Update core/visual_engine.py
Browse files- core/visual_engine.py +202 -125
core/visual_engine.py
CHANGED
@@ -10,86 +10,164 @@ import requests
|
|
10 |
import io
|
11 |
import time
|
12 |
import random
|
13 |
-
|
14 |
-
|
15 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
class VisualEngine:
|
17 |
def __init__(self, output_dir="temp_cinegen_media"):
|
18 |
-
self.output_dir = output_dir
|
19 |
-
self.
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
self.
|
24 |
-
self.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
25 |
self.video_frame_size = (1280, 720)
|
26 |
-
|
27 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
28 |
|
29 |
def set_openai_api_key(self,k):
|
30 |
-
self.openai_api_key=k
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
38 |
def set_pexels_api_key(self,k):
|
39 |
self.pexels_api_key=k; self.USE_PEXELS=bool(k)
|
40 |
-
print(f"Pexels {'Ready' if k else 'Disabled'}.")
|
41 |
-
|
42 |
-
def _get_text_dimensions(self,
|
43 |
-
if not
|
44 |
try:
|
45 |
-
if hasattr(
|
46 |
-
|
47 |
-
|
48 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
49 |
|
50 |
-
def _create_placeholder_image_content(self,
|
51 |
-
img=Image.new('RGB',
|
52 |
-
if not
|
53 |
-
|
54 |
-
for
|
55 |
-
|
56 |
-
if self._get_text_dimensions(
|
57 |
else:
|
58 |
-
if
|
59 |
-
|
60 |
-
if
|
61 |
-
if not
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
74 |
if not self.USE_PEXELS or not self.pexels_api_key: return None
|
75 |
headers = {"Authorization": self.pexels_api_key}
|
76 |
-
params = {"query": query, "per_page":
|
77 |
-
pexels_filename =
|
78 |
filepath = os.path.join(self.output_dir, pexels_filename)
|
79 |
try:
|
80 |
print(f"Searching Pexels for: '{query}'")
|
81 |
-
|
82 |
params["query"] = effective_query
|
83 |
-
|
|
|
84 |
response.raise_for_status(); data = response.json()
|
85 |
-
if data.get("photos"):
|
86 |
-
photo_url = data["photos"][0]["src"]["large2x"]
|
87 |
-
image_response = requests.get(photo_url, timeout=
|
88 |
img_data = Image.open(io.BytesIO(image_response.content))
|
89 |
if img_data.mode != 'RGB': img_data = img_data.convert('RGB')
|
90 |
img_data.save(filepath); print(f"Pexels image saved: {filepath}"); return filepath
|
91 |
-
else: print(f"No photos on Pexels for: '{effective_query}'")
|
92 |
-
except Exception as e: print(f"Pexels error for '{query}': {e}")
|
93 |
return None
|
94 |
|
95 |
def generate_image_visual(self, image_prompt_text, scene_data, scene_identifier_filename):
|
@@ -101,87 +179,82 @@ class VisualEngine:
|
|
101 |
print(f"Attempt {attempt+1}: DALL-E ({self.dalle_model}) for: {image_prompt_text[:120]}...")
|
102 |
client = openai.OpenAI(api_key=self.openai_api_key, timeout=90.0)
|
103 |
response = client.images.generate(
|
104 |
-
model=self.dalle_model,
|
105 |
-
|
106 |
-
n=1,
|
107 |
-
size=self.image_size_dalle3,
|
108 |
-
quality="hd",
|
109 |
-
response_format="url",
|
110 |
-
style="vivid"
|
111 |
)
|
112 |
image_url = response.data[0].url
|
113 |
revised_prompt = getattr(response.data[0], 'revised_prompt', None)
|
114 |
-
if revised_prompt:
|
115 |
-
print(f"DALL-E 3 revised_prompt: {revised_prompt[:100]}...")
