Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -12,9 +12,9 @@ import re
|
|
12 |
import requests
|
13 |
import io
|
14 |
import shutil
|
15 |
-
from urllib.parse import quote
|
16 |
import numpy as np
|
17 |
-
from bs4 import BeautifulSoup #
|
18 |
import base64
|
19 |
from gtts import gTTS
|
20 |
import gradio as gr
|
@@ -33,7 +33,6 @@ from pydub import AudioSegment
|
|
33 |
from pydub.generators import Sine
|
34 |
|
35 |
# ---------------- Global Configuration ---------------- #
|
36 |
-
# --- API Keys (Replace with your actual keys) ---
|
37 |
PEXELS_API_KEY = 'BhJqbcdm9Vi90KqzXKAhnEHGsuFNv4irXuOjWtT761U49lRzo03qBGna'
|
38 |
OPENROUTER_API_KEY = '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'
|
39 |
OPENROUTER_MODEL = "mistralai/mistral-small-3.1-24b-instruct:free"
|
@@ -71,8 +70,7 @@ except Exception as e:
|
|
71 |
|
72 |
def generate_script(user_input):
|
73 |
"""Generate documentary script using OpenRouter API."""
|
74 |
-
#
|
75 |
-
# (No changes needed here based on the request)
|
76 |
headers = {
|
77 |
'Authorization': f'Bearer {OPENROUTER_API_KEY}',
|
78 |
'Content-Type': 'application/json',
|
@@ -144,8 +142,7 @@ Now generate the script based on: {user_input}
|
|
144 |
|
145 |
def parse_script(script_text):
|
146 |
"""Parse the generated script into segments."""
|
147 |
-
#
|
148 |
-
# (No changes needed here based on the request)
|
149 |
segments = []
|
150 |
current_title = None
|
151 |
current_narration = ""
|
@@ -178,219 +175,414 @@ def parse_script(script_text):
|
|
178 |
print(f"Parsed {len(segments)} segments from script.")
|
179 |
return segments
|
180 |
|
181 |
-
# ---
|
182 |
-
|
183 |
-
|
184 |
-
|
185 |
-
|
|
|
186 |
return None
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
187 |
|
188 |
-
|
189 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
190 |
params = {"query": query, "per_page": 15, "orientation": "landscape"}
|
191 |
-
if search_type == "videos":
|
192 |
-
params["size"] = "medium" # Request medium or large
|
193 |
|
194 |
max_retries = 3
|
195 |
-
retry_delay =
|
196 |
-
timeout_duration =
|
197 |
|
198 |
-
print(f"Searching Pexels
|
199 |
|
200 |
for attempt in range(max_retries):
|
201 |
try:
|
202 |
-
response = requests.get(
|
203 |
-
|
204 |
-
|
205 |
-
|
206 |
-
|
207 |
-
|
208 |
-
|
209 |
-
|
210 |
-
|
211 |
-
|
212 |
-
|
213 |
-
|
214 |
-
|
215 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
216 |
time.sleep(retry_delay)
|
217 |
retry_delay *= 2
|
218 |
-
continue # Go to next attempt
|
219 |
-
|
220 |
-
response.raise_for_status() # Raise exceptions for other 4xx/5xx errors
|
221 |
-
|
222 |
-
data = response.json()
|
223 |
-
items = data.get(search_type, [])
|
224 |
-
|
225 |
-
if not items:
|
226 |
-
print(f"No Pexels {search_type} found for query: {query}")
|
227 |
-
return None # No results found
|
228 |
-
|
229 |
-
# --- Filtering logic remains the same ---
|
230 |
-
valid_items = []
|
231 |
-
if search_type == "videos":
|
232 |
-
for video in items:
|
233 |
-
hd_link = next((f['link'] for f in video.get('video_files', []) if f.get('quality') == 'hd' and f.get('width', 0) >= 1080), None)
|
234 |
-
large_link = next((f['link'] for f in video.get('video_files', []) if f.get('quality') == 'large' and f.get('width', 0) >= 1080), None)
|
235 |
-
medium_link = next((f['link'] for f in video.get('video_files', []) if f.get('quality') == 'medium'), None)
|
236 |
-
link = hd_link or large_link or medium_link
|
237 |
-
if link: valid_items.append(link)
|
238 |
-
else: # images
|
239 |
-
for photo in items:
|
240 |
-
link = photo.get("src", {}).get("large2x") or photo.get("src", {}).get("original")
|
241 |
-
if link: valid_items.append(link)
|
242 |
-
|
243 |
-
if valid_items:
|
244 |
-
print(f"Found {len(valid_items)} suitable Pexels {search_type} for '{query}'. Selecting one.")
|
245 |
-
return random.choice(valid_items)
|
246 |
else:
|
247 |
-
print(f"
|
248 |
-
|
|
|
|
|
|
|
|
|
|
|
249 |
|
250 |
except requests.exceptions.Timeout:
|
251 |
-
print(f"
|
252 |
time.sleep(retry_delay)
|
253 |
retry_delay *= 2
|
254 |
except requests.exceptions.RequestException as e:
|
255 |
-
print(f"
|
256 |
-
|
257 |
-
|
258 |
-
|
259 |
-
|
260 |
-
|
261 |
-
|
262 |
|
263 |
-
print(f"Pexels
|
264 |
return None
|
265 |
|
266 |
-
#
|
267 |
-
|
268 |
-
|
269 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
270 |
|
271 |
-
def
|
272 |
-
"""Download
|
273 |
-
# --- Retain previous download_media function ---
|
274 |
-
# (No changes needed here based on the request, it handles Pexels URLs)
|
275 |
try:
|
276 |
headers = {"User-Agent": USER_AGENT}
|
277 |
-
|
|
|
278 |
response.raise_for_status()
|
279 |
-
|
280 |
-
|
281 |
-
|
282 |
-
|
283 |
-
elif 'image/png' in content_type: file_extension = ".png"
|
284 |
-
elif 'image/webp' in content_type: file_extension = ".webp"
|
285 |
-
else: # Guess from URL
|
286 |
-
media_url_lower = media_url.lower()
|
287 |
-
if '.mp4' in media_url_lower: file_extension = ".mp4"
|
288 |
-
elif '.mov' in media_url_lower: file_extension = ".mov"
|
289 |
-
elif '.jpg' in media_url_lower or '.jpeg' in media_url_lower: file_extension = ".jpg"
|
290 |
-
elif '.png' in media_url_lower: file_extension = ".png"
|
291 |
-
elif '.webp' in media_url_lower: file_extension = ".webp"
|
292 |
-
|
293 |
-
filename = os.path.join(target_folder, f"{filename_prefix}{file_extension}")
|
294 |
with open(filename, 'wb') as f:
|
295 |
-
for chunk in response.iter_content(chunk_size=8192):
|
296 |
-
|
297 |
|
298 |
-
|
299 |
-
|
300 |
-
|
301 |
-
|
302 |
-
|
303 |
-
|
304 |
-
|
305 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
306 |
rgb_img = img.convert('RGB')
|
307 |
-
|
308 |
-
|
|
|
309 |
rgb_img.close()
|
310 |
img.close()
|
311 |
-
|
312 |
-
|
313 |
-
|
314 |
-
|
315 |
-
|
316 |
-
|
317 |
-
|
318 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
319 |
except requests.exceptions.RequestException as e_download:
|
320 |
-
print(f"
|
|
|
321 |
return None
|
322 |
-
except Exception as
|
323 |
-
print(f"General error during
|
324 |
return None
|
325 |
|
326 |
-
|
327 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
328 |
"""
|
329 |
-
|
330 |
-
|
331 |
-
Uses a single generic Pexels image search as the final fallback.
|
332 |
"""
|
333 |
-
|
334 |
-
|
335 |
-
|
336 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
337 |
|
338 |
-
#
|
339 |
if random.random() < video_clip_probability:
|
340 |
-
|
|
|
|
|
341 |
if video_url:
|
342 |
-
|
343 |
-
if
|
344 |
-
print(f"Using Pexels video
|
345 |
-
return {"path":
|
|
|
|
|
346 |
else:
|
347 |
-
print(f"Pexels video search
|
|
|
|
|
348 |
|
349 |
-
|
350 |
-
|
|
|
|
|
|
|
351 |
if image_url:
|
352 |
-
|
353 |
-
if
|
354 |
-
print(f"Using Pexels image
|
355 |
-
return {"path":
|
|
|
|
|
356 |
else:
|
357 |
-
print(f"Pexels image search
|
358 |
-
|
359 |
-
# ---
|
360 |
-
#
|
361 |
-
# google_image_url = search_google_images(prompt)
|
362 |
-
# ... (removed logic) ...
|
363 |
-
|
364 |
-
# 3. Absolute Fallback: Generic Pexels Image Search
|
365 |
-
# Only use this if the specific prompt searches failed.
|
366 |
-
# Avoid searching for problematic terms like "Subscribe CTA".
|
367 |
-
fallback_terms = ["technology", "abstract", "nature", "background"]
|
368 |
-
# Don't use fallback for prompts that are clearly instructions/CTAs
|
369 |
if "subscribe" not in prompt.lower() and "cta" not in prompt.lower():
|
370 |
-
|
371 |
-
|
372 |
-
|
|
|
|
|
373 |
if fallback_url:
|
374 |
-
|
375 |
-
if
|
376 |
-
print(f"Using Pexels fallback image
|
377 |
-
return {"path":
|
378 |
else:
|
379 |
-
|
380 |
else:
|
381 |
-
print(f"Pexels fallback image search failed for term '{
|
382 |
else:
|
383 |
-
|
|
|
384 |
|
|
|
|
|
|
|
385 |
|
386 |
-
|
387 |
-
print(f"FATAL: Could not retrieve any suitable media for prompt: '{prompt}' after all attempts.")
|
388 |
-
return None # Indicate failure
|
389 |
|
390 |
|
391 |
def generate_tts(text, voice_id, speed):
|
392 |
"""Generate TTS audio using Kokoro, falling back to gTTS."""
|
393 |
-
#
|
394 |
safe_text_prefix = re.sub(r'[^\w\s-]', '', text[:20]).strip().replace(' ', '_')
|
395 |
output_filename = os.path.join(TEMP_FOLDER, f"tts_{safe_text_prefix}_{voice_id}.wav")
|
396 |
if pipeline:
|
@@ -421,10 +613,17 @@ def generate_tts(text, voice_id, speed):
|
|
421 |
wav_path = output_filename
|
422 |
tts.save(mp3_path)
|
423 |
audio = AudioSegment.from_mp3(mp3_path)
|
|
|
|
|
424 |
audio.export(wav_path, format="wav")
|
425 |
os.remove(mp3_path)
|
426 |
print(f"gTTS audio saved and converted to {wav_path}")
|
427 |
-
|
|
|
|
|
|
|
|
|
|
|
428 |
except ImportError:
|
429 |
print("Error: gTTS or pydub might not be installed. Cannot use gTTS fallback.")
