Update media_processing.py
Browse files- media_processing.py +310 -1109
media_processing.py
CHANGED
@@ -1,1167 +1,368 @@
|
|
|
|
|
|
|
|
|
|
1 |
import os
|
|
|
2 |
import base64
|
3 |
-
import cv2
|
4 |
-
import numpy as np
|
5 |
-
from PIL import Image
|
6 |
-
import pytesseract
|
7 |
import requests
|
8 |
-
from urllib.parse import urlparse, urljoin
|
9 |
-
from bs4 import BeautifulSoup
|
10 |
-
import html2text
|
11 |
-
import json
|
12 |
-
import time
|
13 |
-
import webbrowser
|
14 |
-
import urllib.parse
|
15 |
-
import copy
|
16 |
-
import html
|
17 |
import tempfile
|
18 |
-
import
|
19 |
-
import
|
20 |
-
import
|
21 |
-
import atexit
|
22 |
-
from huggingface_hub import HfApi
|
23 |
import gradio as gr
|
24 |
-
import subprocess
|
25 |
-
import re
|
26 |
-
|
27 |
-
# ---------------------------------------------------------------------------
|
28 |
-
# Video temp-file management (per-session tracking and cleanup)
|
29 |
-
# ---------------------------------------------------------------------------
|
30 |
-
VIDEO_TEMP_DIR = os.path.join(tempfile.gettempdir(), "anycoder_videos")
|
31 |
-
VIDEO_FILE_TTL_SECONDS = 6 * 60 * 60 # 6 hours
|
32 |
-
_SESSION_VIDEO_FILES: Dict[str, List[str]] = {}
|
33 |
-
_VIDEO_FILES_LOCK = threading.Lock()
|
34 |
-
|
35 |
-
def _ensure_video_dir_exists() -> None:
|
36 |
-
try:
|
37 |
-
os.makedirs(VIDEO_TEMP_DIR, exist_ok=True)
|
38 |
-
except Exception:
|
39 |
-
pass
|
40 |
-
|
41 |
-
def _register_video_for_session(session_id: Optional[str], file_path: str) -> None:
|
42 |
-
if not session_id or not file_path:
|
43 |
-
return
|
44 |
-
with _VIDEO_FILES_LOCK:
|
45 |
-
if session_id not in _SESSION_VIDEO_FILES:
|
46 |
-
_SESSION_VIDEO_FILES[session_id] = []
|
47 |
-
_SESSION_VIDEO_FILES[session_id].append(file_path)
|
48 |
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
with _VIDEO_FILES_LOCK:
|
53 |
-
file_list = _SESSION_VIDEO_FILES.pop(session_id, [])
|
54 |
-
for path in file_list:
|
55 |
-
try:
|
56 |
-
if path and os.path.exists(path):
|
57 |
-
os.unlink(path)
|
58 |
-
except Exception:
|
59 |
-
# Best-effort cleanup
|
60 |
-
pass
|
61 |
-
|
62 |
-
def reap_old_videos(ttl_seconds: int = VIDEO_FILE_TTL_SECONDS) -> None:
|
63 |
-
"""Delete old video files in the temp directory based on modification time."""
|
64 |
-
try:
|
65 |
-
_ensure_video_dir_exists()
|
66 |
-
now_ts = time.time()
|
67 |
-
for name in os.listdir(VIDEO_TEMP_DIR):
|
68 |
-
path = os.path.join(VIDEO_TEMP_DIR, name)
|
69 |
-
try:
|
70 |
-
if not os.path.isfile(path):
|
71 |
-
continue
|
72 |
-
mtime = os.path.getmtime(path)
|
73 |
-
if now_ts - mtime > ttl_seconds:
|
74 |
-
os.unlink(path)
|
75 |
-
except Exception:
|
76 |
-
pass
|
77 |
-
except Exception:
|
78 |
-
# Temp dir might not exist or be accessible; ignore
|
79 |
-
pass
|
80 |
-
|
81 |
-
# ---------------------------------------------------------------------------
|
82 |
-
# Audio temp-file management (per-session tracking and cleanup)
|
83 |
-
# ---------------------------------------------------------------------------
|
84 |
-
AUDIO_TEMP_DIR = os.path.join(tempfile.gettempdir(), "anycoder_audio")
|
85 |
-
AUDIO_FILE_TTL_SECONDS = 6 * 60 * 60 # 6 hours
|
86 |
-
_SESSION_AUDIO_FILES: Dict[str, List[str]] = {}
|
87 |
-
_AUDIO_FILES_LOCK = threading.Lock()
|
88 |
-
|
89 |
-
def _ensure_audio_dir_exists() -> None:
|
90 |
-
try:
|
91 |
-
os.makedirs(AUDIO_TEMP_DIR, exist_ok=True)
|
92 |
-
except Exception:
|
93 |
-
pass
|
94 |
-
|
95 |
-
def _register_audio_for_session(session_id: Optional[str], file_path: str) -> None:
|
96 |
-
if not session_id or not file_path:
|
97 |
-
return
|
98 |
-
with _AUDIO_FILES_LOCK:
|
99 |
-
if session_id not in _SESSION_AUDIO_FILES:
|
100 |
-
_SESSION_AUDIO_FILES[session_id] = []
|
101 |
-
_SESSION_AUDIO_FILES[session_id].append(file_path)
|
102 |
-
|
103 |
-
def cleanup_session_audio(session_id: Optional[str]) -> None:
|
104 |
-
if not session_id:
|
105 |
-
return
|
106 |
-
with _AUDIO_FILES_LOCK:
|
107 |
-
file_list = _SESSION_AUDIO_FILES.pop(session_id, [])
|
108 |
-
for path in file_list:
|
109 |
-
try:
|
110 |
-
if path and os.path.exists(path):
|
111 |
-
os.unlink(path)
|
112 |
-
except Exception:
|
113 |
-
pass
|
114 |
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
os.unlink(path)
|
127 |
-
except Exception:
|
128 |
-
pass
|
129 |
-
except Exception:
|
130 |
-
pass
|
131 |
-
|
132 |
-
# ---------------------------------------------------------------------------
|
133 |
-
# General temp media file management (per-session tracking and cleanup)
|
134 |
-
# ---------------------------------------------------------------------------
|
135 |
-
MEDIA_TEMP_DIR = os.path.join(tempfile.gettempdir(), "anycoder_media")
|
136 |
-
MEDIA_FILE_TTL_SECONDS = 6 * 60 * 60 # 6 hours
|
137 |
-
_SESSION_MEDIA_FILES: Dict[str, List[str]] = {}
|
138 |
-
_MEDIA_FILES_LOCK = threading.Lock()
|
139 |
-
|
140 |
-
# Global dictionary to store temporary media files for the session
|
141 |
-
temp_media_files = {}
|
142 |
-
|
143 |
-
def _ensure_media_dir_exists() -> None:
|
144 |
-
"""Ensure the media temp directory exists."""
|
145 |
-
try:
|
146 |
-
os.makedirs(MEDIA_TEMP_DIR, exist_ok=True)
|
147 |
-
except Exception:
|
148 |
-
pass
|
149 |
-
|
150 |
-
def track_session_media_file(session_id: Optional[str], file_path: str) -> None:
|
151 |
-
"""Track a media file for session-based cleanup."""
|
152 |
-
if not session_id or not file_path:
|
153 |
-
return
|
154 |
-
with _MEDIA_FILES_LOCK:
|
155 |
-
if session_id not in _SESSION_MEDIA_FILES:
|
156 |
-
_SESSION_MEDIA_FILES[session_id] = []
|
157 |
-
_SESSION_MEDIA_FILES[session_id].append(file_path)
|
158 |
-
|
159 |
-
def cleanup_session_media(session_id: Optional[str]) -> None:
|
160 |
-
"""Clean up media files for a specific session."""
|
161 |
-
if not session_id:
|
162 |
-
return
|
163 |
-
with _MEDIA_FILES_LOCK:
|
164 |
-
files_to_clean = _SESSION_MEDIA_FILES.pop(session_id, [])
|
165 |
|
166 |
-
|
|
|
|
|
167 |
try:
|
168 |
-
if
|
169 |
-
|
170 |
-
except Exception:
|
171 |
-
# Best-effort cleanup
|
172 |
-
pass
|
173 |
-
|
174 |
-
def reap_old_media(ttl_seconds: int = MEDIA_FILE_TTL_SECONDS) -> None:
|
175 |
-
"""Delete old media files in the temp directory based on modification time."""
|
176 |
-
try:
|
177 |
-
_ensure_media_dir_exists()
|
178 |
-
now_ts = time.time()
|
179 |
-
for name in os.listdir(MEDIA_TEMP_DIR):
|
180 |
-
path = os.path.join(MEDIA_TEMP_DIR, name)
|
181 |
-
if os.path.isfile(path):
|
182 |
-
try:
|
183 |
-
mtime = os.path.getmtime(path)
|
184 |
-
if (now_ts - mtime) > ttl_seconds:
|
185 |
-
os.unlink(path)
|
186 |
-
except Exception:
|
187 |
-
pass
|
188 |
-
except Exception:
|
189 |
-
# Temp dir might not exist or be accessible; ignore
|
190 |
-
pass
|
191 |
-
|
192 |
-
def cleanup_all_temp_media_on_startup() -> None:
|
193 |
-
"""Clean up all temporary media files on app startup."""