|
116 |
|
117 |
image_response = requests.get(image_url, timeout=120)
|
118 |
image_response.raise_for_status()
|
119 |
img_data = Image.open(io.BytesIO(image_response.content))
|
120 |
-
if img_data.mode != 'RGB':
|
121 |
-
img_data = img_data.convert('RGB')
|
122 |
|
123 |
-
img_data.save(filepath)
|
124 |
-
print(f"AI Image (DALL-E) saved: {filepath}")
|
125 |
-
return filepath
|
126 |
-
|
127 |
except openai.RateLimitError as e:
|
128 |
print(f"OpenAI Rate Limit: {e}. Retrying after {5*(attempt+1)}s...")
|
129 |
time.sleep(5 * (attempt + 1))
|
130 |
-
|
131 |
-
|
132 |
-
|
133 |
-
|
134 |
-
|
135 |
-
continue # Go to the next attempt in the for loop
|
136 |
-
|
137 |
-
except openai.APIError as e:
|
138 |
-
print(f"OpenAI API Error: {e}")
|
139 |
-
break # Break from loop, will try Pexels/placeholder
|
140 |
-
except requests.exceptions.RequestException as e:
|
141 |
-
print(f"Requests Error (DALL-E image download): {e}")
|
142 |
-
break # Break from loop
|
143 |
-
except Exception as e:
|
144 |
-
print(f"Generic error (DALL-E gen): {e}")
|
145 |
-
break # Break from loop
|
146 |
|
147 |
-
|
148 |
-
|
149 |
-
|
150 |
-
pexels_query_text = scene_data.get('pexels_search_query_감독', f"{scene_data.get('emotional_beat','')} {scene_data.get('setting_description','')}")
|
151 |
pexels_path = self._search_pexels_image(pexels_query_text, scene_identifier_filename)
|
152 |
-
if pexels_path:
|
153 |
-
return pexels_path
|
154 |
|
155 |
print("Pexels also failed/disabled. Using placeholder.")
|
156 |
return self._create_placeholder_image_content(
|
157 |
f"[AI/Pexels Failed] Original Prompt: {image_prompt_text[:100]}...",
|
158 |
scene_identifier_filename, size=self.video_frame_size
|
159 |
)
|
160 |
-
else:
|
161 |
return self._create_placeholder_image_content(
|
162 |
image_prompt_text, scene_identifier_filename, size=self.video_frame_size
|
163 |
)
|
164 |
|
165 |
def generate_narration_audio(self, text_to_narrate, output_filename="narration_overall.mp3"):
|
166 |
-
if not self.USE_ELEVENLABS or not self.
|
167 |
-
print("ElevenLabs
|
|
|
|
|
168 |
audio_filepath = os.path.join(self.output_dir, output_filename)
|
169 |
try:
|
170 |
print(f"Generating ElevenLabs audio (Voice: {self.elevenlabs_voice_id}) for: {text_to_narrate[:70]}...")
|
171 |
-
|
172 |
-
|
173 |
-
|
174 |
-
|
175 |
-
|
176 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
177 |
return None
|
178 |
|
179 |
def create_video_from_images(self, image_data_list, overall_narration_path=None, output_filename="final_video.mp4", fps=24, duration_per_image=4.5):
|
180 |
-
if not image_data_list: return None
|
181 |
-
print(f"Creating video from {len(image_data_list)} image sets.")