|
430 |
return None
|
@@ -434,27 +633,57 @@ def generate_tts(text, voice_id, speed):
|
|
434 |
|
435 |
def apply_kenburns_effect(clip, target_resolution, duration):
|
436 |
"""Apply a randomized Ken Burns effect (zoom/pan) to an image clip."""
|
437 |
-
#
|
438 |
target_w, target_h = target_resolution
|
439 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
440 |
scale_factor = 1.2
|
441 |
-
|
442 |
-
if
|
|
|
|
|
|
|
|
|
|
|
443 |
final_h = target_h * scale_factor
|
444 |
-
final_w = final_h *
|
445 |
else:
|
446 |
final_w = target_w * scale_factor
|
447 |
-
final_h = final_w
|
|
|
448 |
final_w, final_h = int(final_w), int(final_h)
|
|
|
|
|
|
|
|
|
449 |
try:
|
450 |
-
|
451 |
-
resized_pil = pil_img.resize((final_w, final_h), Image.Resampling.LANCZOS)
|
452 |
-
resized_clip = ImageClip(np.array(resized_pil)).set_duration(duration)
|
453 |
-
except Exception as e:
|
454 |
-
print(f"Warning: Error during high-quality resize for Ken Burns, using MoviePy default: {e}")
|
455 |
resized_clip = clip.resize(newsize=(final_w, final_h)).set_duration(duration)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
456 |
max_move_x = final_w - target_w
|
457 |
max_move_y = final_h - target_h
|
|
|
|
|
|
|
|
|
458 |
effect = random.choice(['zoom_in', 'zoom_out', 'pan_lr', 'pan_rl', 'pan_td', 'pan_dt'])
|
459 |
if effect == 'zoom_in': zoom_start, zoom_end = 1.0, scale_factor; x_start, x_end = max_move_x / 2, max_move_x / 2; y_start, y_end = max_move_y / 2, max_move_y / 2
|
460 |
elif effect == 'zoom_out': zoom_start, zoom_end = scale_factor, 1.0; x_start, x_end = max_move_x / 2, max_move_x / 2; y_start, y_end = max_move_y / 2, max_move_y / 2
|
@@ -462,54 +691,170 @@ def apply_kenburns_effect(clip, target_resolution, duration):
|
|
462 |
elif effect == 'pan_rl': zoom_start, zoom_end = scale_factor, scale_factor; x_start, x_end = max_move_x, 0; y_start, y_end = max_move_y / 2, max_move_y / 2
|
463 |
elif effect == 'pan_td': zoom_start, zoom_end = scale_factor, scale_factor; x_start, x_end = max_move_x / 2, max_move_x / 2; y_start, y_end = 0, max_move_y
|
464 |
else: zoom_start, zoom_end = scale_factor, scale_factor; x_start, x_end = max_move_x / 2, max_move_x / 2; y_start, y_end = max_move_y, 0
|
|
|
465 |
def make_frame(t):
|
466 |
-
|
|
|
|
|
467 |
current_zoom = zoom_start + (zoom_end - zoom_start) * interp
|
468 |
current_x = x_start + (x_end - x_start) * interp
|
469 |
current_y = y_start + (y_end - y_start) * interp
|
470 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
471 |
crop_w = max(1, int(crop_w)); crop_h = max(1, int(crop_h))
|
|
|
|
|
472 |
x1 = current_x; y1 = current_y
|
473 |
x1 = max(0, min(x1, final_w - crop_w)); y1 = max(0, min(y1, final_h - crop_h))
|
|
|
|
|
474 |
frame = resized_clip.get_frame(t)
|
475 |
-
|
476 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
477 |
return final_frame
|
478 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
479 |
|
480 |
def resize_to_fill(clip, target_resolution):
|
481 |
"""Resize and crop a video clip to fill the target resolution."""
|
482 |
-
#
|
483 |
target_w, target_h = target_resolution
|
484 |
-
|
485 |
-
|
486 |
-
|
487 |
-
|
488 |
-
|
489 |
-
|
490 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
491 |
|
492 |
def add_background_music(video_clip, music_file_path, volume):
|
493 |
"""Add background music, looping if necessary."""
|
494 |
-
#
|
495 |
if not music_file_path or not os.path.exists(music_file_path):
|
496 |
print("No background music file found or provided. Skipping.")
|
497 |
return video_clip
|
498 |
try:
|
499 |
print(f"Adding background music from: {music_file_path}")
|
500 |
bg_music = AudioFileClip(music_file_path)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
501 |
if bg_music.duration > video_clip.duration: bg_music = bg_music.subclip(0, video_clip.duration)
|
502 |
elif bg_music.duration < video_clip.duration:
|
503 |
loops_needed = math.ceil(video_clip.duration / bg_music.duration)
|
504 |
-
|
505 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
506 |
bg_music = bg_music.volumex(volume)
|
507 |
-
|
508 |
if video_clip.audio:
|
509 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
510 |
else:
|
511 |
-
# If original clip has no audio
|
512 |
-
|
513 |
video_clip = video_clip.set_audio(final_audio)
|
514 |
print("Background music added successfully.")
|
515 |
return video_clip
|
@@ -519,39 +864,65 @@ def add_background_music(video_clip, music_file_path, volume):
|
|
519 |
|
520 |
def create_segment_clip(media_info, tts_path, narration_text):
|
521 |
"""Create a single video segment (clip) with visuals, audio, and subtitles."""
|
522 |
-
#
|
523 |
try:
|
524 |
media_path = media_info['path']
|
525 |
asset_type = media_info['asset_type']
|
526 |
print(f"Creating clip segment: Type={asset_type}, Media={os.path.basename(media_path)}")
|
527 |
-
if not os.path.exists(tts_path)
|
|
|
528 |
audio_clip = AudioFileClip(tts_path)
|
|
|
|
|
529 |
segment_duration = audio_clip.duration + 0.3
|
|
|
530 |
if asset_type == "video":
|
531 |
-
if not os.path.exists(media_path)
|
532 |
-
|
533 |
-
|
534 |
-
|
535 |
-
|
536 |
-
|
537 |
-
|
538 |
-
|
539 |
-
|
540 |
-
|
541 |
-
|
542 |
-
|
543 |
-
|
544 |
-
|
545 |
-
|
546 |
-
|
547 |
-
|
548 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
549 |
elif asset_type == "image":
|
550 |
-
if not os.path.exists(media_path)
|
551 |
-
|
552 |
-
|
553 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
554 |
else: print(f"Error: Unknown asset type: {asset_type}"); return None
|
|
|
555 |
visual_clip = visual_clip.fadein(0.15).fadeout(0.15)
|
556 |
subtitle_clips = []
|
557 |
if USE_CAPTIONS and narration_text:
|
@@ -566,35 +937,54 @@ def create_segment_clip(media_info, tts_path, narration_text):
|
|
566 |
num_chunks = len(chunks); chunk_duration = audio_clip.duration / num_chunks
|
567 |
start_time = 0.1
|
568 |
for i, chunk_text in enumerate(chunks):
|
569 |
-
|
|
|
|
|
|
|
570 |
try:
|
571 |
txt_clip = TextClip(txt=chunk_text, fontsize=font_size, font=caption_font, color=caption_style_text_color,
|
572 |
bg_color=caption_style_bg_color, method='label', align='center',
|
573 |
size=(TARGET_RESOLUTION[0] * 0.8, None))
|
574 |
txt_clip = txt_clip.set_position(('center', TARGET_RESOLUTION[1] * 0.80))
|
575 |
-
txt_clip = txt_clip.set_start(start_time).set_duration(
|
576 |
subtitle_clips.append(txt_clip)
|
577 |
-
start_time =
|
578 |
except Exception as txt_err:
|
579 |
print(f"ERROR creating TextClip for '{chunk_text}': {txt_err}. Skipping subtitle chunk.")
|
580 |
-
# If one subtitle fails, continue without it
|
581 |
|
582 |
-
final_clip = CompositeVideoClip([visual_clip] + subtitle_clips) if subtitle_clips else visual_clip
|
583 |
-
|
|
|
|
|
|
|
|
|
584 |
print(f"Clip segment created successfully. Duration: {final_clip.duration:.2f}s")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
585 |
return final_clip
|
586 |
except Exception as e:
|
587 |
print(f"Error creating clip segment: {e}")
|
588 |
import traceback
|
589 |
traceback.print_exc()
|
|
|
|
|
|
|
|
|
590 |
return None
|
591 |
|
|
|
592 |
# ---------------- Main Video Generation Function ---------------- #
|
593 |
|
594 |
def generate_full_video(user_input, resolution_choice, caption_choice, music_file_info):
|
595 |
"""Main function orchestrating the video generation process."""
|
596 |
-
#
|
597 |
-
# (Ensure it handles None from generate_media correctly)
|
598 |
global TARGET_RESOLUTION, TEMP_FOLDER, USE_CAPTIONS
|
599 |
print("\n--- Starting Video Generation ---"); start_time = time.time()
|
600 |
if resolution_choice == "Short (9:16)": TARGET_RESOLUTION = (1080, 1920); print("Resolution set to: Short (1080x1920)")
|
@@ -605,6 +995,8 @@ def generate_full_video(user_input, resolution_choice, caption_choice, music_fil
|
|
605 |
if music_file_info is not None:
|
606 |
try:
|
607 |
music_file_path = os.path.join(TEMP_FOLDER, "background_music.mp3")
|
|
|
|
|
608 |
shutil.copy(music_file_info.name, music_file_path)
|
609 |
print(f"Background music copied to: {music_file_path}")
|
610 |
except Exception as e: print(f"Error handling uploaded music file: {e}"); music_file_path = None
|
@@ -624,57 +1016,84 @@ def generate_full_video(user_input, resolution_choice, caption_choice, music_fil
|
|
624 |
print(f" Prompt: {segment['prompt']}")
|
625 |
print(f" Narration: {segment['narration']}")
|
626 |
|
627 |
-
media_info = generate_media(segment['prompt'])
|
628 |
-
# --- Crucial Check ---
|
629 |
if not media_info:
|
630 |
print(f"Warning: Failed to get media for segment {i+1} ('{segment['prompt']}'). Skipping this segment.")