|
194 |
-
try:
|
195 |
-
# Clean up temp_media_files registry
|
196 |
-
temp_media_files.clear()
|
197 |
-
|
198 |
-
# Clean up actual files from disk (assume all are orphaned on startup)
|
199 |
-
_ensure_media_dir_exists()
|
200 |
-
for name in os.listdir(MEDIA_TEMP_DIR):
|
201 |
-
path = os.path.join(MEDIA_TEMP_DIR, name)
|
202 |
-
if os.path.isfile(path):
|
203 |
-
try:
|
204 |
-
os.unlink(path)
|
205 |
-
except Exception:
|
206 |
-
pass
|
207 |
-
|
208 |
-
# Clear session tracking
|
209 |
-
with _MEDIA_FILES_LOCK:
|
210 |
-
_SESSION_MEDIA_FILES.clear()
|
211 |
|
212 |
-
|
213 |
-
|
214 |
-
|
215 |
-
|
216 |
-
|
217 |
-
|
218 |
-
|
219 |
-
|
220 |
-
|
221 |
-
|
222 |
-
|
223 |
-
|
224 |
-
|
225 |
-
|
226 |
-
|
227 |
-
|
228 |
-
|
229 |
-
|
230 |
-
|
231 |
-
|
232 |
-
|
233 |
-
|
234 |
-
|
235 |
-
|
236 |
-
|
237 |
-
|
238 |
-
|
239 |
-
|
240 |
-
|
241 |
-
|
242 |
-
|
243 |
-
|
244 |
-
|
245 |
-
|
246 |
-
atexit.register(cleanup_all_temp_media_on_shutdown)
|
247 |
-
|
248 |
-
def create_temp_media_url(media_bytes: bytes, filename: str, media_type: str = "image", session_id: Optional[str] = None) -> str:
|
249 |
-
"""Create a temporary file and return a local URL for preview."""
|
250 |
-
try:
|
251 |
-
# Create unique filename with timestamp and UUID
|
252 |
-
timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
|
253 |
-
unique_id = str(uuid.uuid4())[:8]
|
254 |
-
base_name, ext = os.path.splitext(filename)
|
255 |
-
unique_filename = f"{media_type}_{timestamp}_{unique_id}_{base_name}{ext}"
|
256 |
-
|
257 |
-
# Create temporary file in the dedicated directory
|
258 |
-
_ensure_media_dir_exists()
|
259 |
-
temp_path = os.path.join(MEDIA_TEMP_DIR, unique_filename)
|
260 |
-
|
261 |
-
# Write media bytes to temporary file
|
262 |
-
with open(temp_path, 'wb') as f:
|
263 |
-
f.write(media_bytes)
|
264 |
-
|
265 |
-
# Track file for session-based cleanup
|
266 |
-
if session_id:
|
267 |
-
track_session_media_file(session_id, temp_path)
|
268 |
-
|
269 |
-
# Store the file info for later upload
|
270 |
-
file_id = f"{media_type}_{unique_id}"
|
271 |
-
temp_media_files[file_id] = {
|
272 |
-
'path': temp_path,
|
273 |
-
'filename': filename,
|
274 |
-
'media_type': media_type,
|
275 |
-
'media_bytes': media_bytes
|
276 |
-
}
|
277 |
-
|
278 |
-
# Return file:// URL for preview
|
279 |
-
file_url = f"file://{temp_path}"
|
280 |
-
print(f"[TempMedia] Created temporary {media_type} file: {file_url}")
|
281 |
-
return file_url
|
282 |
-
|
283 |
-
except Exception as e:
|
284 |
-
print(f"[TempMedia] Failed to create temporary file: {str(e)}")
|
285 |
-
return f"Error creating temporary {media_type} file: {str(e)}"
|
286 |
-
|
287 |
-
def upload_media_to_hf(media_bytes: bytes, filename: str, media_type: str = "image", token: gr.OAuthToken | None = None, use_temp: bool = True) -> str:
|
288 |
-
"""Upload media file to user's Hugging Face account or create temporary file."""
|
289 |
-
try:
|
290 |
-
# If use_temp is True, create temporary file for preview
|
291 |
-
if use_temp:
|
292 |
-
return create_temp_media_url(media_bytes, filename, media_type)
|
293 |
-
|
294 |
-
# Otherwise, upload to Hugging Face for permanent URL
|
295 |
-
# Try to get token from OAuth first, then fall back to environment variable
|
296 |
-
hf_token = None
|
297 |
-
if token and token.token:
|
298 |
-
hf_token = token.token
|
299 |
-
else:
|
300 |
-
hf_token = os.getenv('HF_TOKEN')
|
301 |
-
|
302 |
-
if not hf_token:
|
303 |
-
return "Error: Please log in with your Hugging Face account to upload media, or set HF_TOKEN environment variable."
|
304 |
-
|
305 |
-
# Initialize HF API
|
306 |
-
api = HfApi(token=hf_token)
|
307 |
-
|
308 |
-
# Get current user info to determine username
|
309 |
try:
|
310 |
-
|
311 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
312 |
except Exception as e:
|
313 |
-
print(f"
|
314 |
-
|
315 |
-
|
316 |
-
|
317 |
-
|
318 |
-
|
319 |
-
|
320 |
try:
|
321 |
-
|
322 |
-
|
323 |
-
|
324 |
-
|
325 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
326 |
)
|
327 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
328 |
except Exception as e:
|
329 |
-
print(f"[
|
330 |
-
|
331 |
-
|
332 |
-
|
333 |
-
|
334 |
-
|
335 |
-
base_name, ext = os.path.splitext(filename)
|
336 |
-
unique_filename = f"{media_type}/{timestamp}_{unique_id}_{base_name}{ext}"
|
337 |
-
|
338 |
-
# Create temporary file for upload
|
339 |
-
with tempfile.NamedTemporaryFile(delete=False, suffix=ext) as temp_file:
|
340 |
-
temp_file.write(media_bytes)
|
341 |
-
temp_path = temp_file.name
|
342 |
-
|
343 |
try:
|
344 |
-
|
345 |
-
|
346 |
-
|
347 |
-
|
348 |
-
|
349 |
-
|
350 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
351 |
)
|
352 |
|
353 |
-
|
354 |
-
permanent_url = f"https://huggingface.co/datasets/{repo_name}/resolve/main/{unique_filename}"
|
355 |
-
print(f"[HFUpload] Successfully uploaded {media_type} to {permanent_url}")
|
356 |
-
return permanent_url
|
357 |
|
358 |
-
|
359 |
-
|
360 |
-
|
361 |
-
|
362 |
-
|
363 |
-
|
364 |
-
|
365 |
-
|
366 |
-
|
367 |
-
|
368 |
-
|
369 |
-
|
370 |
-
|
371 |
-
|
372 |
-
|
373 |
-
|
374 |
-
|
375 |
-
|
376 |
-
|
377 |
-
|
378 |
-
|
379 |
-
|
380 |
-
|
381 |
-
|
382 |
-
|
383 |
-
|
384 |
-
|
385 |
-
|
386 |
-
|
387 |
-
|
388 |
-
|
389 |
-
|
390 |
-
|
391 |
-
|
392 |
-
|
393 |
-
|
394 |
-
|
395 |
-
else
|
396 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
397 |
|
398 |
-
except Exception as e:
|
399 |
-
print(f"[DeployUpload] Error uploading {file_id}: {str(e)}")
|
400 |
-
continue
|
401 |
-
|
402 |
-
# Clean up temporary files after upload
|
403 |
-
cleanup_temp_media_files()
|
404 |
-
|
405 |
-
return updated_content
|
406 |
-
|
407 |
-
except Exception as e:
|
408 |
-
print(f"[DeployUpload] Failed to upload temporary files: {str(e)}")
|
409 |
-
return html_content
|
410 |
-
|
411 |
-
def cleanup_temp_media_files():
|
412 |
-
"""Clean up temporary media files from disk and memory."""
|
413 |
-
try:
|
414 |
-
for file_id, file_info in temp_media_files.items():
|
415 |
try:
|
416 |
-
|
417 |
-
os.remove(file_info['path'])
|
418 |
-
print(f"[TempCleanup] Removed {file_info['path']}")
|
419 |
except Exception as e:
|
420 |
-
|
421 |
-
|
422 |
-
|
423 |
-
|
424 |
-
|
425 |
-
|
426 |
-
|
427 |
-
|
428 |
-
|
429 |
-
|
430 |
-
|
431 |
-
|
432 |
-
|
433 |
-
|
434 |
-
|
435 |
-
|
436 |
-
|
437 |
-
|
438 |
-
|
439 |
-
|
440 |
-
bill_to="huggingface",
|
441 |
-
)
|
442 |
-
|
443 |
-
# Generate image using Qwen/Qwen-Image model
|
444 |
-
image = client.text_to_image(
|
445 |
-
prompt,
|
446 |
-
model="Qwen/Qwen-Image",
|
447 |
-
)
|
448 |
-
|
449 |
-
# Resize image to reduce size while maintaining quality
|
450 |
-
max_size = 1024 # Increased size since we're not using data URIs
|
451 |
-
if image.width > max_size or image.height > max_size:
|
452 |
-
image.thumbnail((max_size, max_size), Image.Resampling.LANCZOS)
|
453 |
-
|
454 |
-
# Convert PIL Image to bytes for upload
|
455 |
-
import io
|
456 |
-
buffer = io.BytesIO()
|
457 |
-
# Save as JPEG with good quality since we're not embedding
|
458 |
-
image.convert('RGB').save(buffer, format='JPEG', quality=90, optimize=True)
|
459 |
-
image_bytes = buffer.getvalue()
|
460 |
-
|
461 |
-
# Create temporary URL for preview (will be uploaded to HF during deploy)
|
462 |
-
filename = f"generated_image_{image_index}.jpg"
|
463 |
-
temp_url = upload_media_to_hf(image_bytes, filename, "image", token, use_temp=True)
|
464 |
-
|
465 |
-
# Check if creation was successful
|
466 |
-
if temp_url.startswith("Error"):
|
467 |
-
return temp_url
|
468 |
-
|
469 |
-
# Return HTML img tag with temporary URL
|
470 |
-
return f'<img src="{temp_url}" alt="{prompt}" style="max-width: 100%; height: auto; border-radius: 8px; margin: 10px 0;" loading="lazy" />'
|
471 |
-
|
472 |
-
except Exception as e:
|
473 |
-
print(f"Image generation error: {str(e)}")
|
474 |
-
return f"Error generating image: {str(e)}"
|
475 |
-
|
476 |
-
def generate_image_to_image(input_image_data, prompt: str, token: gr.OAuthToken | None = None) -> str:
|
477 |
-
"""Generate an image using image-to-image with Qwen-Image-Edit via Hugging Face InferenceClient."""