|
182 |
-
processed_clips = []
|
183 |
-
narration_audio_clip = None
|
184 |
-
final_video_clip_obj = None
|
185 |
|
186 |
for i, data in enumerate(image_data_list):
|
187 |
img_path, scene_num, key_action = data.get('path'), data.get('scene_num', i+1), data.get('key_action', '')
|
@@ -191,12 +264,12 @@ class VisualEngine:
|
|
191 |
if pil_img.mode != 'RGB': pil_img = pil_img.convert('RGB')
|
192 |
img_copy = pil_img.copy()
|
193 |
img_copy.thumbnail(self.video_frame_size, Image.Resampling.LANCZOS)
|
194 |
-
canvas = Image.new('RGB', self.video_frame_size, (random.randint(0,
|
195 |
xo, yo = (self.video_frame_size[0]-img_copy.width)//2, (self.video_frame_size[1]-img_copy.height)//2
|
196 |
canvas.paste(img_copy, (xo,yo))
|
197 |
frame_np = np.array(canvas)
|
198 |
img_clip = ImageClip(frame_np).set_duration(duration_per_image)
|
199 |
-
end_scale = random.uniform(1.05, 1.12)
|
200 |
img_clip = img_clip.fx(vfx.resize, lambda t: 1 + (end_scale - 1) * (t / duration_per_image))
|
201 |
img_clip = img_clip.set_position('center')
|
202 |
if key_action:
|
@@ -208,9 +281,9 @@ class VisualEngine:
|
|
208 |
final_scene_clip = CompositeVideoClip([img_clip, txt_clip], size=self.video_frame_size)
|
209 |
else: final_scene_clip = img_clip
|
210 |
processed_clips.append(final_scene_clip)
|
211 |
-
except Exception as e: print(f"Error clip for {img_path}: {e}.")
|
212 |
|
213 |
-
if not processed_clips: print("No clips for video."); return None
|
214 |
transition = 0.8
|
215 |
final_video_clip_obj = concatenate_videoclips(processed_clips, padding=-transition, method="compose")
|
216 |
if final_video_clip_obj.duration > transition*2:
|
@@ -219,22 +292,26 @@ class VisualEngine:
|
|
219 |
if overall_narration_path and os.path.exists(overall_narration_path):
|
220 |
try:
|
221 |
narration_audio_clip = AudioFileClip(overall_narration_path)
|
222 |
-
|
223 |
-
|
|
|
224 |
final_video_clip_obj = final_video_clip_obj.subclip(0, narration_audio_clip.duration)
|
225 |
-
|
226 |
-
|
|
|
|
|
227 |
|
228 |
output_path = os.path.join(self.output_dir, output_filename)
|
229 |
try:
|
230 |
-
|
|
|
231 |
audio_codec='aac',
|
232 |
temp_audiofile=os.path.join(self.output_dir, f'temp-audio-{os.urandom(4).hex()}.m4a'),
|
233 |
-
remove_temp=True, threads=os.cpu_count() or 2, logger='bar', bitrate="5000k")
|
234 |
-
print(f"Video created: {output_path}"); return output_path
|
235 |
-
except Exception as e: print(f"Error writing video: {e}"); return None
|
236 |
finally:
|
237 |
-
for
|
238 |
-
if hasattr(
|
239 |
if narration_audio_clip and hasattr(narration_audio_clip, 'close'): narration_audio_clip.close()
|
240 |
if final_video_clip_obj and hasattr(final_video_clip_obj, 'close'): final_video_clip_obj.close()
|
|
|
10 |
import io
|
11 |
import time
|
12 |
import random
|
13 |
+
import subprocess # For the dummy video fallback
|
14 |
+
|
15 |
+
# --- ElevenLabs Import ---
|
16 |
+
ELEVENLABS_CLIENT_IMPORTED = False
|
17 |
+
ElevenLabsAPIClient = None # Placeholder for the class
|
18 |
+
Voice = None # Placeholder for the class
|
19 |
+
VoiceSettings = None # Placeholder for the class
|
20 |
+
|
21 |
+
try:
|
22 |
+
from elevenlabs.client import ElevenLabs as ImportedElevenLabsClient
|
23 |
+
from elevenlabs import Voice as ImportedVoice, VoiceSettings as ImportedVoiceSettings
|
24 |
+
|
25 |
+
ElevenLabsAPIClient = ImportedElevenLabsClient
|
26 |
+
Voice = ImportedVoice
|
27 |
+
VoiceSettings = ImportedVoiceSettings
|
28 |
+
ELEVENLABS_CLIENT_IMPORTED = True
|
29 |
+
print("Successfully imported ElevenLabs client components (SDK v1.x.x pattern).")