|
631 |
-
continue
|
632 |
|
633 |
tts_path = generate_tts(segment['narration'], selected_voice, voice_speed)
|
634 |
if not tts_path:
|
635 |
print(f"Warning: Failed to generate TTS for segment {i+1}. Skipping segment.")
|
636 |
if media_info and os.path.exists(media_info['path']):
|
637 |
try: os.remove(media_info['path']); print(f"Cleaned up unused media: {media_info['path']}")
|
638 |
-
except OSError:
|
639 |
continue
|
640 |
|
|
|
641 |
clip = create_segment_clip(media_info, tts_path, segment['narration'])
|
642 |
if clip:
|
643 |
segment_clips.append(clip)
|
644 |
else:
|
645 |
print(f"Warning: Failed to create video clip for segment {i+1}. Skipping.")
|
|
|
646 |
if media_info and os.path.exists(media_info['path']):
|
647 |
try: os.remove(media_info['path']); print(f"Cleaned up media for failed clip: {media_info['path']}")
|
648 |
-
except OSError:
|
649 |
if tts_path and os.path.exists(tts_path):
|
650 |
try: os.remove(tts_path); print(f"Cleaned up TTS for failed clip: {tts_path}")
|
651 |
-
except OSError:
|
|
|
652 |
|
653 |
if not segment_clips:
|
654 |
print("ERROR: No video clips were successfully created. Aborting.")
|
655 |
-
shutil.rmtree(TEMP_FOLDER)
|
|
|
656 |
return None, "Error: Failed to create any video segments. Check logs for media/TTS issues."
|
657 |
|
658 |
print("\nStep 4: Concatenating video segments...");
|
|
|
659 |
try:
|
660 |
-
# Filter out potential None values just in case, although the loop should prevent them
|
661 |
valid_clips = [c for c in segment_clips if c is not None]
|
662 |
-
if not valid_clips:
|
663 |
-
|
664 |
final_video = concatenate_videoclips(valid_clips, method="compose")
|
665 |
print("Segments concatenated successfully.")
|
|
|
|
|
|
|
|
|
666 |
except Exception as e:
|
667 |
-
print(f"ERROR: Failed to concatenate video clips: {e}");
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
668 |
|
669 |
-
print("\nStep 5: Adding background music..."); final_video = add_background_music(final_video, music_file_path, bg_music_volume)
|
670 |
|
671 |
print(f"\nStep 6: Exporting final video to '{OUTPUT_VIDEO_FILENAME}'..."); export_success = False
|
672 |
-
|
673 |
-
|
674 |
-
|
675 |
-
|
676 |
-
|
677 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
678 |
|
679 |
print("\nStep 7: Cleaning up temporary files...");
|
680 |
try: shutil.rmtree(TEMP_FOLDER); print(f"Temporary folder {TEMP_FOLDER} removed.")
|
@@ -683,11 +1102,12 @@ def generate_full_video(user_input, resolution_choice, caption_choice, music_fil
|
|
683 |
end_time = time.time(); total_time = end_time - start_time
|
684 |
print(f"\n--- Video Generation Finished ---"); print(f"Total time: {total_time:.2f} seconds")
|
685 |
if export_success: return OUTPUT_VIDEO_FILENAME, f"Video generation complete! Time: {total_time:.2f}s"
|
|
|
686 |
else: return None, f"Error: Video export failed. Check logs. Time: {total_time:.2f}s"
|
687 |
|
688 |
|
689 |
# ---------------- Gradio Interface Definition ---------------- #
|
690 |
-
#
|
691 |
VOICE_CHOICES = {
|
692 |
'Emma (US Female)': 'af_heart', 'Bella (US Female)': 'af_bella', 'Nicole (US Female)': 'af_nicole',
|
693 |
'Sarah (US Female)': 'af_sarah', 'Michael (US Male)': 'am_michael', 'Eric (US Male)': 'am_eric',
|
@@ -696,8 +1116,18 @@ VOICE_CHOICES = {
|
|
696 |
}
|
697 |
def gradio_interface_handler(user_prompt, resolution, captions, bg_music, voice_name, video_prob, music_vol, video_fps, export_preset, tts_speed, caption_size):
|
698 |
print("\n--- Received Request from Gradio ---")
|
699 |
-
print(f"Prompt: {user_prompt[:50]}...")
|
700 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
701 |
global selected_voice, voice_speed, font_size, video_clip_probability, bg_music_volume, fps, preset
|
702 |
selected_voice = VOICE_CHOICES.get(voice_name, 'af_heart')
|
703 |
voice_speed = tts_speed; font_size = caption_size; video_clip_probability = video_prob / 100.0
|
@@ -745,4 +1175,4 @@ if __name__ == "__main__":
|
|
745 |
print("!!! Please replace 'YOUR_PEXELS_API_KEY' and !!!")
|
746 |
print("!!! 'YOUR_OPENROUTER_API_KEY' with your actual keys. !!!")
|
747 |
print("!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!\n")
|
748 |
-
iface.launch(share=True, debug=True)
|
|
|
12 |
import requests
|
13 |
import io
|
14 |
import shutil
|
15 |
+
from urllib.parse import quote # Needed for search_google_images
|
16 |
import numpy as np
|
17 |
+
from bs4 import BeautifulSoup # Needed for search_google_images
|
18 |
import base64
|
19 |
from gtts import gTTS
|
20 |
import gradio as gr
|
|
|
33 |
from pydub.generators import Sine
|
34 |
|
35 |
# ---------------- Global Configuration ---------------- #
|
|
|
36 |
PEXELS_API_KEY = 'BhJqbcdm9Vi90KqzXKAhnEHGsuFNv4irXuOjWtT761U49lRzo03qBGna'
|
37 |
OPENROUTER_API_KEY = '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'
|
38 |
OPENROUTER_MODEL = "mistralai/mistral-small-3.1-24b-instruct:free"
|
|
|
70 |
|
71 |
def generate_script(user_input):
|
72 |
"""Generate documentary script using OpenRouter API."""
|
73 |
+
# (Retained from previous versions)
|
|
|
74 |
headers = {
|
75 |
'Authorization': f'Bearer {OPENROUTER_API_KEY}',
|
76 |
'Content-Type': 'application/json',
|
|
|
142 |
|
143 |
def parse_script(script_text):
|
144 |
"""Parse the generated script into segments."""
|
145 |
+
# (Retained from previous versions)
|
|
|
146 |
segments = []
|
147 |
current_title = None
|
148 |
current_narration = ""
|
|
|
175 |
print(f"Parsed {len(segments)} segments from script.")
|
176 |
return segments
|
177 |
|
178 |
+
# --- Start: User Provided Functions ---
|
179 |
+
|
180 |
+
def search_pexels_videos(query, pexels_api_key):
|
181 |
+
"""Search for a video on Pexels by query and return a random HD video."""
|
182 |
+
if not pexels_api_key or pexels_api_key == 'YOUR_PEXELS_API_KEY':
|
183 |
+
print(f"Pexels API key not provided or is default. Skipping Pexels video search.")
|
184 |
return None
|
185 |
+
headers = {'Authorization': pexels_api_key}
|
186 |
+
base_url = "https://api.pexels.com/videos/search"
|
187 |
+
num_pages = 3
|
188 |
+
videos_per_page = 15
|
189 |
+
|
190 |
+
max_retries = 3
|
191 |
+
retry_delay = 1
|
192 |
+
timeout_duration = 15 # Increased timeout slightly
|
193 |
+
|
194 |
+
search_query = query
|
195 |
+
all_videos = []
|
196 |
+
print(f"Searching Pexels videos for '{query}' (up to {num_pages} pages)...")
|
197 |
+
|
198 |
+
for page in range(1, num_pages + 1):
|
199 |
+
print(f" Pexels Video Search: Page {page}")
|
200 |
+
for attempt in range(max_retries):
|
201 |
+
try:
|
202 |
+
params = {"query": search_query, "per_page": videos_per_page, "page": page, "orientation": "landscape"}
|
203 |
+
response = requests.get(base_url, headers=headers, params=params, timeout=timeout_duration)
|
204 |
+
|
205 |
+
if response.status_code == 200:
|
206 |
+
data = response.json()
|
207 |
+
videos = data.get("videos", [])
|
208 |
+
|
209 |
+
if not videos:
|
210 |
+
print(f" No videos found on page {page} for '{query}'.")
|
211 |
+
# Don't break inner loop immediately, maybe next page has results
|
212 |
+
# Break outer loop if no videos found on *this* page attempt
|
213 |
+
break # Break attempt loop for this page
|
214 |
+
|
215 |
+
found_on_page = 0
|
216 |
+
for video in videos:
|
217 |
+
video_files = video.get("video_files", [])
|
218 |
+
# Prioritize HD, then large, then medium
|
219 |
+
hd_link = next((f['link'] for f in video_files if f.get('quality') == 'hd' and f.get('width', 0) >= 1080), None)
|
220 |
+
large_link = next((f['link'] for f in video_files if f.get('quality') == 'large' and f.get('width', 0) >= 1080), None)
|
221 |
+
medium_link = next((f['link'] for f in video_files if f.get('quality') == 'medium'), None) # Fallback if no HD/Large
|
222 |
+
link_to_add = hd_link or large_link or medium_link
|
223 |
+
|
224 |
+
if link_to_add:
|
225 |
+
all_videos.append(link_to_add)
|
226 |
+
found_on_page += 1
|
227 |
+
# Don't break inner loop, collect all suitable videos from the page
|
228 |
|
229 |
+
print(f" Found {found_on_page} suitable videos on page {page}.")
|
230 |
+
break # Break attempt loop successfully after processing page
|
231 |
+
|
232 |
+
elif response.status_code == 401:
|
233 |
+
print(f" Pexels API Error: Unauthorized (401). Check your API Key.")
|
234 |
+
return None # Stop searching if key is bad
|
235 |
+
elif response.status_code == 429:
|
236 |
+
print(f" Rate limit hit (attempt {attempt+1}/{max_retries}). Retrying in {retry_delay} seconds...")
|
237 |
+
time.sleep(retry_delay)
|
238 |
+
retry_delay *= 2
|
239 |
+
elif response.status_code == 522:
|
240 |
+
print(f" Pexels API Error: Connection Timed Out (522) (attempt {attempt+1}/{max_retries}). Retrying in {retry_delay} seconds...")
