brands or locations.
|
| 202 |
+
|
| 203 |
+
Avoid generic filler such as:
|
| 204 |
+
masterpiece, stunning, breathtaking, amazing, incredible.
|
| 205 |
+
|
| 206 |
+
Avoid repetition and contradictory attributes.
|
| 207 |
+
|
| 208 |
+
Do not radically change the user's concept.
|
| 209 |
+
|
| 210 |
+
Return ONLY the finished image prompt.
|
| 211 |
+
|
| 212 |
+
Do not explain your choices.
|
| 213 |
+
Do not mention being an AI.
|
| 214 |
+
Do not use quotation marks around the answer.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 215 |
"""
|
| 216 |
|
| 217 |
|
|
|
|
| 229 |
if client is None:
|
| 230 |
return (
|
| 231 |
"",
|
| 232 |
+
"Hugging Face AI unavailable.",
|
| 233 |
)
|
| 234 |
|
| 235 |
+
request = f"""
|
| 236 |
USER IDEA:
|
| 237 |
{idea}
|
| 238 |
|
|
|
|
| 242 |
ASPECT RATIO:
|
| 243 |
{aspect}
|
| 244 |
|
| 245 |
+
MOOD:
|
| 246 |
{mood}
|
| 247 |
|
| 248 |
DETAIL LEVEL:
|
|
|
|
| 257 |
messages=[
|
| 258 |
{
|
| 259 |
"role": "system",
|
| 260 |
+
"content": AI_SYSTEM,
|
| 261 |
},
|
| 262 |
{
|
| 263 |
"role": "user",
|
| 264 |
+
"content": request,
|
| 265 |
},
|
| 266 |
],
|
| 267 |
max_tokens=1200,
|
|
|
|
| 269 |
)
|
| 270 |
|
| 271 |
if not response:
|
| 272 |
+
return "", "AI returned no response."
|
| 273 |
|
| 274 |
if not response.choices:
|
| 275 |
+
return "", "AI returned no choices."
|
| 276 |
|
| 277 |
message = response.choices[0].message
|
| 278 |
|
| 279 |
if not message:
|
| 280 |
+
return "", "AI returned no message."
|
| 281 |
|
| 282 |
content = message.content
|
| 283 |
|
| 284 |
if not content:
|
| 285 |
+
return "", "AI returned empty content."
|
| 286 |
|
| 287 |
+
content = clean_ai_response(content)
|
| 288 |
|
| 289 |
if not content:
|
| 290 |
+
return "", "AI returned an empty prompt."
|
| 291 |
|
| 292 |
return content, None
|
| 293 |
|
| 294 |
except Exception as exc:
|
| 295 |
+
|
| 296 |
print(
|
| 297 |
+
"[HF ERROR]",
|
| 298 |
+
type(exc).__name__,
|
| 299 |
+
str(exc),
|
| 300 |
)
|
| 301 |
|
| 302 |
return (
|
| 303 |
"",
|
| 304 |
+
"Hugging Face request failed.",
|
| 305 |
)
|
| 306 |
|
| 307 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 308 |
def generate_prompt(
|
| 309 |
idea: str,
|
| 310 |
style: str,
|
|
|
|
| 313 |
detail: str,
|
| 314 |
camera: str,
|
| 315 |
use_ai: bool,
|
| 316 |
+
):
|
| 317 |
|
| 318 |
if not idea or not idea.strip():
|
| 319 |
+
|
| 320 |
return (
|
| 321 |
"",
|
| 322 |
+
NEGATIVE_PROMPT,
|
| 323 |
"⚠️ Enter an image idea first.",
|
| 324 |
)
|
| 325 |
|
| 326 |
idea = clean_text(idea)
|
| 327 |
|
| 328 |
if use_ai:
|
| 329 |
+
|
| 330 |
ai_prompt, error = build_ai_prompt(
|
| 331 |
+
idea,
|
| 332 |
+
style,
|
| 333 |
+
aspect,
|
| 334 |
+
mood,
|
| 335 |
+
detail,
|
| 336 |
+
camera,
|
| 337 |
)
|
| 338 |
|
| 339 |
if ai_prompt:
|
| 340 |
+
|
| 341 |
return (
|
| 342 |
ai_prompt,
|
| 343 |
+
NEGATIVE_PROMPT,
|
| 344 |
+
f"✓ Generated with {MODEL}",
|
| 345 |
)
|
| 346 |
|
| 347 |
+
local_prompt = build_local_prompt(
|
| 348 |
+
idea,
|
| 349 |
+
style,
|
| 350 |
+
aspect,
|
| 351 |
+
mood,
|
| 352 |
+
detail,
|
| 353 |
+
camera,
|
| 354 |
)
|
| 355 |
|
| 356 |
return (
|
| 357 |
local_prompt,
|
| 358 |
+
NEGATIVE_PROMPT,
|
| 359 |
+
f"✓ Built-in engine used • {error}",
|
| 360 |
)
|
| 361 |
|
| 362 |
+
local_prompt = build_local_prompt(
|
| 363 |
+
idea,
|
| 364 |
+
style,
|
| 365 |
+
aspect,
|
| 366 |
+
mood,
|
| 367 |
+
detail,
|
| 368 |
+
camera,
|
| 369 |
)
|
| 370 |
|
| 371 |
return (
|
| 372 |
local_prompt,
|
| 373 |
+
NEGATIVE_PROMPT,
|
| 374 |
+
"✓ Generated with built-in engine",
|
| 375 |
)
|
| 376 |
|
| 377 |
|
| 378 |
+
TRANSFORM_INSTRUCTIONS = {
|
|
|
|
|
|
|
| 379 |
|
|
|
|
| 380 |
"Improve": """
|
| 381 |
+
Improve this image-generation prompt.