|
30 |
+
except ImportError as e_eleven:
|
31 |
+
print(f"WARNING: Could not import ElevenLabs client components: {e_eleven}. ElevenLabs audio generation will be disabled.")
|
32 |
+
except Exception as e_gen_eleven: # Catch any other general import error for elevenlabs
|
33 |
+
print(f"WARNING: General error importing ElevenLabs: {e_gen_eleven}. ElevenLabs audio generation will be disabled.")
|
34 |
+
|
35 |
+
|
36 |
class VisualEngine:
|
37 |
def __init__(self, output_dir="temp_cinegen_media"):
|
38 |
+
self.output_dir = output_dir
|
39 |
+
os.makedirs(self.output_dir, exist_ok=True)
|
40 |
+
|
41 |
+
self.font_filename = "arial.ttf"
|
42 |
+
self.font_path_in_container = f"/usr/local/share/fonts/truetype/mycustomfonts/{self.font_filename}"
|
43 |
+
self.font_size_pil = 20
|
44 |
+
self.video_overlay_font_size = 30
|
45 |
+
self.video_overlay_font_color = 'white'
|
46 |
+
self.video_overlay_font = 'Arial-Bold'
|
47 |
+
|
48 |
+
try:
|
49 |
+
self.font = ImageFont.truetype(self.font_path_in_container, self.font_size_pil)
|
50 |
+
# print(f"Placeholder font loaded: {self.font_path_in_container}.") # Less verbose
|
51 |
+
except IOError:
|
52 |
+
print(f"Warning: Placeholder font '{self.font_path_in_container}' not loaded. Using default.")
|
53 |
+
self.font = ImageFont.load_default()
|
54 |
+
self.font_size_pil = 10
|
55 |
+
|
56 |
+
self.openai_api_key = None
|
57 |
+
self.USE_AI_IMAGE_GENERATION = False
|
58 |
+
self.dalle_model = "dall-e-3"
|
59 |
+
self.image_size_dalle3 = "1792x1024"
|
60 |
self.video_frame_size = (1280, 720)
|
61 |
+
|
62 |
+
# ElevenLabs Client
|
63 |
+
self.elevenlabs_api_key = None
|
64 |
+
self.USE_ELEVENLABS = False
|
65 |
+
self.elevenlabs_client = None
|
66 |
+
self.elevenlabs_voice_id = "Rachel" # Default, can be name or ID
|
67 |
+
if VoiceSettings: # Check if VoiceSettings was successfully imported
|
68 |
+
self.elevenlabs_voice_settings = VoiceSettings(
|
69 |
+
stability=0.65, similarity_boost=0.75,
|
70 |
+
style=0.1, use_speaker_boost=True
|
71 |
+
)
|
72 |
+
else:
|
73 |
+
self.elevenlabs_voice_settings = None
|
74 |
+
|
75 |
+
self.pexels_api_key = None
|
76 |
+
self.USE_PEXELS = False
|
77 |
|
78 |
def set_openai_api_key(self,k):
|
79 |
+
self.openai_api_key=k
|
80 |
+
self.USE_AI_IMAGE_GENERATION=bool(k)
|
81 |
+
# print(f"DALL-E ({self.dalle_model}) {'Ready' if k else 'Disabled'}.")
|
82 |
+
|
83 |
+
def set_elevenlabs_api_key(self,api_key):
|
84 |
+
self.elevenlabs_api_key=api_key
|
85 |
+
if api_key and ELEVENLABS_CLIENT_IMPORTED and ElevenLabsAPIClient:
|
86 |
+
try:
|
87 |
+
self.elevenlabs_client = ElevenLabsAPIClient(api_key=api_key)