|
241 |
+
time.sleep(retry_delay)
|
242 |
+
retry_delay *= 2
|
243 |
+
else:
|
244 |
+
print(f" Error fetching videos: {response.status_code} {response.text}")
|
245 |
+
if attempt < max_retries - 1:
|
246 |
+
print(f" Retrying in {retry_delay} seconds...")
|
247 |
+
time.sleep(retry_delay)
|
248 |
+
retry_delay *= 2
|
249 |
+
else:
|
250 |
+
print(f" Max retries reached for page {page}.")
|
251 |
+
break # Break attempt loop for this page after max retries
|
252 |
+
|
253 |
+
except requests.exceptions.Timeout:
|
254 |
+
print(f" Request timed out (attempt {attempt+1}/{max_retries}). Retrying in {retry_delay} seconds...")
|
255 |
+
time.sleep(retry_delay)
|
256 |
+
retry_delay *= 2
|
257 |
+
except requests.exceptions.RequestException as e:
|
258 |
+
print(f" Request exception: {e}")
|
259 |
+
if attempt < max_retries - 1:
|
260 |
+
print(f" Retrying in {retry_delay} seconds...")
|
261 |
+
time.sleep(retry_delay)
|
262 |
+
retry_delay *= 2
|
263 |
+
else:
|
264 |
+
print(f" Max retries reached for page {page} due to request exception.")
|
265 |
+
break # Break attempt loop for this page
|
266 |
+
|
267 |
+
# Reset retry delay for the next page
|
268 |
+
retry_delay = 1
|
269 |
+
|
270 |
+
if all_videos:
|
271 |
+
random_video = random.choice(all_videos)
|
272 |
+
print(f"Selected random video from {len(all_videos)} suitable videos found across pages.")
|
273 |
+
return random_video
|
274 |
+
else:
|
275 |
+
print(f"No suitable videos found for '{query}' after searching {num_pages} pages.")
|
276 |
+
return None
|
277 |
+
|
278 |
+
def search_pexels_images(query, pexels_api_key):
|
279 |
+
"""Search for an image on Pexels by query."""
|
280 |
+
if not pexels_api_key or pexels_api_key == 'YOUR_PEXELS_API_KEY':
|
281 |
+
print(f"Pexels API key not provided or is default. Skipping Pexels image search.")
|
282 |
+
return None
|
283 |
+
headers = {'Authorization': pexels_api_key}
|
284 |
+
url = "https://api.pexels.com/v1/search"
|
285 |
+
# Fetch more results to increase chance of finding good ones
|
286 |
params = {"query": query, "per_page": 15, "orientation": "landscape"}
|
|
|
|
|
287 |
|
288 |
max_retries = 3
|
289 |
+
retry_delay = 1
|
290 |
+
timeout_duration = 15 # Increased timeout slightly
|
291 |
|
292 |
+
print(f"Searching Pexels images for '{query}'...")
|
293 |
|
294 |
for attempt in range(max_retries):
|
295 |
try:
|
296 |
+
response = requests.get(url, headers=headers, params=params, timeout=timeout_duration)
|
297 |
+
|
298 |
+
if response.status_code == 200:
|
299 |
+
data = response.json()
|
300 |
+
photos = data.get("photos", [])
|
301 |
+
if photos:
|
302 |
+
# Select from all returned photos, preferring larger sizes
|
303 |
+
valid_photos = []
|
304 |
+
for photo in photos:
|
305 |
+
# Prefer large2x or original, fallback to large
|
306 |
+
large2x_url = photo.get("src", {}).get("large2x")
|
307 |
+
original_url = photo.get("src", {}).get("original")
|
308 |
+
large_url = photo.get("src", {}).get("large")
|
309 |
+
img_url = large2x_url or original_url or large_url
|
310 |
+
if img_url:
|
311 |
+
valid_photos.append(img_url)
|
312 |
+
|
313 |
+
if valid_photos:
|
314 |
+
selected_photo = random.choice(valid_photos)
|
315 |
+
print(f"Selected random image from {len(valid_photos)} suitable images found.")
|
316 |
+
return selected_photo
|
317 |
+
else:
|
318 |
+
print(f"No suitable image URLs found in results for query: {query}")
|
319 |
+
return None # Found photos but no usable URLs
|
320 |
+
else:
|
321 |
+
print(f"No images found for query: {query}")
|
322 |
+
return None # API returned empty 'photos' list
|
323 |
+
|
324 |
+
elif response.status_code == 401:
|
325 |
+
print(f" Pexels API Error: Unauthorized (401). Check your API Key.")
|
326 |
+
return None # Stop searching if key is bad
|
327 |
+
elif response.status_code == 429:
|
328 |
+
print(f" Rate limit hit (attempt {attempt+1}/{max_retries}). Retrying in {retry_delay} seconds...")
|
329 |
+
time.sleep(retry_delay)
|
330 |
+
retry_delay *= 2
|
331 |
+
elif response.status_code == 522:
|
332 |
+
print(f" Pexels API Error: Connection Timed Out (522) (attempt {attempt+1}/{max_retries}). Retrying in {retry_delay} seconds...")
|
333 |
time.sleep(retry_delay)
|
334 |
retry_delay *= 2
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
335 |
else:
|
336 |
+
print(f" Error fetching images: {response.status_code} {response.text}")
|
337 |
+
if attempt < max_retries - 1:
|
338 |
+
print(f" Retrying in {retry_delay} seconds...")
|
339 |
+
time.sleep(retry_delay)
|
340 |
+
retry_delay *= 2
|
341 |
+
else:
|
342 |
+
break # Max retries for other errors
|
343 |
|
344 |
except requests.exceptions.Timeout:
|
345 |
+
print(f" Request timed out (attempt {attempt+1}/{max_retries}). Retrying in {retry_delay} seconds...")
|
346 |
time.sleep(retry_delay)
|
347 |
retry_delay *= 2
|
348 |
except requests.exceptions.RequestException as e:
|
349 |
+
print(f" Request exception: {e}")
|
350 |
+
if attempt < max_retries - 1:
|
351 |
+
print(f" Retrying in {retry_delay} seconds...")
|
352 |
+
time.sleep(retry_delay)
|
353 |
+
retry_delay *= 2
|
354 |
+
else:
|
355 |
+
break # Max retries after request exception
|
356 |
|
357 |
+
print(f"No Pexels images found for query: {query} after all attempts")
|
358 |
return None
|
359 |
|
360 |
+
# Added back search_google_images as it's used in the provided generate_media
|
361 |
+
def search_google_images(query):
|
362 |
+
"""Search for images on Google Images (use sparingly and ethically)."""
|
363 |
+
print(f"Attempting Google Image search for: {query} (Use with caution)")
|
364 |
+
try:
|
365 |
+
search_url = f"https://www.google.com/search?q={quote(query)}&tbm=isch&safe=active"
|
366 |
+
headers = {"User-Agent": USER_AGENT}
|
367 |
+
response = requests.get(search_url, headers=headers, timeout=10)
|
368 |
+
response.raise_for_status()
|
369 |
+
soup = BeautifulSoup(response.text, "html.parser")
|
370 |
+
|
371 |
+
img_tags = soup.find_all("img")
|
372 |
+
image_urls = []
|
373 |
+
for img in img_tags:
|
374 |
+
src = img.get("data-src") or img.get("src")
|
375 |
+
if src and src.startswith("http") and not "gstatic" in src and not src.startswith("data:image"):
|
376 |
+
if any(ext in src.lower() for ext in ['.jpg', '.jpeg', '.png', '.webp']):
|
377 |
+
image_urls.append(src)
|
378 |
+
|
379 |
+
if image_urls:
|
380 |
+
print(f"Found {len(image_urls)} potential Google Images for '{query}'.")
|
381 |
+
# Select randomly from top 10 potential URLs
|
382 |
+
return random.choice(image_urls[:min(len(image_urls), 10)])
|
383 |
+
else:
|
384 |
+
print(f"No suitable Google Images found for query: {query}")
|
385 |
+
return None
|
386 |
+
except requests.exceptions.RequestException as e:
|
387 |
+
print(f"Error during Google Images search request: {e}")
|
388 |
+
return None
|
389 |
+
except Exception as e:
|
390 |
+
print(f"Error parsing Google Images search results: {e}")
|
391 |
+
return None
|
392 |
+
|
393 |
|
394 |
+
def download_image(image_url, filename):
|
395 |
+
"""Download an image from a URL to a local file with enhanced error handling."""
|
|
|
|
|
396 |
try:
|
397 |
headers = {"User-Agent": USER_AGENT}
|
398 |
+
print(f"Downloading image from: {image_url} to {filename}")
|
399 |
+
response = requests.get(image_url, headers=headers, stream=True, timeout=20) # Increased timeout
|
400 |
response.raise_for_status()
|
401 |
+
|
402 |
+
# Ensure the target directory exists
|
403 |
+
os.makedirs(os.path.dirname(filename), exist_ok=True)
|
404 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
405 |
with open(filename, 'wb') as f:
|
406 |
+
for chunk in response.iter_content(chunk_size=8192):
|
407 |
+
f.write(chunk)
|
408 |
|
409 |
+
print(f" Image downloaded successfully to: {filename}")
|
410 |
+
|
411 |
+
# Validate and convert image
|
412 |
+
try:
|
413 |
+
img = Image.open(filename)
|
414 |
+
img.verify() # Check if it's a valid image file format
|
415 |
+
img.close() # Close file handle after verify
|
416 |
+
|
417 |
+
# Re-open to check mode and convert if needed
|
418 |
+
img = Image.open(filename)
|
419 |
+
if img.mode != 'RGB':
|
420 |
+
print(f" Converting image {os.path.basename(filename)} to RGB.")
|
421 |
+
# Ensure conversion doesn't create an empty file on error
|
422 |
+
try:
|
423 |
rgb_img = img.convert('RGB')
|
424 |
+
# Save to a temporary file first, then replace original
|
425 |
+
temp_filename = filename + ".tmp.jpg"
|
426 |
+
rgb_img.save(temp_filename, "JPEG")
|
427 |
rgb_img.close()
|
428 |
img.close()
|
429 |
+
os.replace(temp_filename, filename) # Atomic replace if possible
|
430 |
+
print(f" Image successfully converted and saved as {os.path.basename(filename)}")
|
431 |
+
except Exception as e_convert:
|
432 |
+
print(f" Error converting image to RGB: {e_convert}")
|
433 |
+
img.close() # Close original image handle
|
434 |
+
if os.path.exists(filename): os.remove(filename) # Remove partially converted/original
|
435 |
+
if os.path.exists(temp_filename): os.remove(temp_filename) # Clean up temp
|
436 |
+
return None
|
437 |
+
else:
|
438 |
+
img.close() # Close if already RGB
|
439 |
+
print(f" Image {os.path.basename(filename)} validated (already RGB).")