|
478 |
-
try:
|
479 |
-
# Check token
|
480 |
-
if not os.getenv('HF_TOKEN'):
|
481 |
-
return "Error: HF_TOKEN environment variable is not set. Please set it to your Hugging Face API token."
|
482 |
-
|
483 |
-
# Prepare client
|
484 |
-
client = InferenceClient(
|
485 |
-
provider="auto",
|
486 |
-
api_key=os.getenv('HF_TOKEN'),
|
487 |
-
bill_to="huggingface",
|
488 |
-
)
|
489 |
-
|
490 |
-
# Normalize input image to bytes
|
491 |
-
import io
|
492 |
-
from PIL import Image
|
493 |
-
try:
|
494 |
-
import numpy as np
|
495 |
-
except Exception:
|
496 |
-
np = None
|
497 |
-
|
498 |
if hasattr(input_image_data, 'read'):
|
499 |
-
# File-like object
|
500 |
raw = input_image_data.read()
|
501 |
pil_image = Image.open(io.BytesIO(raw))
|
502 |
elif hasattr(input_image_data, 'mode') and hasattr(input_image_data, 'size'):
|
503 |
-
# PIL Image
|
504 |
pil_image = input_image_data
|
505 |
-
elif
|
506 |
pil_image = Image.fromarray(input_image_data)
|
507 |
elif isinstance(input_image_data, (bytes, bytearray)):
|
508 |
pil_image = Image.open(io.BytesIO(input_image_data))
|
509 |
else:
|
510 |
-
# Fallback: try to convert via bytes
|
511 |
pil_image = Image.open(io.BytesIO(bytes(input_image_data)))
|
512 |
-
|
513 |
# Ensure RGB
|
514 |
if pil_image.mode != 'RGB':
|
515 |
pil_image = pil_image.convert('RGB')
|
516 |
-
|
517 |
-
# Resize input image to avoid request body size limits
|
518 |
-
max_input_size = 1024
|
519 |
-
if pil_image.width > max_input_size or pil_image.height > max_input_size:
|
520 |
-
pil_image.thumbnail((max_input_size, max_input_size), Image.Resampling.LANCZOS)
|
521 |
-
|
522 |
-
buf = io.BytesIO()
|
523 |
-
pil_image.save(buf, format='JPEG', quality=85, optimize=True)
|
524 |
-
input_bytes = buf.getvalue()
|
525 |
-
|
526 |
-
# Call image-to-image
|
527 |
-
image = client.image_to_image(
|
528 |
-
input_bytes,
|
529 |
-
prompt=prompt,
|
530 |
-
model="Qwen/Qwen-Image-Edit",
|
531 |
-
)
|
532 |
-
|
533 |
-
# Resize/optimize (larger since not using data URIs)
|
534 |
-
max_size = 1024
|
535 |
-
if image.width > max_size or image.height > max_size:
|
536 |
-
image.thumbnail((max_size, max_size), Image.Resampling.LANCZOS)
|
537 |
-
|
538 |
-
out_buf = io.BytesIO()
|
539 |
-
image.convert('RGB').save(out_buf, format='JPEG', quality=90, optimize=True)
|
540 |
-
image_bytes = out_buf.getvalue()
|
541 |
-
|
542 |
-
# Create temporary URL for preview (will be uploaded to HF during deploy)
|
543 |
-
filename = "image_to_image_result.jpg"
|
544 |
-
temp_url = upload_media_to_hf(image_bytes, filename, "image", token, use_temp=True)
|
545 |
|
546 |
-
|
547 |
-
|
548 |
-
|
549 |
-
|
550 |
-
|
551 |
-
|
552 |
-
print(f"Image-to-image generation error: {str(e)}")
|
553 |
-
return f"Error generating image (image-to-image): {str(e)}"
|
554 |
-
|
555 |
-
def generate_video_from_image(input_image_data, prompt: str, session_id: Optional[str] = None, token: gr.OAuthToken | None = None) -> str:
|
556 |
-
"""Generate a video from an input image and prompt using Hugging Face InferenceClient."""
|
557 |
-
try:
|
558 |
-
print("[Image2Video] Starting video generation")
|
559 |
-
if not os.getenv('HF_TOKEN'):
|
560 |
-
print("[Image2Video] Missing HF_TOKEN")
|
561 |
-
return "Error: HF_TOKEN environment variable is not set. Please set it to your Hugging Face API token."
|
562 |
-
|
563 |
-
# Prepare client
|
564 |
-
client = InferenceClient(
|
565 |
-
provider="auto",
|
566 |
-
api_key=os.getenv('HF_TOKEN'),
|
567 |
-
bill_to="huggingface",
|
568 |
-
)
|
569 |
-
print(f"[Image2Video] InferenceClient initialized (provider=auto)")
|
570 |
-
|
571 |
-
# Normalize input image to bytes, with downscale/compress to cap request size
|
572 |
-
import io
|
573 |
-
from PIL import Image
|
574 |
-
try:
|
575 |
-
import numpy as np
|
576 |
-
except Exception:
|
577 |
-
np = None
|
578 |
-
|
579 |
-
def _load_pil(img_like) -> Image.Image:
|
580 |
-
if hasattr(img_like, 'read'):
|
581 |
-
return Image.open(io.BytesIO(img_like.read()))
|
582 |
-
if hasattr(img_like, 'mode') and hasattr(img_like, 'size'):
|
583 |
-
return img_like
|
584 |
-
if np is not None and isinstance(img_like, np.ndarray):
|
585 |
-
return Image.fromarray(img_like)
|
586 |
-
if isinstance(img_like, (bytes, bytearray)):
|
587 |
-
return Image.open(io.BytesIO(img_like))
|
588 |
-
return Image.open(io.BytesIO(bytes(img_like)))
|
589 |
-
|
590 |
-
pil_image = _load_pil(input_image_data)
|
591 |
-
if pil_image.mode != 'RGB':
|
592 |
-
pil_image = pil_image.convert('RGB')
|
593 |
-
try:
|
594 |
-
print(f"[Image2Video] Input PIL image size={pil_image.size} mode={pil_image.mode}")
|
595 |
-
except Exception:
|
596 |
-
pass
|
597 |
-
|
598 |
-
# Progressive encode to keep payload under ~3.9MB (below 4MB limit)
|
599 |
-
MAX_BYTES = 3_900_000
|
600 |
-
max_dim = 1024 # initial cap on longest edge
|
601 |
quality = 90
|
602 |
-
|
603 |
def encode_current(pil: Image.Image, q: int) -> bytes:
|
604 |
tmp = io.BytesIO()
|
605 |
pil.save(tmp, format='JPEG', quality=q, optimize=True)
|
606 |
return tmp.getvalue()
|
607 |
-
|
608 |
-
# Downscale while
|
609 |
while max(pil_image.size) > max_dim:
|
610 |
ratio = max_dim / float(max(pil_image.size))
|
611 |
new_size = (max(1, int(pil_image.size[0] * ratio)), max(1, int(pil_image.size[1] * ratio)))
|
612 |
pil_image = pil_image.resize(new_size, Image.Resampling.LANCZOS)
|
613 |
-
|
614 |
encoded = encode_current(pil_image, quality)
|
615 |
-
|
|
|
616 |
while len(encoded) > MAX_BYTES and (quality > 40 or max(pil_image.size) > 640):
|
617 |
if quality > 40:
|
618 |
quality -= 10
|
619 |
else:
|
620 |
-
# reduce dims by 15% if already at low quality
|
621 |
new_w = max(1, int(pil_image.size[0] * 0.85))
|
622 |
new_h = max(1, int(pil_image.size[1] * 0.85))
|
623 |
pil_image = pil_image.resize((new_w, new_h), Image.Resampling.LANCZOS)
|
624 |
encoded = encode_current(pil_image, quality)
|
625 |
-
|
626 |
-
input_bytes = encoded
|
627 |
-
|
628 |
-
# Call image-to-video; require method support
|
629 |
-
model_id = "Lightricks/LTX-Video-0.9.8-13B-distilled"
|
630 |
-
image_to_video_method = getattr(client, "image_to_video", None)
|
631 |
-
if not callable(image_to_video_method):
|
632 |
-
print("[Image2Video] InferenceClient.image_to_video not available in this huggingface_hub version")
|
633 |
-
return (
|
634 |
-
"Error generating video (image-to-video): Your installed huggingface_hub version "
|
635 |
-
"does not expose InferenceClient.image_to_video. Please upgrade with "
|
636 |
-
"`pip install -U huggingface_hub` and try again."