|
88 |
+
# Optional: Test client (e.g., fetch voices) can be added here for robust init
|
89 |
+
# voices_test = self.elevenlabs_client.voices.get_all() # This makes an API call
|
90 |
+
# if voices_test and voices_test.voices: print("ElevenLabs client connected.")
|
91 |
+
self.USE_ELEVENLABS=True
|
92 |
+
# print("ElevenLabs Client Ready.")
|
93 |
+
except Exception as e:
|
94 |
+
print(f"Error initializing ElevenLabs client with API key: {e}. ElevenLabs Disabled.");
|
95 |
+
self.USE_ELEVENLABS=False; self.elevenlabs_client = None
|
96 |
+
else:
|
97 |
+
self.USE_ELEVENLABS=False; self.elevenlabs_client = None
|
98 |
+
# if not ELEVENLABS_CLIENT_IMPORTED or not ElevenLabsAPIClient:
|
99 |
+
# print("ElevenLabs Client class was not imported. ElevenLabs Disabled.") # Already printed at import
|
100 |
+
# else:
|
101 |
+
# print("ElevenLabs API Key not provided. ElevenLabs Disabled.") # Less verbose
|
102 |
+
|
103 |
def set_pexels_api_key(self,k):
|
104 |
self.pexels_api_key=k; self.USE_PEXELS=bool(k)
|
105 |
+
# print(f"Pexels {'Ready' if k else 'Disabled'}.")
|
106 |
+
|
107 |
+
def _get_text_dimensions(self,text_content,font_obj):
|
108 |
+
if not text_content: return 0,self.font_size_pil
|
109 |
try:
|
110 |
+
if hasattr(font_obj,'getbbox'):
|
111 |
+
bbox=font_obj.getbbox(text_content);w=bbox[2]-bbox[0];h=bbox[3]-bbox[1]
|
112 |
+
return w, h if h > 0 else self.font_size_pil
|
113 |
+
elif hasattr(font_obj,'getsize'):
|
114 |
+
w,h=font_obj.getsize(text_content)
|
115 |
+
return w, h if h > 0 else self.font_size_pil
|
116 |
+
else: # Fallback
|
117 |
+
return int(len(text_content)*self.font_size_pil*0.6),int(self.font_size_pil*1.2 if self.font_size_pil*1.2>0 else self.font_size_pil)
|
118 |
+
except Exception: # Generic fallback on error
|
119 |
+
return int(len(text_content)*self.font_size_pil*0.6),int(self.font_size_pil*1.2)
|
120 |
|
121 |
+
def _create_placeholder_image_content(self,text_description,filename,size=(1280,720)):
|
122 |
+
img=Image.new('RGB',size,color=(20,20,40));d=ImageDraw.Draw(img);padding=25;max_w=size[0]-(2*padding);lines=[];
|
123 |
+
if not text_description: text_description="(Placeholder: No prompt text)"
|
124 |
+
words=text_description.split();current_line=""
|
125 |
+
for word in words:
|
126 |
+
test_line=current_line+word+" "
|
127 |
+
if self._get_text_dimensions(test_line,self.font)[0] <= max_w: current_line=test_line
|
128 |
else:
|
129 |
+
if current_line: lines.append(current_line.strip())
|
130 |
+
current_line=word+" "
|
131 |
+
if current_line: lines.append(current_line.strip())
|
132 |
+
if not lines: lines.append("(Text error or too long for placeholder)")
|
133 |
+
|
134 |
+
_,single_line_h=self._get_text_dimensions("Ay",self.font)
|
135 |
+
single_line_h = single_line_h if single_line_h > 0 else self.font_size_pil + 2
|
136 |
+
|
137 |
+
max_lines_to_display=min(len(lines),(size[1]-(2*padding))//(single_line_h+2)) # Max lines based on height
|
138 |
+
|
139 |
+
y_text=padding + (size[1]-(2*padding) - max_lines_to_display*(single_line_h+2))/2.0
|
140 |
+
|
141 |
+
for i in range(max_lines_to_display):
|
142 |
+
line_content=lines[i];line_w,_=self._get_text_dimensions(line_content,self.font);x_text=(size[0]-line_w)/2.0
|