|
440 |
+
|
441 |
+
# Final check if file exists after processing
|
442 |
+
if os.path.exists(filename) and os.path.getsize(filename) > 0:
|
443 |
+
return filename
|
444 |
+
else:
|
445 |
+
print(f" Image file {os.path.basename(filename)} missing or empty after processing.")
|
446 |
+
return None
|
447 |
+
|
448 |
+
except (IOError, SyntaxError, Image.UnidentifiedImageError) as e_validate:
|
449 |
+
print(f" Downloaded file {os.path.basename(filename)} is not a valid image or corrupted: {e_validate}")
|
450 |
+
if os.path.exists(filename): os.remove(filename)
|
451 |
+
return None
|
452 |
+
except Exception as e_general_process:
|
453 |
+
print(f" Unexpected error during image processing: {e_general_process}")
|
454 |
+
if os.path.exists(filename): os.remove(filename)
|
455 |
+
return None
|
456 |
+
|
457 |
+
|
458 |
except requests.exceptions.RequestException as e_download:
|
459 |
+
print(f" Image download error: {e_download}")
|
460 |
+
# Don't remove file here, might not exist or be partial
|
461 |
return None
|
462 |
+
except Exception as e_general_download:
|
463 |
+
print(f" General error during image download request: {e_general_download}")
|
464 |
return None
|
465 |
|
466 |
+
def download_video(video_url, filename):
|
467 |
+
"""Download a video from a URL to a local file."""
|
468 |
+
try:
|
469 |
+
headers = {"User-Agent": USER_AGENT} # Add User-Agent
|
470 |
+
print(f"Downloading video from: {video_url} to {filename}")
|
471 |
+
response = requests.get(video_url, headers=headers, stream=True, timeout=60) # Longer timeout for videos
|
472 |
+
response.raise_for_status()
|
473 |
+
|
474 |
+
# Ensure the target directory exists
|
475 |
+
os.makedirs(os.path.dirname(filename), exist_ok=True)
|
476 |
+
|
477 |
+
with open(filename, 'wb') as f:
|
478 |
+
for chunk in response.iter_content(chunk_size=1024*1024): # Larger chunk size for videos
|
479 |
+
f.write(chunk)
|
480 |
+
print(f" Video downloaded successfully to: {filename}")
|
481 |
+
# Basic check: file exists and has size > 0
|
482 |
+
if os.path.exists(filename) and os.path.getsize(filename) > 0:
|
483 |
+
return filename
|
484 |
+
else:
|
485 |
+
print(f" Video file {os.path.basename(filename)} missing or empty after download.")
|
486 |
+
if os.path.exists(filename): os.remove(filename) # Clean up empty file
|
487 |
+
return None
|
488 |
+
except requests.exceptions.RequestException as e:
|
489 |
+
print(f"Video download error: {e}")
|
490 |
+
# Don't remove file here, might not exist or be partial
|
491 |
+
return None
|
492 |
+
except Exception as e_general:
|
493 |
+
print(f"General error during video download: {e_general}")
|
494 |
+
return None
|
495 |
+
|
496 |
+
def generate_media(prompt, user_image=None, current_index=0, total_segments=1):
|
497 |
"""
|
498 |
+
Generate a visual asset using user-provided functions.
|
499 |
+
Checks for "news", tries Pexels video/image, falls back to generic Pexels image.
|
|
|
500 |
"""
|
501 |
+
# Sanitize prompt for filename
|
502 |
+
safe_prompt_base = re.sub(r'[^\w\s-]', '', prompt).strip().replace(' ', '_')
|
503 |
+
safe_prompt_base = safe_prompt_base[:50] # Limit length
|
504 |
+
# Add timestamp for uniqueness within the temp folder
|
505 |
+
safe_prompt = f"{safe_prompt_base}_{int(time.time())}"
|
506 |
+
|
507 |
+
# --- News Check (Uses Google Images) ---
|
508 |
+
if "news" in prompt.lower():
|
509 |
+
print(f"News-related query detected: '{prompt}'. Trying Google Images...")
|
510 |
+
image_file = os.path.join(TEMP_FOLDER, f"{safe_prompt}_news.jpg")
|
511 |
+
image_url = search_google_images(prompt) # Use the added back function
|
512 |
+
if image_url:
|
513 |
+
downloaded_image = download_image(image_url, image_file)
|
514 |
+
if downloaded_image:
|
515 |
+
print(f" Using Google Image news asset: {os.path.basename(downloaded_image)}")
|
516 |
+
return {"path": downloaded_image, "asset_type": "image"}
|
517 |
+
else:
|
518 |
+
print(f" Google Image download failed for: {image_url}")
|
519 |
+
else:
|
520 |
+
print(f" Google Images search failed for prompt: '{prompt}'. Proceeding with Pexels.")
|
521 |
+
# Fall through to Pexels if Google fails
|
522 |
|
523 |
+
# --- Pexels Video Attempt ---
|
524 |
if random.random() < video_clip_probability:
|
525 |
+
print(f"Attempting Pexels video search for: '{prompt}'")
|
526 |
+
video_file = os.path.join(TEMP_FOLDER, f"{safe_prompt}_video.mp4")
|
527 |
+
video_url = search_pexels_videos(prompt, PEXELS_API_KEY)
|
528 |
if video_url:
|
529 |
+
downloaded_video = download_video(video_url, video_file)
|
530 |
+
if downloaded_video:
|
531 |
+
print(f" Using Pexels video asset: {os.path.basename(downloaded_video)}")
|
532 |
+
return {"path": downloaded_video, "asset_type": "video"}
|
533 |
+
else:
|
534 |
+
print(f" Pexels video download failed for: {video_url}")
|
535 |
else:
|
536 |
+
print(f" Pexels video search yielded no results for '{prompt}'.")
|
537 |
+
else:
|
538 |
+
print(f"Skipping Pexels video search based on probability for '{prompt}'.")
|
539 |
|
540 |
+
|
541 |
+
# --- Pexels Image Attempt ---
|
542 |
+
print(f"Attempting Pexels image search for: '{prompt}'")
|
543 |
+
image_file = os.path.join(TEMP_FOLDER, f"{safe_prompt}_image.jpg")
|
544 |
+
image_url = search_pexels_images(prompt, PEXELS_API_KEY)
|
545 |
if image_url:
|
546 |
+
downloaded_image = download_image(image_url, image_file)
|
547 |
+
if downloaded_image:
|
548 |
+
print(f" Using Pexels image asset: {os.path.basename(downloaded_image)}")
|
549 |
+
return {"path": downloaded_image, "asset_type": "image"}
|
550 |
+
else:
|
551 |
+
print(f" Pexels image download failed for: {image_url}")
|
552 |
else:
|
553 |
+
print(f" Pexels image search yielded no results for '{prompt}'.")
|
554 |
+
|
555 |
+
# --- Fallback Pexels Image Attempt ---
|
556 |
+
# Avoid fallback for instructional prompts
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
557 |
if "subscribe" not in prompt.lower() and "cta" not in prompt.lower():
|
558 |
+
fallback_terms = ["technology", "abstract", "nature", "background", "texture"]
|
559 |
+
term = random.choice(fallback_terms)
|
560 |
+
print(f"All specific searches failed for '{prompt}'. Trying Pexels fallback term: '{term}'")
|
561 |
+
fallback_file = os.path.join(TEMP_FOLDER, f"{safe_prompt}_fallback_{term}.jpg")
|
562 |
+
fallback_url = search_pexels_images(term, PEXELS_API_KEY)
|
563 |
if fallback_url:
|
564 |
+
downloaded_fallback = download_image(fallback_url, fallback_file)
|
565 |
+
if downloaded_fallback:
|
566 |
+
print(f" Using Pexels fallback image asset: {os.path.basename(downloaded_fallback)}")
|
567 |
+
return {"path": downloaded_fallback, "asset_type": "image"}
|
568 |
else:
|
569 |
+
print(f" Pexels fallback image download failed for term: '{term}'")
|
570 |
else:
|
571 |
+
print(f" Pexels fallback image search failed for term: '{term}'")
|
572 |
else:
|
573 |
+
print(f"Skipping fallback search for instructional prompt: '{prompt}'")
|
574 |
+
|
575 |
|
576 |
+
# --- Final Failure ---
|
577 |
+
print(f"FATAL: Failed to generate any visual asset for prompt: '{prompt}'")
|
578 |
+
return None
|
579 |
|
580 |
+
# --- End: User Provided Functions ---
|
|
|
|
|
581 |
|
582 |
|
583 |
def generate_tts(text, voice_id, speed):
|
584 |
"""Generate TTS audio using Kokoro, falling back to gTTS."""
|
585 |
+
# (Retained from previous versions)
|
586 |
safe_text_prefix = re.sub(r'[^\w\s-]', '', text[:20]).strip().replace(' ', '_')
|
587 |
output_filename = os.path.join(TEMP_FOLDER, f"tts_{safe_text_prefix}_{voice_id}.wav")
|
588 |
if pipeline:
|
|
|
613 |
wav_path = output_filename
|
614 |
tts.save(mp3_path)
|
615 |
audio = AudioSegment.from_mp3(mp3_path)
|
616 |
+
# Ensure target directory exists for export
|
617 |
+
os.makedirs(os.path.dirname(wav_path), exist_ok=True)
|
618 |
audio.export(wav_path, format="wav")
|
619 |
os.remove(mp3_path)
|
620 |
print(f"gTTS audio saved and converted to {wav_path}")
|
621 |
+
# Check if file exists and has size
|
622 |
+
if os.path.exists(wav_path) and os.path.getsize(wav_path) > 0:
|
623 |
+
return wav_path
|
624 |
+
else:
|
625 |
+
print(f"Error: gTTS output file missing or empty: {wav_path}")
|
626 |
+
return None
|
627 |
except ImportError:
|
628 |
print("Error: gTTS or pydub might not be installed. Cannot use gTTS fallback.")
|
629 |
return None
|
|
|
633 |
|
634 |
def apply_kenburns_effect(clip, target_resolution, duration):
|
635 |
"""Apply a randomized Ken Burns effect (zoom/pan) to an image clip."""