|
637 |
-
)
|
638 |
-
print(f"[Image2Video] Calling image_to_video with model={model_id}, prompt length={len(prompt or '')}")
|
639 |
-
video_bytes = image_to_video_method(
|
640 |
-
input_bytes,
|
641 |
-
prompt=prompt,
|
642 |
-
model=model_id,
|
643 |
-
)
|
644 |
-
print(f"[Image2Video] Received video bytes: {len(video_bytes) if hasattr(video_bytes, '__len__') else 'unknown length'}")
|
645 |
-
|
646 |
-
# Create temporary URL for preview (will be uploaded to HF during deploy)
|
647 |
-
filename = "image_to_video_result.mp4"
|
648 |
-
temp_url = upload_media_to_hf(video_bytes, filename, "video", token, use_temp=True)
|
649 |
-
|
650 |
-
# Check if creation was successful
|
651 |
-
if temp_url.startswith("Error"):
|
652 |
-
return temp_url
|
653 |
-
|
654 |
-
video_html = (
|
655 |
-
f'<video controls autoplay muted loop playsinline '
|
656 |
-
f'style="max-width: 100%; height: auto; border-radius: 8px; margin: 10px 0; display: block;" '
|
657 |
-
f'onloadstart="this.style.backgroundColor=\'#f0f0f0\'" '
|
658 |
-
f'onerror="this.style.display=\'none\'; console.error(\'Video failed to load\')">'
|
659 |
-
f'<source src="{temp_url}" type="video/mp4" />'
|
660 |
-
f'<p style="text-align: center; color: #666;">Your browser does not support the video tag.</p>'
|
661 |
-
f'</video>'
|
662 |
-
)
|
663 |
-
|
664 |
-
print(f"[Image2Video] Successfully generated video HTML tag with temporary URL: {temp_url}")
|
665 |
-
|
666 |
-
# Validate the generated video HTML
|
667 |
-
if not validate_video_html(video_html):
|
668 |
-
print("[Image2Video] Generated video HTML failed validation")
|
669 |
-
return "Error: Generated video HTML is malformed"
|
670 |
-
|
671 |
-
return video_html
|
672 |
-
except Exception as e:
|
673 |
-
import traceback
|
674 |
-
print("[Image2Video] Exception during generation:")
|
675 |
-
traceback.print_exc()
|
676 |
-
print(f"Image-to-video generation error: {str(e)}")
|
677 |
-
return f"Error generating video (image-to-video): {str(e)}"
|
678 |
-
|
679 |
-
def generate_video_from_text(prompt: str, session_id: Optional[str] = None, token: gr.OAuthToken | None = None) -> str:
|
680 |
-
"""Generate a video from a text prompt using Hugging Face InferenceClient."""
|
681 |
-
try:
|
682 |
-
print("[Text2Video] Starting video generation from text")
|
683 |
-
if not os.getenv('HF_TOKEN'):
|
684 |
-
print("[Text2Video] Missing HF_TOKEN")
|
685 |
-
return "Error: HF_TOKEN environment variable is not set. Please set it to your Hugging Face API token."
|
686 |
-
|
687 |
-
client = InferenceClient(
|
688 |
-
provider="auto",
|
689 |
-
api_key=os.getenv('HF_TOKEN'),
|
690 |
-
bill_to="huggingface",
|
691 |
-
)
|
692 |
-
print("[Text2Video] InferenceClient initialized (provider=auto)")
|
693 |
-
|
694 |
-
# Ensure the client has text_to_video (newer huggingface_hub)
|
695 |
-
text_to_video_method = getattr(client, "text_to_video", None)
|
696 |
-
if not callable(text_to_video_method):
|
697 |
-
print("[Text2Video] InferenceClient.text_to_video not available in this huggingface_hub version")
|
698 |
-
return (
|
699 |
-
"Error generating video (text-to-video): Your installed huggingface_hub version "
|
700 |
-
"does not expose InferenceClient.text_to_video. Please upgrade with "
|
701 |
-
"`pip install -U huggingface_hub` and try again."
|
702 |
-
)
|
703 |
-
|
704 |
-
model_id = "Wan-AI/Wan2.2-T2V-A14B"
|
705 |
-
prompt_str = (prompt or "").strip()
|
706 |
-
print(f"[Text2Video] Calling text_to_video with model={model_id}, prompt length={len(prompt_str)}")
|
707 |
-
video_bytes = text_to_video_method(
|
708 |
-
prompt_str,
|
709 |
-
model=model_id,
|
710 |
-
)
|
711 |
-
print(f"[Text2Video] Received video bytes: {len(video_bytes) if hasattr(video_bytes, '__len__') else 'unknown length'}")
|
712 |
-
|
713 |
-
# Create temporary URL for preview (will be uploaded to HF during deploy)
|
714 |
-
filename = "text_to_video_result.mp4"
|
715 |
-
temp_url = upload_media_to_hf(video_bytes, filename, "video", token, use_temp=True)
|
716 |
-
|
717 |
-
# Check if creation was successful
|
718 |
-
if temp_url.startswith("Error"):
|
719 |
-
return temp_url
|
720 |
-
|
721 |
-
video_html = (
|
722 |
-
f'<video controls autoplay muted loop playsinline '
|
723 |
-
f'style="max-width: 100%; height: auto; border-radius: 8px; margin: 10px 0; display: block;" '
|
724 |
-
f'onloadstart="this.style.backgroundColor=\'#f0f0f0\'" '
|
725 |
-
f'onerror="this.style.display=\'none\'; console.error(\'Video failed to load\')">'
|
726 |
-
f'<source src="{temp_url}" type="video/mp4" />'
|
727 |
-
f'<p style="text-align: center; color: #666;">Your browser does not support the video tag.</p>'
|
728 |
-
f'</video>'
|
729 |
-
)
|
730 |
-
|
731 |
-
print(f"[Text2Video] Successfully generated video HTML tag with temporary URL: {temp_url}")
|
732 |
-
|
733 |
-
# Validate the generated video HTML
|
734 |
-
if not validate_video_html(video_html):
|
735 |
-
print("[Text2Video] Generated video HTML failed validation")
|
736 |
-
return "Error: Generated video HTML is malformed"
|
737 |
-
|
738 |
-
return video_html
|
739 |
-
except Exception as e:
|
740 |
-
import traceback
|
741 |
-
print("[Text2Video] Exception during generation:")
|
742 |
-
traceback.print_exc()
|
743 |
-
print(f"Text-to-video generation error: {str(e)}")
|
744 |
-
return f"Error generating video (text-to-video): {str(e)}"
|
745 |
-
|
746 |
-
def generate_music_from_text(prompt: str, music_length_ms: int = 30000, session_id: Optional[str] = None, token: gr.OAuthToken | None = None) -> str:
|
747 |
-
"""Generate music from a text prompt using ElevenLabs Music API and return an HTML <audio> tag."""
|
748 |
-
try:
|
749 |
-
api_key = os.getenv('ELEVENLABS_API_KEY')
|
750 |
-
if not api_key:
|
751 |
-
return "Error: ELEVENLABS_API_KEY environment variable is not set."
|
752 |
-
|
753 |
-
headers = {
|
754 |
-
'Content-Type': 'application/json',
|
755 |
-
'xi-api-key': api_key,
|
756 |
-
}
|
757 |
-
payload = {
|
758 |
-
'prompt': (prompt or 'Epic orchestral theme with soaring strings and powerful brass'),
|
759 |
-
'music_length_ms': int(music_length_ms) if music_length_ms else 30000,
|
760 |
-
}
|
761 |
-
|
762 |
-
resp = requests.post('https://api.elevenlabs.io/v1/music/compose', headers=headers, json=payload)
|
763 |
-
try:
|
764 |
-
resp.raise_for_status()
|
765 |
-
except Exception as e:
|
766 |
-
return f"Error generating music: {getattr(e, 'response', resp).text if hasattr(e, 'response') else resp.text}"
|
767 |
-
|
768 |
-
# Create temporary URL for preview (will be uploaded to HF during deploy)
|
769 |
-
filename = "generated_music.mp3"
|
770 |
-
temp_url = upload_media_to_hf(resp.content, filename, "audio", token, use_temp=True)
|
771 |
-
|
772 |
-
# Check if creation was successful
|
773 |
-
if temp_url.startswith("Error"):
|
774 |
-
return temp_url
|
775 |
|
776 |
-
|
777 |
-
"<div class=\"anycoder-music\" style=\"max-width:420px;margin:16px auto;padding:12px 16px;border:1px solid #e5e7eb;border-radius:12px;background:linear-gradient(180deg,#fafafa,#f3f4f6);box-shadow:0 2px 8px rgba(0,0,0,0.06)\">"
|
778 |
-
" <div style=\"font-size:13px;color:#374151;margin-bottom:8px;display:flex:align-items:center;gap:6px\">"
|
779 |
-
" <span>🎵 Generated music</span>"
|
780 |
-
" </div>"
|
781 |
-
f" <audio controls autoplay loop style=\"width:100%;outline:none;\">"
|
782 |
-
f" <source src=\"{temp_url}\" type=\"audio/mpeg\" />"
|
783 |
-
" Your browser does not support the audio element."