143 |
+
d.text((x_text,y_text),line_content,font=self.font,fill=(200,200,180));y_text+=single_line_h+2
|
144 |
+
if i==6 and max_lines_to_display > 7: # Show ellipsis if more text
|
145 |
+
d.text((x_text,y_text),"...",font=self.font,fill=(200,200,180));break
|
146 |
+
filepath=os.path.join(self.output_dir,filename)
|
147 |
+
try:img.save(filepath);return filepath
|
148 |
+
except Exception as e:print(f"Error saving placeholder image {filepath}: {e}");return None
|
149 |
+
|
150 |
+
def _search_pexels_image(self, query, output_filename_base):
|
151 |
if not self.USE_PEXELS or not self.pexels_api_key: return None
|
152 |
headers = {"Authorization": self.pexels_api_key}
|
153 |
+
params = {"query": query, "per_page": 1, "orientation": "landscape", "size": "large"} # Get only 1 relevant image
|
154 |
+
pexels_filename = output_filename_base.replace(".png", f"_pexels_{random.randint(100,999)}.jpg")
|
155 |
filepath = os.path.join(self.output_dir, pexels_filename)
|
156 |
try:
|
157 |
print(f"Searching Pexels for: '{query}'")
|
158 |
+
effective_query = " ".join(query.split()[:5]) # Use first 5 words for Pexels query
|
159 |
params["query"] = effective_query
|
160 |
+
|
161 |
+
response = requests.get("https://api.pexels.com/v1/search", headers=headers, params=params, timeout=20)
|
162 |
response.raise_for_status(); data = response.json()
|
163 |
+
if data.get("photos") and len(data["photos"]) > 0:
|
164 |
+
photo_url = data["photos"][0]["src"]["large2x"] # High quality
|
165 |
+
image_response = requests.get(photo_url, timeout=60); image_response.raise_for_status()
|
166 |
img_data = Image.open(io.BytesIO(image_response.content))
|
167 |
if img_data.mode != 'RGB': img_data = img_data.convert('RGB')
|
168 |
img_data.save(filepath); print(f"Pexels image saved: {filepath}"); return filepath
|
169 |
+
else: print(f"No photos found on Pexels for query: '{effective_query}'")
|
170 |
+
except Exception as e: print(f"Pexels search/download error for query '{query}': {e}")
|
171 |
return None
|
172 |
|
173 |
def generate_image_visual(self, image_prompt_text, scene_data, scene_identifier_filename):
|
|
|
179 |
print(f"Attempt {attempt+1}: DALL-E ({self.dalle_model}) for: {image_prompt_text[:120]}...")
|
180 |
client = openai.OpenAI(api_key=self.openai_api_key, timeout=90.0)
|
181 |
response = client.images.generate(
|
182 |
+
model=self.dalle_model, prompt=image_prompt_text, n=1,
|
183 |
+
size=self.image_size_dalle3, quality="hd", response_format="url", style="vivid"
|
|
|
|
|
|
|
|
|
|
|
184 |
)
|
185 |
image_url = response.data[0].url
|
186 |
revised_prompt = getattr(response.data[0], 'revised_prompt', None)
|
187 |
+
if revised_prompt: print(f"DALL-E 3 revised_prompt: {revised_prompt[:100]}...")
|
|
|
188 |
|
189 |
image_response = requests.get(image_url, timeout=120)
|
190 |
image_response.raise_for_status()
|
191 |
img_data = Image.open(io.BytesIO(image_response.content))
|
192 |
+
if img_data.mode != 'RGB': img_data = img_data.convert('RGB')
|
|
|
193 |
|
194 |
+
img_data.save(filepath); print(f"AI Image (DALL-E) saved: {filepath}"); return filepath
|
|
|
|
|
|
|
195 |
except openai.RateLimitError as e:
|
196 |
print(f"OpenAI Rate Limit: {e}. Retrying after {5*(attempt+1)}s...")