|
636 |
+
# (Retained from previous versions)
|
637 |
target_w, target_h = target_resolution
|
638 |
+
try: # Add try-except around accessing clip properties
|
639 |
+
img_w, img_h = clip.size
|
640 |
+
except Exception as e:
|
641 |
+
print(f"Error accessing image clip size: {e}")
|
642 |
+
# Cannot apply effect if size is unknown, return original clip resized
|
643 |
+
return clip.resize(newsize=target_resolution).set_duration(duration)
|
644 |
+
|
645 |
scale_factor = 1.2
|
646 |
+
# Handle potential division by zero if img_h is 0
|
647 |
+
if img_h == 0: img_h = 1
|
648 |
+
img_aspect = img_w / img_h
|
649 |
+
target_aspect = target_w / target_h
|
650 |
+
|
651 |
+
# Resize logic
|
652 |
+
if img_aspect > target_aspect:
|
653 |
final_h = target_h * scale_factor
|
654 |
+
final_w = final_h * img_aspect
|
655 |
else:
|
656 |
final_w = target_w * scale_factor
|
657 |
+
final_h = final_w / img_aspect # Use img_aspect here
|
658 |
+
|
659 |
final_w, final_h = int(final_w), int(final_h)
|
660 |
+
# Ensure final dimensions are not zero
|
661 |
+
if final_w <= 0: final_w = 1
|
662 |
+
if final_h <= 0: final_h = 1
|
663 |
+
|
664 |
try:
|
665 |
+
# Use MoviePy's resize for ImageClip directly before applying fl
|
|
|
|
|
|
|
|
|
666 |
resized_clip = clip.resize(newsize=(final_w, final_h)).set_duration(duration)
|
667 |
+
# pil_img = Image.fromarray(clip.get_frame(0)) # Avoid get_frame here
|
668 |
+
# resized_pil = pil_img.resize((final_w, final_h), Image.Resampling.LANCZOS)
|
669 |
+
# resized_clip = ImageClip(np.array(resized_pil)).set_duration(duration)
|
670 |
+
except Exception as e:
|
671 |
+
print(f"Warning: Error during resize for Ken Burns, using MoviePy default: {e}")
|
672 |
+
try:
|
673 |
+
# Fallback resize
|
674 |
+
resized_clip = clip.resize(newsize=(final_w, final_h)).set_duration(duration)
|
675 |
+
except Exception as resize_err:
|
676 |
+
print(f"FATAL error during fallback resize for Ken Burns: {resize_err}")
|
677 |
+
# Return original clip resized to target as last resort
|
678 |
+
return clip.resize(newsize=target_resolution).set_duration(duration)
|
679 |
+
|
680 |
+
|
681 |
max_move_x = final_w - target_w
|
682 |
max_move_y = final_h - target_h
|
683 |
+
# Ensure max_move is not negative if resizing failed slightly
|
684 |
+
max_move_x = max(0, max_move_x)
|
685 |
+
max_move_y = max(0, max_move_y)
|
686 |
+
|
687 |
effect = random.choice(['zoom_in', 'zoom_out', 'pan_lr', 'pan_rl', 'pan_td', 'pan_dt'])
|
688 |
if effect == 'zoom_in': zoom_start, zoom_end = 1.0, scale_factor; x_start, x_end = max_move_x / 2, max_move_x / 2; y_start, y_end = max_move_y / 2, max_move_y / 2
|
689 |
elif effect == 'zoom_out': zoom_start, zoom_end = scale_factor, 1.0; x_start, x_end = max_move_x / 2, max_move_x / 2; y_start, y_end = max_move_y / 2, max_move_y / 2
|
|
|
691 |
elif effect == 'pan_rl': zoom_start, zoom_end = scale_factor, scale_factor; x_start, x_end = max_move_x, 0; y_start, y_end = max_move_y / 2, max_move_y / 2
|
692 |
elif effect == 'pan_td': zoom_start, zoom_end = scale_factor, scale_factor; x_start, x_end = max_move_x / 2, max_move_x / 2; y_start, y_end = 0, max_move_y
|
693 |
else: zoom_start, zoom_end = scale_factor, scale_factor; x_start, x_end = max_move_x / 2, max_move_x / 2; y_start, y_end = max_move_y, 0
|
694 |
+
|
695 |
def make_frame(t):
|
696 |
+
# Protect against duration being zero
|
697 |
+
interp = t / duration if duration > 0 else 0
|
698 |
+
|
699 |
current_zoom = zoom_start + (zoom_end - zoom_start) * interp
|
700 |
current_x = x_start + (x_end - x_start) * interp
|
701 |
current_y = y_start + (y_end - y_start) * interp
|
702 |
+
|
703 |
+
# Avoid division by zero for zoom
|
704 |
+
if current_zoom <= 0: current_zoom = 1e-6 # Small positive number
|
705 |
+
|
706 |
+
# Calculate crop box dimensions based on current zoom relative to the base scale_factor
|
707 |
+
crop_w = target_w / (current_zoom / scale_factor)
|
708 |
+
crop_h = target_h / (current_zoom / scale_factor)
|
709 |
crop_w = max(1, int(crop_w)); crop_h = max(1, int(crop_h))
|
710 |
+
|
711 |
+
# Calculate top-left corner (x1, y1)
|
712 |
x1 = current_x; y1 = current_y
|
713 |
x1 = max(0, min(x1, final_w - crop_w)); y1 = max(0, min(y1, final_h - crop_h))
|
714 |
+
|
715 |
+
# Get frame from the *resized* clip
|
716 |
frame = resized_clip.get_frame(t)
|
717 |
+
|
718 |
+
# Crop using numpy slicing
|
719 |
+
try:
|
720 |
+
# Ensure indices are integers and within bounds
|
721 |
+
y1_int, y2_int = int(y1), int(y1 + crop_h)
|
722 |
+
x1_int, x2_int = int(x1), int(x1 + crop_w)
|
723 |
+
# Clamp coordinates to frame dimensions BEFORE slicing
|
724 |
+
y1_int = max(0, min(y1_int, frame.shape[0] - 1))
|
725 |
+
y2_int = max(y1_int + 1, min(y2_int, frame.shape[0])) # Ensure y2 > y1
|
726 |
+
x1_int = max(0, min(x1_int, frame.shape[1] - 1))
|
727 |
+
x2_int = max(x1_int + 1, min(x2_int, frame.shape[1])) # Ensure x2 > x1
|
728 |
+
|
729 |
+
# Check if dimensions are valid after clamping
|
730 |
+
if y2_int <= y1_int or x2_int <= x1_int:
|
731 |
+
print(f"Warning: Invalid crop dimensions after clamping ({y1_int}:{y2_int}, {x1_int}:{x2_int}). Returning uncropped frame.")
|
732 |
+
# Resize original frame to target as fallback for this frame
|
733 |
+
return cv2.resize(frame, (target_w, target_h), interpolation=cv2.INTER_AREA)
|
734 |
+
|
735 |
+
|
736 |
+
cropped_frame = frame[y1_int:y2_int, x1_int:x2_int]
|
737 |
+
|
738 |
+
# Check if cropped frame is empty
|
739 |
+
if cropped_frame.size == 0:
|
740 |
+
print(f"Warning: Cropped frame is empty ({y1_int}:{y2_int}, {x1_int}:{x2_int}). Returning uncropped frame.")
|
741 |
+
return cv2.resize(frame, (target_w, target_h), interpolation=cv2.INTER_AREA)
|
742 |
+
|
743 |
+
|
744 |
+
# Resize cropped frame to target
|
745 |
+
final_frame = cv2.resize(cropped_frame, (target_w, target_h), interpolation=cv2.INTER_LANCZOS4)
|
746 |
+
except IndexError as ie:
|
747 |
+
print(f"Error during frame cropping/resizing: {ie}. Frame shape: {frame.shape}, Crop: y={y1_int}:{y2_int}, x={x1_int}:{x2_int}")
|
748 |
+
# Fallback: return the original frame resized
|
749 |
+
final_frame = cv2.resize(frame, (target_w, target_h), interpolation=cv2.INTER_AREA)
|
750 |
+
except Exception as crop_resize_err:
|
751 |
+
print(f"Unexpected error during frame cropping/resizing: {crop_resize_err}")
|
752 |
+
final_frame = cv2.resize(frame, (target_w, target_h), interpolation=cv2.INTER_AREA)
|
753 |
+
|
754 |
+
|
755 |
return final_frame
|
756 |
+
|
757 |
+
# Apply the transformation using fl
|
758 |
+
try:
|
759 |
+
return resized_clip.fl(make_frame, apply_to=['mask'])
|
760 |
+
except Exception as fl_err:
|
761 |
+
print(f"Error applying Ken Burns effect via fl: {fl_err}")
|
762 |
+
# Return the resized clip without the effect as fallback
|
763 |
+
return resized_clip.resize(newsize=target_resolution)
|
764 |
+
|
765 |
|
766 |
def resize_to_fill(clip, target_resolution):
|
767 |
"""Resize and crop a video clip to fill the target resolution."""
|
768 |
+
# (Retained from previous versions)
|
769 |
target_w, target_h = target_resolution
|
770 |
+
try:
|
771 |
+
# Ensure clip dimensions are valid
|
772 |
+
if clip.w <= 0 or clip.h <= 0:
|
773 |
+
print(f"Warning: Invalid clip dimensions ({clip.w}x{clip.h}). Cannot resize/crop.")
|
774 |
+
# Return a black clip of target size? Or fail? Let's return black.
|
775 |
+
from moviepy.editor import ColorClip
|
776 |
+
return ColorClip(size=target_resolution, color=(0,0,0), duration=clip.duration)
|
777 |
+
|
778 |
+
target_aspect = target_w / target_h
|
779 |
+
clip_aspect = clip.w / clip.h
|
780 |
+
|
781 |
+
if clip_aspect > target_aspect: resized_clip = clip.resize(height=target_h)
|
782 |
+
else: resized_clip = clip.resize(width=target_w)
|
783 |
+
|
784 |
+
# Ensure resized dimensions are valid
|
785 |
+
if resized_clip.w <= 0 or resized_clip.h <= 0:
|
786 |
+
print(f"Warning: Invalid resized clip dimensions ({resized_clip.w}x{resized_clip.h}).")