|
784 |
-
" </audio>"
|
785 |
-
"</div>"
|
786 |
-
)
|
787 |
-
|
788 |
-
print(f"[Music] Successfully generated music HTML tag with temporary URL: {temp_url}")
|
789 |
-
return audio_html
|
790 |
-
except Exception as e:
|
791 |
-
return f"Error generating music: {str(e)}"
|
792 |
-
|
793 |
-
def extract_image_prompts_from_text(text: str, num_images_needed: int = 1) -> list:
|
794 |
-
"""Extract image generation prompts from the full text based on number of images needed"""
|
795 |
-
# Use the entire text as the base prompt for image generation
|
796 |
-
# Clean up the text and create variations for the required number of images
|
797 |
-
|
798 |
-
# Clean the text
|
799 |
-
cleaned_text = text.strip()
|
800 |
-
if not cleaned_text:
|
801 |
-
return []
|
802 |
-
|
803 |
-
# Create variations of the prompt for the required number of images
|
804 |
-
prompts = []
|
805 |
-
|
806 |
-
# Generate exactly the number of images needed
|
807 |
-
for i in range(num_images_needed):
|
808 |
-
if i == 0:
|
809 |
-
# First image: Use the full prompt as-is
|
810 |
-
prompts.append(cleaned_text)
|
811 |
-
elif i == 1:
|
812 |
-
# Second image: Add "visual representation" to make it more image-focused
|
813 |
-
prompts.append(f"Visual representation of {cleaned_text}")
|
814 |
-
elif i == 2:
|
815 |
-
# Third image: Add "illustration" to create a different style
|
816 |
-
prompts.append(f"Illustration of {cleaned_text}")
|
817 |
-
else:
|
818 |
-
# For additional images, use different variations
|
819 |
-
variations = [
|
820 |
-
f"Digital art of {cleaned_text}",
|
821 |
-
f"Modern design of {cleaned_text}",
|
822 |
-
f"Professional illustration of {cleaned_text}",
|
823 |
-
f"Clean design of {cleaned_text}",
|
824 |
-
f"Beautiful visualization of {cleaned_text}",
|
825 |
-
f"Stylish representation of {cleaned_text}",
|
826 |
-
f"Contemporary design of {cleaned_text}",
|
827 |
-
f"Elegant illustration of {cleaned_text}"
|
828 |
-
]
|
829 |
-
variation_index = (i - 3) % len(variations)
|
830 |
-
prompts.append(variations[variation_index])
|
831 |
-
|
832 |
-
return prompts
|
833 |
-
|
834 |
-
def create_image_replacement_blocks(html_content: str, user_prompt: str) -> str:
|
835 |
-
"""Create search/replace blocks to replace placeholder images with generated Qwen images"""
|
836 |
-
if not user_prompt:
|
837 |
-
return ""
|
838 |
-
|
839 |
-
# Find existing image placeholders in the HTML first
|
840 |
-
import re
|
841 |
-
|
842 |
-
# Common patterns for placeholder images
|
843 |
-
placeholder_patterns = [
|
844 |
-
r'<img[^>]*src=["\'](?:placeholder|dummy|sample|example)[^"\']*["\'][^>]*>',
|
845 |
-
r'<img[^>]*src=["\']https?://via\.placeholder\.com[^"\']*["\'][^>]*>',
|
846 |
-
r'<img[^>]*src=["\']https?://picsum\.photos[^"\']*["\'][^>]*>',
|
847 |
-
r'<img[^>]*src=["\']https?://dummyimage\.com[^"\']*["\'][^>]*>',
|
848 |
-
r'<img[^>]*alt=["\'][^"\']*placeholder[^"\']*["\'][^>]*>',
|
849 |
-
r'<img[^>]*class=["\'][^"\']*placeholder[^"\']*["\'][^>]*>',
|
850 |
-
r'<img[^>]*id=["\'][^"\']*placeholder[^"\']*["\'][^>]*>',
|
851 |
-
r'<img[^>]*src=["\']data:image[^"\']*["\'][^>]*>', # Base64 images
|
852 |
-
r'<img[^>]*src=["\']#["\'][^>]*>', # Empty src
|
853 |
-
r'<img[^>]*src=["\']about:blank["\'][^>]*>', # About blank
|
854 |
-
]
|
855 |
-
|
856 |
-
# Find all placeholder images
|
857 |
-
placeholder_images = []
|
858 |
-
for pattern in placeholder_patterns:
|
859 |
-
matches = re.findall(pattern, html_content, re.IGNORECASE)
|
860 |
-
placeholder_images.extend(matches)
|
861 |
-
|
862 |
-
# Filter out HF URLs from placeholders (they are real generated content)
|
863 |
-
placeholder_images = [img for img in placeholder_images if 'huggingface.co/datasets/' not in img]
|
864 |
-
|
865 |
-
# If no placeholder images found, look for any img tags
|
866 |
-
if not placeholder_images:
|
867 |
-
img_pattern = r'<img[^>]*>'
|
868 |
-
# Case-insensitive to catch <IMG> or mixed-case tags
|
869 |
-
placeholder_images = re.findall(img_pattern, html_content, re.IGNORECASE)
|
870 |
-
|
871 |
-
# Also look for div elements that might be image placeholders
|
872 |
-
div_placeholder_patterns = [
|
873 |
-
r'<div[^>]*class=["\'][^"\']*(?:image|img|photo|picture)[^"\']*["\'][^>]*>.*?</div>',
|
874 |
-
r'<div[^>]*id=["\'][^"\']*(?:image|img|photo|picture)[^"\']*["\'][^>]*>.*?</div>',
|
875 |
-
]
|
876 |
-
|
877 |
-
for pattern in div_placeholder_patterns:
|
878 |
-
matches = re.findall(pattern, html_content, re.IGNORECASE | re.DOTALL)
|
879 |
-
placeholder_images.extend(matches)
|
880 |
-
|
881 |
-
# Count how many images we need to generate
|
882 |
-
num_images_needed = len(placeholder_images)
|
883 |
-
|
884 |
-
if num_images_needed == 0:
|
885 |
-
return ""
|
886 |
-
|
887 |
-
# Generate image prompts based on the number of images found
|
888 |
-
image_prompts = extract_image_prompts_from_text(user_prompt, num_images_needed)
|
889 |
-
|
890 |
-
# Generate images for each prompt
|
891 |
-
generated_images = []
|
892 |
-
for i, prompt in enumerate(image_prompts):
|
893 |
-
image_html = generate_image_with_qwen(prompt, i, token=None) # TODO: Pass token from parent context
|
894 |
-
if not image_html.startswith("Error"):
|
895 |
-
generated_images.append((i, image_html))
|
896 |
-
|
897 |
-
if not generated_images:
|
898 |
-
return ""
|
899 |
-
|
900 |
-
# Create search/replace blocks
|
901 |
-
replacement_blocks = []
|
902 |
-
|
903 |
-
for i, (prompt_index, generated_image) in enumerate(generated_images):
|
904 |
-
if i < len(placeholder_images):
|
905 |
-
# Replace existing placeholder
|
906 |
-
placeholder = placeholder_images[i]
|
907 |
-
# Clean up the placeholder for better matching
|
908 |
-
placeholder_clean = re.sub(r'\s+', ' ', placeholder.strip())
|
909 |
-
|
910 |
-
# Try multiple variations of the placeholder for better matching
|
911 |
-
placeholder_variations = [
|
912 |
-
placeholder_clean,
|
913 |
-
placeholder_clean.replace('"', "'"),
|
914 |
-
placeholder_clean.replace("'", '"'),
|
915 |
-
re.sub(r'\s+', ' ', placeholder_clean),
|
916 |
-
placeholder_clean.replace(' ', ' '),
|
917 |
-
]
|
918 |
-
|
919 |
-
# Create a replacement block for each variation
|
920 |
-
for variation in placeholder_variations:
|
921 |
-
replacement_blocks.append(f"""{SEARCH_START}
|
922 |
-
{variation}
|
923 |
-
{DIVIDER}
|
924 |
-
{generated_image}
|
925 |
-
{REPLACE_END}""")
|
926 |
-
else:
|
927 |
-
# Add new image if we have more generated images than placeholders
|
928 |
-
# Find a good insertion point (after body tag or main content)
|
929 |
-
if '<body' in html_content:
|
930 |
-
body_end = html_content.find('>', html_content.find('<body')) + 1
|
931 |
-
insertion_point = html_content[:body_end] + '\n '
|
932 |
-
replacement_blocks.append(f"""{SEARCH_START}
|
933 |
-
{insertion_point}
|
934 |
-
{DIVIDER}
|
935 |
-
{insertion_point}
|
936 |
-
{generated_image}
|
937 |
-
{REPLACE_END}""")
|
938 |
|
939 |
-
|
940 |
-
|
941 |
-
|
942 |
-
|
943 |
-
|
944 |
-
|
945 |
-
|
946 |
-
|
947 |
-
|
948 |
-
# Detect placeholders similarly to the multi-image version
|
949 |
-
placeholder_patterns = [
|
950 |
-
r'<img[^>]*src=["\'](?:placeholder|dummy|sample|example)[^"\']*["\'][^>]*>',
|
951 |
-
r'<img[^>]*src=["\']https?://via\.placeholder\.com[^"\']*["\'][^>]*>',
|
952 |
-
r'<img[^>]*src=["\']https?://picsum\.photos[^"\']*["\'][^>]*>',
|
953 |
-
r'<img[^>]*src=["\']https?://dummyimage\.com[^"\']*["\'][^>]*>',
|
954 |
-
r'<img[^>]*alt=["\'][^"\']*placeholder[^"\']*["\'][^>]*>',
|
955 |
-
r'<img[^>]*class=["\'][^"\']*placeholder[^"\']*["\'][^>]*>',
|
956 |
-
r'<img[^>]*id=["\'][^"\']*placeholder[^"\']*["\'][^>]*>',
|
957 |
-
r'<img[^>]*src=["\']data:image[^"\']*["\'][^>]*>',
|
958 |
-
r'<img[^>]*src=["\']#["\'][^>]*>',
|
959 |
-
r'<img[^>]*src=["\']about:blank["\'][^>]*>',
|
960 |
-
]
|
961 |
-
|
962 |
-
placeholder_images = []
|
963 |
-
for pattern in placeholder_patterns:
|
964 |
-
matches = re.findall(pattern, html_content, re.IGNORECASE)
|
965 |
-
if matches:
|
966 |
-
placeholder_images.extend(matches)
|
967 |
|
968 |
-
|
969 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
970 |
|
971 |
-
|
972 |
-
|
973 |
-
|
974 |
-
|
975 |
-
|
976 |
-
|
977 |
-
|
978 |
-
|
979 |
-
|
980 |
-
|
981 |
-
|
982 |
-
|
983 |
-
|
984 |
-
|
985 |
-
|
986 |
-
|
987 |
-
|
988 |
-
|
989 |
-
|
990 |
-
|
991 |
-
|
992 |
-
|
993 |
-
|
994 |
-
|
995 |
-
|
996 |
-
|
997 |
-
|
998 |
-
|
999 |
-
|
1000 |
-
|
1001 |
-
|
1002 |
-
|
1003 |
-
|
1004 |
-
# Otherwise insert after <body>
|
1005 |
-
if '<body' in html_content:
|
1006 |
-
body_end = html_content.find('>', html_content.find('<body')) + 1
|
1007 |
-
insertion_point = html_content[:body_end] + '\n '
|
1008 |
-
return f"""{SEARCH_START}
|
1009 |
-
{insertion_point}
|
1010 |
-
{DIVIDER}
|
1011 |
-
{insertion_point}
|
1012 |
-
{image_html}
|
1013 |
-
{REPLACE_END}"""
|
1014 |
-
|
1015 |
-
# If no <body>, just append
|
1016 |
-
return f"{SEARCH_START}\n\n{DIVIDER}\n{image_html}\n{REPLACE_END}"
|
1017 |
-
|
1018 |
-
def create_video_replacement_blocks_text_to_video(html_content: str, prompt: str, session_id: Optional[str] = None) -> str:
|
1019 |
-
"""Create search/replace blocks that generate and insert ONLY ONE text-to-video result."""