|
197 |
time.sleep(5 * (attempt + 1))
|
198 |
+
if attempt == max_retries - 1: print("Max retries for RateLimitError."); break
|
199 |
+
else: continue
|
200 |
+
except openai.APIError as e: print(f"OpenAI API Error: {e}"); break
|
201 |
+
except requests.exceptions.RequestException as e: print(f"Requests Error (DALL-E download): {e}"); break
|
202 |
+
except Exception as e: print(f"Generic error (DALL-E gen): {e}"); break
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
203 |
|
204 |
+
print("DALL-E generation failed. Trying Pexels fallback...")
|
205 |
+
pexels_query_text = scene_data.get('pexels_search_query_감독',
|
206 |
+
f"{scene_data.get('emotional_beat','')} {scene_data.get('setting_description','')}")
|
|
|
207 |
pexels_path = self._search_pexels_image(pexels_query_text, scene_identifier_filename)
|
208 |
+
if pexels_path: return pexels_path
|
|
|
209 |
|
210 |
print("Pexels also failed/disabled. Using placeholder.")
|
211 |
return self._create_placeholder_image_content(
|
212 |
f"[AI/Pexels Failed] Original Prompt: {image_prompt_text[:100]}...",
|
213 |
scene_identifier_filename, size=self.video_frame_size
|
214 |
)
|
215 |
+
else:
|
216 |
return self._create_placeholder_image_content(
|
217 |
image_prompt_text, scene_identifier_filename, size=self.video_frame_size
|
218 |
)
|
219 |
|
220 |
def generate_narration_audio(self, text_to_narrate, output_filename="narration_overall.mp3"):
|
221 |
+
if not self.USE_ELEVENLABS or not self.elevenlabs_client or not text_to_narrate:
|
222 |
+
# print("ElevenLabs not enabled, client not initialized, or no text. Skipping audio.") # Less verbose
|
223 |
+
return None
|
224 |
+
|
225 |
audio_filepath = os.path.join(self.output_dir, output_filename)
|
226 |
try:
|
227 |
print(f"Generating ElevenLabs audio (Voice: {self.elevenlabs_voice_id}) for: {text_to_narrate[:70]}...")
|
228 |
+
|
229 |
+
voice_param = self.elevenlabs_voice_id # Default to string ID
|
230 |
+
if Voice and self.elevenlabs_voice_settings: # Check if Voice & VoiceSettings were imported
|
231 |
+
voice_param = Voice(
|
232 |
+
voice_id=self.elevenlabs_voice_id,
|
233 |
+
settings=self.elevenlabs_voice_settings
|
234 |
+
)
|
235 |
+
|
236 |
+
audio_data_iterator = self.elevenlabs_client.generate(
|
237 |
+
text=text_to_narrate,
|
238 |
+
voice=voice_param,
|
239 |
+
model="eleven_multilingual_v2" # Or other models e.g. "eleven_turbo_v2"
|
240 |
+
)
|
241 |
+
|
242 |
+
with open(audio_filepath, "wb") as f:
|
243 |
+
for chunk in audio_data_iterator:
|
244 |
+
if chunk: f.write(chunk)
|
245 |
+
|
246 |
+
print(f"ElevenLabs audio saved: {audio_filepath}")
|
247 |
+
return audio_filepath
|
248 |
+
except AttributeError as ae:
|
249 |
+
print(f"AttributeError with ElevenLabs client (method name like 'generate' might differ): {ae}")
|
250 |
+
except Exception as e:
|
251 |
+
print(f"Error generating ElevenLabs audio: {e}")
|
252 |
return None
|
253 |
|
254 |
def create_video_from_images(self, image_data_list, overall_narration_path=None, output_filename="final_video.mp4", fps=24, duration_per_image=4.5):
|
255 |
+
if not image_data_list: print("No image data for video."); return None