|
| 382 |
|
| 383 |
Preserve the exact core subject and concept.
|
| 384 |
|
| 385 |
+
Increase useful visual specificity.
|
| 386 |
+
Strengthen composition, lighting, depth, atmosphere, materials and camera
|
| 387 |
+
language where appropriate.
|
| 388 |
|
| 389 |
+
Remove repetition and generic filler.
|
| 390 |
|
| 391 |
Do not radically change the concept.
|
| 392 |
|
| 393 |
+
Return ONLY the improved prompt.
|
| 394 |
""",
|
| 395 |
|
| 396 |
"Remix": """
|
| 397 |
+
Remix this image-generation prompt.
|
| 398 |
|
| 399 |
Preserve the core subject and concept.
|
| 400 |
|
| 401 |
+
Create a fresh interpretation by intelligently changing elements such as
|
| 402 |
+
environment, composition, lighting, atmosphere, color or camera perspective.
|
|
|
|
| 403 |
|
| 404 |
Do not replace the main subject.
|
| 405 |
|
| 406 |
+
Return ONLY the remixed prompt.
|
|
|
|
|
|
|
| 407 |
""",
|
| 408 |
|
| 409 |
"Simplify": """
|
| 410 |
+
Simplify this image-generation prompt.
|
| 411 |
|
| 412 |
+
Preserve the important subject, environment, style, mood and visual intent.
|
| 413 |
|
| 414 |
+
Remove unnecessary adjectives, repetition and generic AI language.
|
|
|
|
| 415 |
|
| 416 |
+
Make the prompt concise and effective.
|
| 417 |
|
| 418 |
+
Return ONLY the simplified prompt.
|
| 419 |
""",
|
| 420 |
}
|
| 421 |
|
| 422 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 423 |
def local_transform(
|
| 424 |
prompt: str,
|
| 425 |
action: str,
|
|
|
|
| 429 |
|
| 430 |
if action == "Improve":
|
| 431 |
|
|
|
|
|
|
|
| 432 |
addition = (
|
| 433 |
+
" Enhance coherent depth, intentional composition, "
|
| 434 |
"natural lighting, refined textures, believable materials, "
|
| 435 |
"and strong focal detail."