|
787 |
+
from moviepy.editor import ColorClip
|
788 |
+
return ColorClip(size=target_resolution, color=(0,0,0), duration=clip.duration)
|
789 |
+
|
790 |
+
|
791 |
+
crop_x = max(0, (resized_clip.w - target_w) / 2)
|
792 |
+
crop_y = max(0, (resized_clip.h - target_h) / 2)
|
793 |
+
|
794 |
+
# Use integer coordinates for cropping
|
795 |
+
cropped_clip = resized_clip.crop(x1=int(crop_x), y1=int(crop_y), width=target_w, height=target_h)
|
796 |
+
return cropped_clip
|
797 |
+
except Exception as e:
|
798 |
+
print(f"Error during resize_to_fill: {e}")
|
799 |
+
# Fallback: try simple resize to target resolution
|
800 |
+
try:
|
801 |
+
return clip.resize(newsize=target_resolution)
|
802 |
+
except Exception as fallback_e:
|
803 |
+
print(f"Fallback resize also failed: {fallback_e}")
|
804 |
+
# Last resort: return black clip
|
805 |
+
from moviepy.editor import ColorClip
|
806 |
+
return ColorClip(size=target_resolution, color=(0,0,0), duration=clip.duration)
|
807 |
+
|
808 |
|
809 |
def add_background_music(video_clip, music_file_path, volume):
|
810 |
"""Add background music, looping if necessary."""
|
811 |
+
# (Retained from previous versions)
|
812 |
if not music_file_path or not os.path.exists(music_file_path):
|
813 |
print("No background music file found or provided. Skipping.")
|
814 |
return video_clip
|
815 |
try:
|
816 |
print(f"Adding background music from: {music_file_path}")
|
817 |
bg_music = AudioFileClip(music_file_path)
|
818 |
+
# Check for valid duration
|
819 |
+
if not bg_music or bg_music.duration is None or bg_music.duration <= 0:
|
820 |
+
print("Warning: Background music file is invalid or has zero duration. Skipping.")
|
821 |
+
return video_clip
|
822 |
+
if not video_clip or video_clip.duration is None or video_clip.duration <=0:
|
823 |
+
print("Warning: Video clip has invalid duration. Cannot add music.")
|
824 |
+
# Return original clip without audio modification
|
825 |
+
return video_clip
|
826 |
+
|
827 |
+
|
828 |
if bg_music.duration > video_clip.duration: bg_music = bg_music.subclip(0, video_clip.duration)
|
829 |
elif bg_music.duration < video_clip.duration:
|
830 |
loops_needed = math.ceil(video_clip.duration / bg_music.duration)
|
831 |
+
# Ensure bg_music can be concatenated
|
832 |
+
try:
|
833 |
+
bg_music = concatenate_audioclips([bg_music] * loops_needed)
|
834 |
+
bg_music = bg_music.subclip(0, video_clip.duration)
|
835 |
+
except Exception as concat_err:
|
836 |
+
print(f"Error looping background music: {concat_err}. Skipping music.")
|
837 |
+
return video_clip # Return original video if looping fails
|
838 |
+
|
839 |
bg_music = bg_music.volumex(volume)
|
840 |
+
|
841 |
if video_clip.audio:
|
842 |
+
# Ensure video audio is valid
|
843 |
+
if video_clip.audio.duration is None or video_clip.audio.duration <= 0:
|
844 |
+
print("Warning: Video clip audio has invalid duration. Replacing with background music.")
|
845 |
+
final_audio = bg_music
|
846 |
+
else:
|
847 |
+
# Make durations match exactly before composing
|
848 |
+
try:
|
849 |
+
video_audio_adjusted = video_clip.audio.set_duration(video_clip.duration)
|
850 |
+
bg_music_adjusted = bg_music.set_duration(video_clip.duration)
|
851 |
+
final_audio = CompositeAudioClip([video_audio_adjusted, bg_music_adjusted])
|
852 |
+
except Exception as audio_comp_err:
|
853 |
+
print(f"Error composing audio clips: {audio_comp_err}. Skipping music.")
|
854 |
+
return video_clip # Return original video
|
855 |
else:
|
856 |
+
final_audio = bg_music # If original clip has no audio
|
857 |
+
|
858 |
video_clip = video_clip.set_audio(final_audio)
|
859 |
print("Background music added successfully.")
|
860 |
return video_clip
|
|
|
864 |
|
865 |
def create_segment_clip(media_info, tts_path, narration_text):
|
866 |
"""Create a single video segment (clip) with visuals, audio, and subtitles."""
|
867 |
+
# (Retained from previous versions, with minor stability checks)
|
868 |
try:
|
869 |
media_path = media_info['path']
|
870 |
asset_type = media_info['asset_type']
|
871 |
print(f"Creating clip segment: Type={asset_type}, Media={os.path.basename(media_path)}")
|
872 |
+
if not os.path.exists(tts_path) or os.path.getsize(tts_path) == 0:
|
873 |
+
print(f"Error: TTS file missing or empty: {tts_path}"); return None
|
874 |
audio_clip = AudioFileClip(tts_path)
|
875 |
+
if audio_clip.duration is None or audio_clip.duration <= 0:
|
876 |
+
print(f"Error: Audio clip has invalid duration: {tts_path}"); return None
|
877 |
segment_duration = audio_clip.duration + 0.3
|
878 |
+
|
879 |
if asset_type == "video":
|
880 |
+
if not os.path.exists(media_path) or os.path.getsize(media_path) == 0:
|
881 |
+
print(f"Error: Video file missing or empty: {media_path}"); return None
|
882 |
+
try:
|
883 |
+
video_clip = VideoFileClip(media_path)
|
884 |
+
if video_clip.duration is None or video_clip.duration <= 0:
|
885 |
+
print(f"Error: Video clip has invalid duration: {media_path}")
|
886 |
+
video_clip.close() # Close the clip
|
887 |
+
return None
|
888 |
+
if video_clip.duration < segment_duration:
|
889 |
+
if video_clip.duration > 0:
|
890 |
+
loops = math.ceil(segment_duration / video_clip.duration)
|
891 |
+
video_clip_looped = concatenate_videoclips([video_clip] * loops)
|
892 |
+
video_clip.close() # Close original after looping
|
893 |
+
video_clip = video_clip_looped
|
894 |
+
else: # Should have been caught above, but double check
|
895 |
+
print(f"Error: Cannot loop zero-duration video: {media_path}")
|
896 |
+
video_clip.close(); return None
|
897 |
+
|
898 |
+
# Ensure subclip duration is valid
|
899 |
+
video_clip_final = video_clip.subclip(0, min(segment_duration, video_clip.duration))
|
900 |
+
video_clip.close() # Close intermediate clip if looped
|
901 |
+
video_clip = video_clip_final
|
902 |
+
|
903 |
+
visual_clip = resize_to_fill(video_clip, TARGET_RESOLUTION)
|
904 |
+
video_clip.close() # Close after resize_to_fill finishes
|
905 |
+
|
906 |
+
except Exception as video_load_err:
|
907 |
+
print(f"Error loading or processing video {media_path}: {video_load_err}")
|
908 |
+
return None
|
909 |
+
|
910 |
elif asset_type == "image":
|
911 |
+
if not os.path.exists(media_path) or os.path.getsize(media_path) == 0:
|
912 |
+
print(f"Error: Image file missing or empty: {media_path}"); return None
|
913 |
+
try:
|
914 |
+
# Load image clip
|
915 |
+
img_clip = ImageClip(media_path).set_duration(segment_duration)
|
916 |
+
# Apply Ken Burns
|
917 |
+
visual_clip = apply_kenburns_effect(img_clip, TARGET_RESOLUTION, segment_duration)
|
918 |
+
# Ensure final size - Ken Burns should handle this, but double check
|
919 |
+
visual_clip = visual_clip.resize(newsize=TARGET_RESOLUTION)
|
920 |
+
# Close the base image clip resource if possible (ImageClip doesn't have explicit close)
|
921 |
+
except Exception as img_load_err:
|
922 |
+
print(f"Error loading or processing image {media_path}: {img_load_err}")
|
923 |
+
return None
|
924 |
else: print(f"Error: Unknown asset type: {asset_type}"); return None
|
925 |
+
|
926 |
visual_clip = visual_clip.fadein(0.15).fadeout(0.15)
|
927 |
subtitle_clips = []
|
928 |
if USE_CAPTIONS and narration_text:
|
|
|
937 |
num_chunks = len(chunks); chunk_duration = audio_clip.duration / num_chunks
|
938 |
start_time = 0.1
|
939 |
for i, chunk_text in enumerate(chunks):
|
940 |
+
# Ensure chunk end time doesn't exceed visual clip duration
|
941 |
+
chunk_end_time = min(start_time + chunk_duration, visual_clip.duration - 0.05) # Leave tiny buffer
|
942 |
+
actual_chunk_duration = max(0.1, chunk_end_time - start_time) # Min duration 0.1s
|
943 |
+
|
944 |
try:
|
945 |
txt_clip = TextClip(txt=chunk_text, fontsize=font_size, font=caption_font, color=caption_style_text_color,
|
946 |
bg_color=caption_style_bg_color, method='label', align='center',
|
947 |
size=(TARGET_RESOLUTION[0] * 0.8, None))
|
948 |
txt_clip = txt_clip.set_position(('center', TARGET_RESOLUTION[1] * 0.80))
|
949 |
+
txt_clip = txt_clip.set_start(start_time).set_duration(actual_chunk_duration)
|
950 |
subtitle_clips.append(txt_clip)
|
951 |
+
start_time = chunk_end_time # Next chunk starts where the last one ended
|
952 |
except Exception as txt_err:
|
953 |
print(f"ERROR creating TextClip for '{chunk_text}': {txt_err}. Skipping subtitle chunk.")