|
1020 |
-
if not prompt or not prompt.strip():
|
1021 |
-
return ""
|
1022 |
-
|
1023 |
-
import re
|
1024 |
-
|
1025 |
-
# Detect the same placeholders as image counterparts, to replace the first image slot with a video
|
1026 |
-
placeholder_patterns = [
|
1027 |
-
r'<img[^>]*src=["\'](?:placeholder|dummy|sample|example)[^"\']*["\'][^>]*>',
|
1028 |
-
r'<img[^>]*src=["\']https?://via\.placeholder\.com[^"\']*["\'][^>]*>',
|
1029 |
-
r'<img[^>]*src=["\']https?://picsum\.photos[^"\']*["\'][^>]*>',
|
1030 |
-
r'<img[^>]*src=["\']https?://dummyimage\.com[^"\']*["\'][^>]*>',
|
1031 |
-
r'<img[^>]*alt=["\'][^"\']*placeholder[^"\']*["\'][^>]*>',
|
1032 |
-
r'<img[^>]*class=["\'][^"\']*placeholder[^"\']*["\'][^>]*>',
|
1033 |
-
r'<img[^>]*id=["\'][^"\']*placeholder[^"\']*["\'][^>]*>',
|
1034 |
-
r'<img[^>]*src=["\']data:image[^"\']*["\'][^>]*>',
|
1035 |
-
r'<img[^>]*src=["\']#["\'][^>]*>',
|
1036 |
-
r'<img[^>]*src=["\']about:blank["\'][^>]*>',
|
1037 |
-
]
|
1038 |
-
|
1039 |
-
placeholder_images = []
|
1040 |
-
for pattern in placeholder_patterns:
|
1041 |
-
matches = re.findall(pattern, html_content, re.IGNORECASE)
|
1042 |
-
if matches:
|
1043 |
-
placeholder_images.extend(matches)
|
1044 |
-
|
1045 |
-
# Filter out HF URLs from placeholders (they are real generated content)
|
1046 |
-
placeholder_images = [img for img in placeholder_images if 'huggingface.co/datasets/' not in img]
|
1047 |
-
|
1048 |
-
if not placeholder_images:
|
1049 |
-
img_pattern = r'<img[^>]*>'
|
1050 |
-
placeholder_images = re.findall(img_pattern, html_content)
|
1051 |
-
|
1052 |
-
video_html = generate_video_from_text(prompt, session_id=session_id, token=None) # TODO: Pass token from parent context
|
1053 |
-
if video_html.startswith("Error"):
|
1054 |
-
return ""
|
1055 |
-
|
1056 |
-
# Replace first placeholder if present
|
1057 |
-
if placeholder_images:
|
1058 |
-
placeholder = placeholder_images[0]
|
1059 |
-
placeholder_clean = re.sub(r'\s+', ' ', placeholder.strip())
|
1060 |
-
placeholder_variations = [
|
1061 |
-
placeholder,
|
1062 |
-
placeholder_clean,
|
1063 |
-
placeholder_clean.replace('"', "'"),
|
1064 |
-
placeholder_clean.replace("'", '"'),
|
1065 |
-
re.sub(r'\s+', ' ', placeholder_clean),
|
1066 |
-
placeholder_clean.replace(' ', ' '),
|
1067 |
-
]
|
1068 |
-
blocks = []
|
1069 |
-
for variation in placeholder_variations:
|
1070 |
-
blocks.append(f"""{SEARCH_START}
|
1071 |
-
{variation}
|
1072 |
-
{DIVIDER}
|
1073 |
-
{video_html}
|
1074 |
-
{REPLACE_END}""")
|
1075 |
-
return '\n\n'.join(blocks)
|
1076 |
-
|
1077 |
-
# Otherwise insert after <body> with proper container
|
1078 |
-
if '<body' in html_content:
|
1079 |
-
body_start = html_content.find('<body')
|
1080 |
-
body_end = html_content.find('>', body_start) + 1
|
1081 |
-
opening_body_tag = html_content[body_start:body_end]
|
1082 |
-
|
1083 |
-
# Look for existing container elements to insert into
|
1084 |
-
body_content_start = body_end
|
1085 |
-
|
1086 |
-
# Try to find a good insertion point within existing content structure
|
1087 |
-
patterns_to_try = [
|
1088 |
-
r'<main[^>]*>',
|
1089 |
-
r'<section[^>]*class="[^"]*hero[^"]*"[^>]*>',
|
1090 |
-
r'<div[^>]*class="[^"]*container[^"]*"[^>]*>',
|
1091 |
-
r'<header[^>]*>',
|
1092 |
-
]
|
1093 |
-
|
1094 |
-
insertion_point = None
|
1095 |
-
for pattern in patterns_to_try:
|
1096 |
-
import re
|
1097 |
-
match = re.search(pattern, html_content[body_content_start:], re.IGNORECASE)
|
1098 |
-
if match:
|
1099 |
-
match_end = body_content_start + match.end()
|
1100 |
-
# Find the end of this tag
|
1101 |
-
tag_content = html_content[body_content_start + match.start():match_end]
|
1102 |
-
insertion_point = html_content[:match_end] + '\n '
|
1103 |
-
break
|
1104 |
-
|
1105 |
-
if not insertion_point:
|
1106 |
-
# Fallback to right after body tag with container div
|
1107 |
-
insertion_point = html_content[:body_end] + '\n '
|
1108 |
-
video_with_container = f'<div class="video-container" style="margin: 20px 0; text-align: center;">\n {video_html}\n </div>'
|
1109 |
-
return f"""{SEARCH_START}
|
1110 |
-
{insertion_point}
|
1111 |
-
{DIVIDER}
|
1112 |
-
{insertion_point}
|
1113 |
-
{video_with_container}
|
1114 |
-
{REPLACE_END}"""
|
1115 |
-
else:
|
1116 |
-
return f"""{SEARCH_START}
|
1117 |
-
{insertion_point}
|
1118 |
-
{DIVIDER}
|
1119 |
-
{insertion_point}
|
1120 |
-
{video_html}
|
1121 |
-
{REPLACE_END}"""
|
1122 |
-
|
1123 |
-
# If no <body>, just append
|
1124 |
-
return f"{SEARCH_START}\n\n{DIVIDER}\n{video_html}\n{REPLACE_END}"
|
1125 |
-
|
1126 |
-
def create_music_replacement_blocks_text_to_music(html_content: str, prompt: str, session_id: Optional[str] = None) -> str:
|
1127 |
-
"""Create search/replace blocks that insert ONE generated <audio> near the top of <body>."""
|
1128 |
-
if not prompt or not prompt.strip():
|
1129 |
-
return ""
|
1130 |
-
|
1131 |
-
audio_html = generate_music_from_text(prompt, session_id=session_id, token=None) # TODO: Pass token from parent context
|
1132 |
-
if audio_html.startswith("Error"):
|
1133 |
-
return ""
|
1134 |
-
|
1135 |
-
# Prefer inserting after the first <section>...</section> if present; else after <body>
|
1136 |
-
import re
|
1137 |
-
section_match = re.search(r"<section\b[\s\S]*?</section>", html_content, flags=re.IGNORECASE)
|
1138 |
-
if section_match:
|
1139 |
-
section_html = section_match.group(0)
|
1140 |
-
section_clean = re.sub(r"\s+", " ", section_html.strip())
|
1141 |
-
variations = [
|
1142 |
-
section_html,
|
1143 |
-
section_clean,
|
1144 |
-
section_clean.replace('"', "'"),
|
1145 |
-
section_clean.replace("'", '"'),
|
1146 |
-
re.sub(r"\s+", " ", section_clean),
|
1147 |
-
]
|
1148 |
-
blocks = []
|
1149 |
-
for v in variations:
|
1150 |
-
blocks.append(f"""{SEARCH_START}
|
1151 |
-
{v}
|
1152 |
-
{DIVIDER}
|
1153 |
-
{v}\n {audio_html}
|
1154 |
-
{REPLACE_END}""")
|
1155 |
-
return "\n\n".join(blocks)
|
1156 |
-
if '<body' in html_content:
|
1157 |
-
body_end = html_content.find('>', html_content.find('<body')) + 1
|
1158 |
-
insertion_point = html_content[:body_end] + '\n '
|
1159 |
-
return f"""{SEARCH_START}
|
1160 |
-
{insertion_point}
|
1161 |
-
{DIVIDER}
|
1162 |
-
{insertion_point}
|
1163 |
-
{audio_html}
|
1164 |
-
{REPLACE_END}"""
|
1165 |
-
|
1166 |
-
# If no <body>, just append
|
1167 |
-
return f"{SEARCH_START}\n\n{DIVIDER}\n{audio_html}\n{REPLACE_END}"
|
|
|
1 |
+
"""
|
2 |
+
Media generation functions for images, videos, and music using various AI models.
|
3 |
+
"""
|
4 |
+
|
5 |
import os
|
6 |
+
import io
|
7 |
import base64
|
|
|
|
|
|
|
|
|
8 |
import requests
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
import tempfile
|
10 |
+
from typing import Optional, Dict, Any
|
11 |
+
from PIL import Image
|
12 |
+
import numpy as np
|
|
|
|
|
13 |
import gradio as gr
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
|
15 |
+
from huggingface_hub import InferenceClient
|
16 |
+
from utils import create_temp_media_url, compress_media_for_data_uri, validate_video_html
|
17 |
+
from config import HF_TOKEN
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
18 |
|
19 |
+
class MediaGenerator:
|
20 |
+
"""Handles generation of images, videos, and music"""
|
21 |
+
|
22 |
+
def __init__(self):
|
23 |
+
self.hf_client = None
|
24 |
+
if HF_TOKEN:
|
25 |
+
self.hf_client = InferenceClient(
|
26 |
+
provider="auto",
|
27 |
+
api_key=HF_TOKEN,
|
28 |
+
bill_to="huggingface"
|
29 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
30 |
|
31 |
+
def generate_image_with_qwen(self, prompt: str, image_index: int = 0,
|
32 |
+
token: Optional[gr.OAuthToken] = None) -> str:
|
33 |
+
"""Generate image using Qwen image model"""
|
34 |
try:
|
35 |
+
if not self.hf_client:
|
36 |
+
return "Error: HF_TOKEN environment variable is not set."