|
| 436 |
)
|
| 437 |
|
| 438 |
return (
|
| 439 |
+
prompt.rstrip(".") + "." + addition,
|
| 440 |
+
"✓ Improved locally",
|
| 441 |
)
|
| 442 |
|
| 443 |
if action == "Simplify":
|
|
|
|
| 452 |
|
| 453 |
return (
|
| 454 |
clean_text(prompt),
|
| 455 |
+
"✓ Simplified locally",
|
| 456 |
)
|
| 457 |
|
| 458 |
if action == "Remix":
|
| 459 |
|
| 460 |
+
choices = [
|
| 461 |
"with dramatic atmospheric depth",
|
| 462 |
"under cinematic environmental lighting",
|
| 463 |
"with a striking editorial composition",
|
| 464 |
"with subtle surreal visual elements",
|
| 465 |
"during a visually dramatic transitional moment",
|
|
|
|
| 466 |
"using an unconventional cinematic perspective",
|
| 467 |
]
|
| 468 |
|
| 469 |
+
addition = random.choice(choices)
|
|
|
|
|
|
|
| 470 |
|
| 471 |
return (
|
| 472 |
+
prompt.rstrip(".") + ", " + addition + ".",
|
| 473 |
+
"✓ Remixed locally",
|
| 474 |
)
|
| 475 |
|
| 476 |
return (
|
| 477 |
prompt,
|
| 478 |
+
"✓ Prompt ready",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 479 |
)
|
| 480 |
|
| 481 |
|
| 482 |
+
def transform_prompt(
|
| 483 |
+
prompt: str,
|
| 484 |
+
action: str,
|
| 485 |
+
) -> Tuple[str, str]:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 486 |
|
| 487 |
+
if not prompt or not prompt.strip():
|
|
|
|
|
|
|
|
|
|
| 488 |
|
| 489 |
+
return (
|
| 490 |
+
"",
|
| 491 |
+
"⚠️ Generate a prompt first.",
|
| 492 |
+
)
|
| 493 |
|
| 494 |
+
prompt = clean_text(prompt)
|
|
|
|
|
|
|
|
|
|
| 495 |
|
| 496 |
+
client = get_client()
|
| 497 |
+
instruction = TRANSFORM_INSTRUCTIONS.get(action)
|
|
|
|
| 498 |
|
| 499 |
+
if client is not None and instruction:
|
|
|
|
|
|
|
|
|
|
| 500 |
|
| 501 |
+
try:
|
|
|
|
|
|
|
|
|
|
|
|
|
| 502 |
|
| 503 |
+
response = client.chat_completion(
|
| 504 |
+
messages=[
|
| 505 |
+
{
|
| 506 |
+
"role": "system",
|
| 507 |
+
"content": (
|
| 508 |
+
"You are an expert image prompt editor. "
|
| 509 |
+
"Return only the requested prompt."
|
| 510 |
+
),
|
| 511 |
+
},
|
| 512 |
+
{
|
| 513 |
+
"role": "user",
|
| 514 |
+
"content": (
|
| 515 |
+
instruction
|
| 516 |
+
+ "\n\nCURRENT PROMPT:\n"
|
| 517 |
+
+ prompt
|
| 518 |
+
),
|
| 519 |
+
},
|
| 520 |
+
],
|
| 521 |
+
max_tokens=1200,
|
| 522 |
+
temperature=(
|
| 523 |
+
0.8
|
| 524 |
+
if action == "Remix"
|
| 525 |
+
else 0.5
|
| 526 |
+
),
|
| 527 |
+
)
|
| 528 |
|
| 529 |
+
if response and response.choices:
|
| 530 |
|
| 531 |
+
content = response.choices[0].message.content
|
|
|
|
|
|
|
| 532 |
|
| 533 |
+
if content:
|
|
|
|
|
|
|
|
|
|
| 534 |
|
| 535 |
+
content = clean_ai_response(content)
|
|
|
|
|
|
|
| 536 |
|
| 537 |
+
if content:
|
|
|
|
|
|
|
| 538 |
|
| 539 |
+
return (
|
| 540 |
+
content,
|
| 541 |
+
f"✓ {action} completed with Smart AI",
|
| 542 |
+
)
|
| 543 |
|
| 544 |
+
except Exception as exc:
|
| 545 |
|
| 546 |
+
print(
|
| 547 |
+
"[TRANSFORM ERROR]",
|
| 548 |
+
type(exc).__name__,
|
| 549 |
+
str(exc),
|
| 550 |
+
)
|
| 551 |
|
| 552 |
+
return local_transform(
|
| 553 |
+
prompt,
|
| 554 |
+
action,
|
| 555 |
)
|
| 556 |
|
|
|
|
|
|
|
|
|
|
| 557 |
|
| 558 |
+
def improve_prompt(prompt):
|
| 559 |
+
return transform_prompt(prompt, "Improve")
|
| 560 |
|
|
|
|
|
|
|
|
|
|
| 561 |
|
| 562 |
+
def remix_prompt(prompt):
|
| 563 |
+
return transform_prompt(prompt, "Remix")
|
| 564 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 565 |
|
| 566 |
+
def simplify_prompt(prompt):
|
| 567 |
+
return transform_prompt(prompt, "Simplify")
|
|
|
|
|
|
|
| 568 |
|
| 569 |
+
|
| 570 |
+
SURPRISE_IDEAS = [
|
| 571 |
+
"a tiny astronaut exploring an enormous abandoned greenhouse",
|
| 572 |
+
"a fox sitting inside an ancient observatory beneath a meteor shower",
|
| 573 |
+
"a futuristic Brisbane street transformed by bioluminescent plants",
|
| 574 |
+
"a mechanical
|