|
|
|
954 |
|
955 |
+
final_clip = CompositeVideoClip([visual_clip] + subtitle_clips, size=TARGET_RESOLUTION) if subtitle_clips else visual_clip
|
956 |
+
# Set audio with slight offset
|
957 |
+
final_clip = final_clip.set_audio(audio_clip.set_start(0.15).set_duration(audio_clip.duration))
|
958 |
+
# Ensure final clip duration matches segment duration closely
|
959 |
+
final_clip = final_clip.set_duration(segment_duration)
|
960 |
+
|
961 |
print(f"Clip segment created successfully. Duration: {final_clip.duration:.2f}s")
|
962 |
+
# Explicitly close audio clip resource
|
963 |
+
audio_clip.close()
|
964 |
+
# Close visual clip resources if they have close methods (VideoClips do)
|
965 |
+
if hasattr(visual_clip, 'close'):
|
966 |
+
visual_clip.close()
|
967 |
+
for sub in subtitle_clips:
|
968 |
+
if hasattr(sub, 'close'): # TextClips might not have close
|
969 |
+
sub.close()
|
970 |
+
|
971 |
return final_clip
|
972 |
except Exception as e:
|
973 |
print(f"Error creating clip segment: {e}")
|
974 |
import traceback
|
975 |
traceback.print_exc()
|
976 |
+
# Clean up resources if possible on error
|
977 |
+
if 'audio_clip' in locals() and hasattr(audio_clip, 'close'): audio_clip.close()
|
978 |
+
if 'visual_clip' in locals() and hasattr(visual_clip, 'close'): visual_clip.close()
|
979 |
+
if 'video_clip' in locals() and hasattr(video_clip, 'close'): video_clip.close()
|
980 |
return None
|
981 |
|
982 |
+
|
983 |
# ---------------- Main Video Generation Function ---------------- #
|
984 |
|
985 |
def generate_full_video(user_input, resolution_choice, caption_choice, music_file_info):
|
986 |
"""Main function orchestrating the video generation process."""
|
987 |
+
# (Retained from previous versions - logic relies on the new generate_media)
|
|
|
988 |
global TARGET_RESOLUTION, TEMP_FOLDER, USE_CAPTIONS
|
989 |
print("\n--- Starting Video Generation ---"); start_time = time.time()
|
990 |
if resolution_choice == "Short (9:16)": TARGET_RESOLUTION = (1080, 1920); print("Resolution set to: Short (1080x1920)")
|
|
|
995 |
if music_file_info is not None:
|
996 |
try:
|
997 |
music_file_path = os.path.join(TEMP_FOLDER, "background_music.mp3")
|
998 |
+
# Ensure temp folder exists before copying
|
999 |
+
os.makedirs(TEMP_FOLDER, exist_ok=True)
|
1000 |
shutil.copy(music_file_info.name, music_file_path)
|
1001 |
print(f"Background music copied to: {music_file_path}")
|
1002 |
except Exception as e: print(f"Error handling uploaded music file: {e}"); music_file_path = None
|
|
|
1016 |
print(f" Prompt: {segment['prompt']}")
|
1017 |
print(f" Narration: {segment['narration']}")
|
1018 |
|
1019 |
+
media_info = generate_media(segment['prompt']) # Using user's generate_media
|
|
|
1020 |
if not media_info:
|
1021 |
print(f"Warning: Failed to get media for segment {i+1} ('{segment['prompt']}'). Skipping this segment.")
|
1022 |
+
continue
|
1023 |
|
1024 |
tts_path = generate_tts(segment['narration'], selected_voice, voice_speed)
|
1025 |
if not tts_path:
|
1026 |
print(f"Warning: Failed to generate TTS for segment {i+1}. Skipping segment.")
|
1027 |
if media_info and os.path.exists(media_info['path']):
|
1028 |
try: os.remove(media_info['path']); print(f"Cleaned up unused media: {media_info['path']}")
|
1029 |
+
except OSError as e: print(f"Error removing unused media {media_info['path']}: {e}")
|
1030 |
continue
|
1031 |
|
1032 |
+
# Pass current_index=i to create_segment if needed by that function (currently not)
|
1033 |
clip = create_segment_clip(media_info, tts_path, segment['narration'])
|
1034 |
if clip:
|
1035 |
segment_clips.append(clip)
|
1036 |
else:
|
1037 |
print(f"Warning: Failed to create video clip for segment {i+1}. Skipping.")
|
1038 |
+
# Clean up files for this failed segment
|
1039 |
if media_info and os.path.exists(media_info['path']):
|
1040 |
try: os.remove(media_info['path']); print(f"Cleaned up media for failed clip: {media_info['path']}")
|
1041 |
+
except OSError as e: print(f"Error removing media for failed clip {media_info['path']}: {e}")
|
1042 |
if tts_path and os.path.exists(tts_path):
|
1043 |
try: os.remove(tts_path); print(f"Cleaned up TTS for failed clip: {tts_path}")
|
1044 |
+
except OSError as e: print(f"Error removing TTS for failed clip {tts_path}: {e}")
|
1045 |
+
|
1046 |
|
1047 |
if not segment_clips:
|
1048 |
print("ERROR: No video clips were successfully created. Aborting.")
|
1049 |
+
try: shutil.rmtree(TEMP_FOLDER)
|
1050 |
+
except Exception as e: print(f"Error removing temp folder during abort: {e}")
|
1051 |
return None, "Error: Failed to create any video segments. Check logs for media/TTS issues."
|
1052 |
|
1053 |
print("\nStep 4: Concatenating video segments...");
|
1054 |
+
final_video = None # Initialize to None
|
1055 |
try:
|
|
|
1056 |
valid_clips = [c for c in segment_clips if c is not None]
|
1057 |
+
if not valid_clips: raise ValueError("No valid clips remained after processing.")
|
1058 |
+
# Concatenate
|
1059 |
final_video = concatenate_videoclips(valid_clips, method="compose")
|
1060 |
print("Segments concatenated successfully.")
|
1061 |
+
# Close individual clips after concatenation
|
1062 |
+
for clip in valid_clips:
|
1063 |
+
if hasattr(clip, 'close'): clip.close()
|
1064 |
+
|
1065 |
except Exception as e:
|
1066 |
+
print(f"ERROR: Failed to concatenate video clips: {e}");
|
1067 |
+
# Clean up individual clips if concatenation failed
|
1068 |
+
for clip in segment_clips:
|
1069 |
+
if clip and hasattr(clip, 'close'): clip.close()
|
1070 |
+
try: shutil.rmtree(TEMP_FOLDER)
|
1071 |
+
except Exception as e_clean: print(f"Error removing temp folder after concat error: {e_clean}")
|
1072 |
+
return None, f"Error: Concatenation failed: {e}"
|
1073 |
+
|
1074 |
+
|
1075 |
+
print("\nStep 5: Adding background music...");
|
1076 |
+
if final_video: # Only add music if concatenation was successful
|
1077 |
+
final_video = add_background_music(final_video, music_file_path, bg_music_volume)
|
1078 |
+
else:
|
1079 |
+
print("Skipping background music as concatenation failed.")
|
1080 |
|
|
|
1081 |
|
1082 |
print(f"\nStep 6: Exporting final video to '{OUTPUT_VIDEO_FILENAME}'..."); export_success = False
|
1083 |
+
if final_video: # Only export if we have a final video object
|
1084 |
+
try:
|
1085 |
+
final_video.write_videofile(OUTPUT_VIDEO_FILENAME, codec='libx264', audio_codec='aac', fps=fps, preset=preset, threads=4, logger='bar', ffmpeg_params=["-vsync", "vfr"]) # Added vsync param
|
1086 |
+
print(f"Final video saved successfully as {OUTPUT_VIDEO_FILENAME}")
|
1087 |
+
export_success = True
|
1088 |
+
except Exception as e:
|
1089 |
+
print(f"ERROR: Failed to write final video file: {e}"); import traceback; traceback.print_exc()
|
1090 |
+
finally:
|
1091 |
+
# Ensure final video resources are closed after export
|
1092 |
+
if hasattr(final_video, 'close'): final_video.close()
|
1093 |
+
print("Closed final video resources.")
|
1094 |
+
else:
|
1095 |
+
print("Skipping export as final video generation failed.")
|
1096 |
+
|
1097 |
|
1098 |
print("\nStep 7: Cleaning up temporary files...");
|
1099 |
try: shutil.rmtree(TEMP_FOLDER); print(f"Temporary folder {TEMP_FOLDER} removed.")
|
|
|
1102 |
end_time = time.time(); total_time = end_time - start_time
|
1103 |
print(f"\n--- Video Generation Finished ---"); print(f"Total time: {total_time:.2f} seconds")
|
1104 |
if export_success: return OUTPUT_VIDEO_FILENAME, f"Video generation complete! Time: {total_time:.2f}s"
|
1105 |
+
elif not final_video: return None, f"Error: Video generation failed before export. Check logs. Time: {total_time:.2f}s"
|
1106 |
else: return None, f"Error: Video export failed. Check logs. Time: {total_time:.2f}s"
|
1107 |
|
1108 |
|
1109 |
# ---------------- Gradio Interface Definition ---------------- #
|
1110 |
+
# (Retained from previous versions)
|
1111 |
VOICE_CHOICES = {
|
1112 |
'Emma (US Female)': 'af_heart', 'Bella (US Female)': 'af_bella', 'Nicole (US Female)': 'af_nicole',
|
1113 |
'Sarah (US Female)': 'af_sarah', 'Michael (US Male)': 'am_michael', 'Eric (US Male)': 'am_eric',
|
|
|
1116 |
}
|
1117 |
def gradio_interface_handler(user_prompt, resolution, captions, bg_music, voice_name, video_prob, music_vol, video_fps, export_preset, tts_speed, caption_size):
|
1118 |
print("\n--- Received Request from Gradio ---")
|
1119 |
+
print(f"Prompt: {user_prompt[:50]}...")
|
1120 |
+
print(f"Resolution: {resolution}")
|
1121 |
+
print(f"Captions: {captions}")
|
1122 |
+
print(f"Music File: {'Provided' if bg_music else 'None'}")
|
1123 |
+
print(f"Voice: {voice_name}")
|
1124 |
+
print(f"Video Probability: {video_prob}%")
|
1125 |
+
print(f"Music Volume: {music_vol}")
|
1126 |
+
print(f"FPS: {video_fps}")
|
1127 |
+
print(f"Preset: {export_preset}")
|
1128 |
+
print(f"TTS Speed: {tts_speed}")
|
1129 |
+
print(f"Caption Size: {caption_size}")
|
1130 |
+
|
1131 |
global selected_voice, voice_speed, font_size, video_clip_probability, bg_music_volume, fps, preset
|
1132 |
selected_voice = VOICE_CHOICES.get(voice_name, 'af_heart')
|
1133 |
voice_speed = tts_speed; font_size = caption_size; video_clip_probability = video_prob / 100.0
|
|
|
1175 |
print("!!! Please replace 'YOUR_PEXELS_API_KEY' and !!!")
|
1176 |
print("!!! 'YOUR_OPENROUTER_API_KEY' with your actual keys. !!!")
|
1177 |
print("!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!\n")
|
1178 |
+
iface.launch(share=True, debug=True) # Share=True for public link, Debug=True for more logs
|