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
|
38 |
+
print(f"[ImageGen] Generating image with prompt: {prompt}")
|
39 |
+
|
40 |
+
# Generate image using Qwen/Qwen-Image model
|
41 |
+
image = self.hf_client.text_to_image(
|
42 |
+
prompt,
|
43 |
+
model="Qwen/Qwen-Image",
|
44 |
+
)
|
45 |
+
|
46 |
+
# Resize image to reduce size while maintaining quality
|
47 |
+
max_size = 1024
|
48 |
+
if image.width > max_size or image.height > max_size:
|
49 |
+
image.thumbnail((max_size, max_size), Image.Resampling.LANCZOS)
|
50 |
+
|
51 |
+
# Convert to bytes
|
52 |
+
buffer = io.BytesIO()
|
53 |
+
image.convert('RGB').save(buffer, format='JPEG', quality=90, optimize=True)
|
54 |
+
image_bytes = buffer.getvalue()
|
55 |
+
|
56 |
+
# Create temporary URL
|
57 |
+
filename = f"generated_image_{image_index}.jpg"
|
58 |
+
temp_url = self._upload_media_to_hf(image_bytes, filename, "image", token, use_temp=True)
|
59 |
+
|
60 |
+
if temp_url.startswith("Error"):
|
61 |
+
return temp_url
|
62 |
+
|
63 |
+
return f'<img src="{temp_url}" alt="{prompt}" style="max-width: 100%; height: auto; border-radius: 8px; margin: 10px 0;" loading="lazy" />'
|
64 |
+
|
65 |
+
except Exception as e:
|
66 |
+
print(f"Image generation error: {str(e)}")
|
67 |
+
return f"Error generating image: {str(e)}"
|
68 |
+
|
69 |
+
def generate_image_to_image(self, input_image_data, prompt: str,
|
70 |
+
token: Optional[gr.OAuthToken] = None) -> str:
|
71 |
+
"""Generate image using image-to-image with Qwen-Image-Edit"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
72 |
try:
|
73 |
+
if not self.hf_client:
|
74 |
+
return "Error: HF_TOKEN environment variable is not set."
|
75 |
+
|
76 |
+
print(f"[Image2Image] Processing with prompt: {prompt}")
|
77 |
+
|
78 |
+
# Normalize input image to bytes
|
79 |
+
pil_image = self._process_input_image(input_image_data)
|
80 |
+
|
81 |
+
# Resize input image to avoid request body size limits
|
82 |
+
max_input_size = 1024
|
83 |
+
if pil_image.width > max_input_size or pil_image.height > max_input_size:
|
84 |
+
pil_image.thumbnail((max_input_size, max_input_size), Image.Resampling.LANCZOS)
|
85 |
+
|
86 |
+
# Convert to bytes
|
87 |
+
buf = io.BytesIO()
|
88 |
+
pil_image.save(buf, format='JPEG', quality=85, optimize=True)
|
89 |
+
input_bytes = buf.getvalue()
|
90 |
+
|
91 |
+
# Call image-to-image
|
92 |
+
image = self.hf_client.image_to_image(
|
93 |
+
input_bytes,
|
94 |
+
prompt=prompt,
|
95 |
+
model="Qwen/Qwen-Image-Edit",
|
96 |
+
)
|
97 |
+
|
98 |
+
# Resize and optimize output
|
99 |
+
max_size = 1024
|
100 |
+
if image.width > max_size or image.height > max_size:
|
101 |
+
image.thumbnail((max_size, max_size), Image.Resampling.LANCZOS)
|
102 |
+
|
103 |
+
out_buf = io.BytesIO()
|
104 |
+
image.convert('RGB').save(out_buf, format='JPEG', quality=90, optimize=True)
|
105 |
+
image_bytes = out_buf.getvalue()
|
106 |
+
|
107 |
+
# Create temporary URL
|
108 |
+
filename = "image_to_image_result.jpg"
|
109 |
+
temp_url = self._upload_media_to_hf(image_bytes, filename, "image", token, use_temp=True)
|
110 |
+
|
111 |
+
if temp_url.startswith("Error"):
|
112 |
+
return temp_url
|
113 |
+
|
114 |
+
return f'<img src="{temp_url}" alt="{prompt}" style="max-width: 100%; height: auto; border-radius: 8px; margin: 10px 0;" loading="lazy" />'
|
115 |
+
|
116 |
except Exception as e:
|
117 |
+
print(f"Image-to-image generation error: {str(e)}")
|
118 |
+
return f"Error generating image (image-to-image): {str(e)}"
|
119 |
+
|
120 |
+
def generate_video_from_image(self, input_image_data, prompt: str,
|
121 |
+
session_id: Optional[str] = None,
|
122 |
+
token: Optional[gr.OAuthToken] = None) -> str:
|
123 |
+
"""Generate video from input image using Lightricks LTX-Video"""
|
124 |
try:
|
125 |
+
print("[Image2Video] Starting video generation")
|
126 |
+
if not self.hf_client:
|
127 |
+
return "Error: HF_TOKEN environment variable is not set."
|
128 |
+
|
129 |
+
# Process input image
|
130 |
+
pil_image = self._process_input_image(input_image_data)
|
131 |
+
print(f"[Image2Video] Input image size: {pil_image.size}")
|
132 |
+
|
133 |
+
# Compress image for API limits
|
134 |
+
input_bytes = self._compress_image_for_video(pil_image, max_size_mb=3.9)
|
135 |
+
|
136 |
+
# Check for image-to-video method
|
137 |
+
image_to_video_method = getattr(self.hf_client, "image_to_video", None)
|
138 |
+
if not callable(image_to_video_method):
|
139 |
+
return ("Error: Your huggingface_hub version does not support image_to_video. "
|
140 |
+
"Please upgrade with `pip install -U huggingface_hub`")
|
141 |
+
|
142 |
+
model_id = "Lightricks/LTX-Video-0.9.8-13B-distilled"
|
143 |
+
print(f"[Image2Video] Calling API with model: {model_id}")
|
144 |
+
|
145 |
+
video_bytes = image_to_video_method(
|
146 |
+
input_bytes,
|
147 |
+
prompt=prompt,
|
148 |
+
model=model_id,
|
149 |
)
|
150 |
+
|
151 |
+
print(f"[Image2Video] Received video bytes: {len(video_bytes) if hasattr(video_bytes, '__len__') else 'unknown'}")
|
152 |
+
|
153 |
+
# Create temporary URL
|
154 |
+
filename = "image_to_video_result.mp4"
|
155 |
+
temp_url = self._upload_media_to_hf(video_bytes, filename, "video", token, use_temp=True)
|
156 |
+
|
157 |
+
if temp_url.startswith("Error"):
|
158 |
+
return temp_url
|
159 |
+
|
160 |
+
video_html = self._create_video_html(temp_url)
|
161 |
+
|
162 |
+
if not validate_video_html(video_html):
|
163 |
+
return "Error: Generated video HTML is malformed"
|
164 |
+
|
165 |
+
print(f"[Image2Video] Successfully generated video: {temp_url}")
|
166 |
+
return video_html
|
167 |
+
|
168 |
except Exception as e:
|
169 |
+
print(f"[Image2Video] Error: {str(e)}")
|
170 |
+
return f"Error generating video (image-to-video): {str(e)}"
|
171 |
+
|
172 |
+
def generate_video_from_text(self, prompt: str, session_id: Optional[str] = None,
|
173 |
+
token: Optional[gr.OAuthToken] = None) -> str:
|
174 |
+
"""Generate video from text prompt using Wan-AI text-to-video model"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
175 |
try:
|
176 |
+
print("[Text2Video] Starting video generation")
|
177 |
+
if not self.hf_client:
|
178 |
+
return "Error: HF_TOKEN environment variable is not set."
|
179 |
+
|
180 |
+
# Check for text-to-video method
|
181 |
+
text_to_video_method = getattr(self.hf_client, "text_to_video", None)
|
182 |
+
if not callable(text_to_video_method):
|
183 |
+
return ("Error: Your huggingface_hub version does not support text_to_video. "
|
184 |
+
"Please upgrade with `pip install -U huggingface_hub`")
|
185 |
+
|
186 |
+
model_id = "Wan-AI/Wan2.2-T2V-A14B"
|
187 |
+
prompt_str = (prompt or "").strip()
|
188 |
+
print(f"[Text2Video] Using model: {model_id}, prompt length: {len(prompt_str)}")
|
189 |
+
|
190 |
+
video_bytes = text_to_video_method(
|
191 |
+
prompt_str,
|
192 |
+
model=model_id,
|
193 |
)
|
194 |
|
195 |
+
print(f"[Text2Video] Received video bytes: {len(video_bytes) if hasattr(video_bytes, '__len__') else 'unknown'}")
|
|
|
|
|
|
|
196 |
|
197 |
+
# Create temporary URL
|
198 |
+
filename = "text_to_video_result.mp4"
|
199 |
+
temp_url = self._upload_media_to_hf(video_bytes, filename, "video", token, use_temp=True)
|
200 |
+
|
201 |
+
if temp_url.startswith("Error"):
|
202 |
+
return temp_url
|
203 |
+
|
204 |
+
video_html = self._create_video_html(temp_url)
|
205 |
+
|
206 |
+
if not validate_video_html(video_html):
|
207 |
+
return "Error: Generated video HTML is malformed"
|
208 |
+
|
209 |
+
print(f"[Text2Video] Successfully generated video: {temp_url}")
|
210 |
+
return video_html
|
211 |
+
|
212 |
+
except Exception as e:
|
213 |
+
print(f"[Text2Video] Error: {str(e)}")
|
214 |
+
return f"Error generating video (text-to-video): {str(e)}"
|
215 |
+
|
216 |
+
def generate_music_from_text(self, prompt: str, music_length_ms: int = 30000,
|
217 |
+
session_id: Optional[str] = None,
|
218 |
+
token: Optional[gr.OAuthToken] = None) -> str:
|
219 |
+
"""Generate music using ElevenLabs Music API"""
|
220 |
+
try:
|
221 |
+
api_key = os.getenv('ELEVENLABS_API_KEY')
|
222 |
+
if not api_key:
|
223 |
+
return "Error: ELEVENLABS_API_KEY environment variable is not set."
|
224 |
+
|
225 |
+
print(f"[MusicGen] Generating music: {prompt}")
|
226 |
+
|
227 |
+
headers = {
|
228 |
+
'Content-Type': 'application/json',
|
229 |
+
'xi-api-key': api_key,
|
230 |
+
}
|
231 |
+
|
232 |
+
payload = {
|
233 |
+
'prompt': prompt or 'Epic orchestral theme with soaring strings and powerful brass',
|
234 |
+
'music_length_ms': int(music_length_ms) if music_length_ms else 30000,
|
235 |
+
}
|
236 |
+
|
237 |
+
resp = requests.post(
|
238 |
+
'https://api.elevenlabs.io/v1/music/compose',
|
239 |
+
headers=headers,
|
240 |
+
json=payload,
|
241 |
+
timeout=60
|
242 |
+
)
|
243 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
244 |
try:
|
245 |
+
resp.raise_for_status()
|
|
|
|
|
246 |
except Exception as e:
|
247 |
+
error_text = getattr(e, 'response', resp).text if hasattr(e, 'response') else resp.text
|
248 |
+
return f"Error generating music: {error_text}"
|
249 |
+
|
250 |
+
# Create temporary URL
|
251 |
+
filename = "generated_music.mp3"
|
252 |
+
temp_url = self._upload_media_to_hf(resp.content, filename, "audio", token, use_temp=True)
|
253 |
+
|
254 |
+
if temp_url.startswith("Error"):
|
255 |
+
return temp_url
|
256 |
+
|
257 |
+
audio_html = self._create_audio_html(temp_url)
|
258 |
+
print(f"[MusicGen] Successfully generated music: {temp_url}")
|
259 |
+
return audio_html
|
260 |
+
|
261 |
+
except Exception as e:
|
262 |
+
print(f"[MusicGen] Error: {str(e)}")
|
263 |
+
return f"Error generating music: {str(e)}"
|
264 |
+
|
265 |
+
def _process_input_image(self, input_image_data) -> Image.Image:
|
266 |
+
"""Convert various image formats to PIL Image"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
267 |
if hasattr(input_image_data, 'read'):
|
|
|
268 |
raw = input_image_data.read()
|
269 |
pil_image = Image.open(io.BytesIO(raw))
|
270 |
elif hasattr(input_image_data, 'mode') and hasattr(input_image_data, 'size'):
|
|
|
271 |
pil_image = input_image_data
|
272 |
+
elif isinstance(input_image_data, np.ndarray):
|
273 |
pil_image = Image.fromarray(input_image_data)
|
274 |
elif isinstance(input_image_data, (bytes, bytearray)):
|
275 |
pil_image = Image.open(io.BytesIO(input_image_data))
|
276 |
else:
|
|
|
277 |
pil_image = Image.open(io.BytesIO(bytes(input_image_data)))
|
278 |
+
|
279 |
# Ensure RGB
|
280 |
if pil_image.mode != 'RGB':
|
281 |
pil_image = pil_image.convert('RGB')
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
282 |
|
283 |
+
return pil_image
|
284 |
+
|
285 |
+
def _compress_image_for_video(self, pil_image: Image.Image, max_size_mb: float = 3.9) -> bytes:
|
286 |
+
"""Compress image for video generation API limits"""
|
287 |
+
MAX_BYTES = int(max_size_mb * 1024 * 1024)
|
288 |
+
max_dim = 1024
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
289 |
quality = 90
|
290 |
+
|
291 |
def encode_current(pil: Image.Image, q: int) -> bytes:
|
292 |
tmp = io.BytesIO()
|
293 |
pil.save(tmp, format='JPEG', quality=q, optimize=True)
|
294 |
return tmp.getvalue()
|
295 |
+
|
296 |
+
# Downscale while too large
|
297 |
while max(pil_image.size) > max_dim:
|
298 |
ratio = max_dim / float(max(pil_image.size))
|
299 |
new_size = (max(1, int(pil_image.size[0] * ratio)), max(1, int(pil_image.size[1] * ratio)))
|
300 |
pil_image = pil_image.resize(new_size, Image.Resampling.LANCZOS)
|
301 |
+
|
302 |
encoded = encode_current(pil_image, quality)
|
303 |
+
|
304 |
+
# Reduce quality or dimensions if still too large
|
305 |
while len(encoded) > MAX_BYTES and (quality > 40 or max(pil_image.size) > 640):
|
306 |
if quality > 40:
|
307 |
quality -= 10
|
308 |
else:
|
|
|
309 |
new_w = max(1, int(pil_image.size[0] * 0.85))
|
310 |
new_h = max(1, int(pil_image.size[1] * 0.85))
|
311 |
pil_image = pil_image.resize((new_w, new_h), Image.Resampling.LANCZOS)
|
312 |
encoded = encode_current(pil_image, quality)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
313 |
|
314 |
+
return encoded
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
315 |
|
316 |
+
def _upload_media_to_hf(self, media_bytes: bytes, filename: str, media_type: str,
|
317 |
+
token: Optional[gr.OAuthToken] = None, use_temp: bool = True) -> str:
|
318 |
+
"""Upload media to HF or create temporary file"""
|
319 |
+
if use_temp:
|
320 |
+
return create_temp_media_url(media_bytes, filename, media_type)
|
321 |
+
|
322 |
+
# HF upload logic would go here for permanent URLs
|
323 |
+
# For now, always use temp files
|
324 |
+
return create_temp_media_url(media_bytes, filename, media_type)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
325 |
|
326 |
+
def _create_video_html(self, video_url: str) -> str:
|
327 |
+
"""Create HTML video element"""
|
328 |
+
return f'''<video controls autoplay muted loop playsinline
|
329 |
+
style="max-width: 100%; height: auto; border-radius: 8px; margin: 10px 0; display: block;"
|
330 |
+
onloadstart="this.style.backgroundColor='#f0f0f0'"
|
331 |
+
onerror="this.style.display='none'; console.error('Video failed to load')">
|
332 |
+
<source src="{video_url}" type="video/mp4" />
|
333 |
+
<p style="text-align: center; color: #666;">Your browser does not support the video tag.</p>
|
334 |
+
</video>'''
|
335 |
|
336 |
+
def _create_audio_html(self, audio_url: str) -> str:
|
337 |
+
"""Create HTML audio player"""
|
338 |
+
return f'''<div class="anycoder-music" style="max-width:420px;margin:16px auto;padding:12px 16px;border:1px solid #e5e7eb;border-radius:12px;background:linear-gradient(180deg,#fafafa,#f3f4f6);box-shadow:0 2px 8px rgba(0,0,0,0.06)">
|
339 |
+
<div style="font-size:13px;color:#374151;margin-bottom:8px;display:flex;align-items:center;gap:6px">
|
340 |
+
<span>🎵 Generated music</span>
|
341 |
+
</div>
|
342 |
+
<audio controls autoplay loop style="width:100%;outline:none;">
|
343 |
+
<source src="{audio_url}" type="audio/mpeg" />
|
344 |
+
Your browser does not support the audio element.
|
345 |
+
</audio>
|
346 |
+
</div>'''
|
347 |
+
|
348 |
+
# Global media generator instance
|
349 |
+
media_generator = MediaGenerator()
|
350 |
+
|
351 |
+
# Export main functions
|
352 |
+
def generate_image_with_qwen(prompt: str, image_index: int = 0, token: Optional[gr.OAuthToken] = None) -> str:
|
353 |
+
return media_generator.generate_image_with_qwen(prompt, image_index, token)
|
354 |
+
|
355 |
+
def generate_image_to_image(input_image_data, prompt: str, token: Optional[gr.OAuthToken] = None) -> str:
|
356 |
+
return media_generator.generate_image_to_image(input_image_data, prompt, token)
|
357 |
+
|
358 |
+
def generate_video_from_image(input_image_data, prompt: str, session_id: Optional[str] = None,
|
359 |
+
token: Optional[gr.OAuthToken] = None) -> str:
|
360 |
+
return media_generator.generate_video_from_image(input_image_data, prompt, session_id, token)
|
361 |
+
|
362 |
+
def generate_video_from_text(prompt: str, session_id: Optional[str] = None,
|
363 |
+
token: Optional[gr.OAuthToken] = None) -> str:
|
364 |
+
return media_generator.generate_video_from_text(prompt, session_id, token)
|
365 |
+
|
366 |
+
def generate_music_from_text(prompt: str, music_length_ms: int = 30000, session_id: Optional[str] = None,
|
367 |
+
token: Optional[gr.OAuthToken] = None) -> str:
|
368 |
+
return media_generator.generate_music_from_text(prompt, music_length_ms, session_id, token)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|