|
256 |
+
# print(f"Creating video from {len(image_data_list)} image sets.") # Less verbose
|
257 |
+
processed_clips = []; narration_audio_clip = None; final_video_clip_obj = None
|
|
|
|
|
258 |
|
259 |
for i, data in enumerate(image_data_list):
|
260 |
img_path, scene_num, key_action = data.get('path'), data.get('scene_num', i+1), data.get('key_action', '')
|
|
|
264 |
if pil_img.mode != 'RGB': pil_img = pil_img.convert('RGB')
|
265 |
img_copy = pil_img.copy()
|
266 |
img_copy.thumbnail(self.video_frame_size, Image.Resampling.LANCZOS)
|
267 |
+
canvas = Image.new('RGB', self.video_frame_size, (random.randint(0,10), random.randint(0,10), random.randint(0,10)))
|
268 |
xo, yo = (self.video_frame_size[0]-img_copy.width)//2, (self.video_frame_size[1]-img_copy.height)//2
|
269 |
canvas.paste(img_copy, (xo,yo))
|
270 |
frame_np = np.array(canvas)
|
271 |
img_clip = ImageClip(frame_np).set_duration(duration_per_image)
|
272 |
+
end_scale = random.uniform(1.05, 1.12)
|
273 |
img_clip = img_clip.fx(vfx.resize, lambda t: 1 + (end_scale - 1) * (t / duration_per_image))
|
274 |
img_clip = img_clip.set_position('center')
|
275 |
if key_action:
|
|
|
281 |
final_scene_clip = CompositeVideoClip([img_clip, txt_clip], size=self.video_frame_size)
|
282 |
else: final_scene_clip = img_clip
|
283 |
processed_clips.append(final_scene_clip)
|
284 |
+
except Exception as e: print(f"Error creating video clip for {img_path}: {e}.")
|
285 |
|
286 |
+
if not processed_clips: print("No clips processed for video."); return None
|
287 |
transition = 0.8
|
288 |
final_video_clip_obj = concatenate_videoclips(processed_clips, padding=-transition, method="compose")
|
289 |
if final_video_clip_obj.duration > transition*2:
|
|
|
292 |
if overall_narration_path and os.path.exists(overall_narration_path):
|
293 |
try:
|
294 |
narration_audio_clip = AudioFileClip(overall_narration_path)
|
295 |
+
current_video_duration = final_video_clip_obj.duration
|
296 |
+
# If narration is shorter, trim video. If narration is longer, audio will be cut by video duration.
|
297 |
+
if narration_audio_clip.duration < current_video_duration:
|
298 |
final_video_clip_obj = final_video_clip_obj.subclip(0, narration_audio_clip.duration)
|
299 |
+
|
300 |
+
final_video_clip_obj = final_video_clip_obj.set_audio(narration_audio_clip)
|
301 |
+
print("Overall narration added to video.")
|
302 |
+
except Exception as e: print(f"Error adding overall narration: {e}.")
|
303 |
|
304 |
output_path = os.path.join(self.output_dir, output_filename)
|
305 |
try:
|
306 |
+
print(f"Writing final video to: {output_path}")
|
307 |
+
final_video_clip_obj.write_videofile(output_path, fps=fps, codec='libx264', preset='medium',
|
308 |
audio_codec='aac',
|
309 |
temp_audiofile=os.path.join(self.output_dir, f'temp-audio-{os.urandom(4).hex()}.m4a'),
|
310 |
+
remove_temp=True, threads=os.cpu_count() or 2, logger='bar', bitrate="5000k") # Consider 'medium' preset
|
311 |
+
print(f"Video successfully created: {output_path}"); return output_path
|
312 |
+
except Exception as e: print(f"Error writing video file: {e}"); return None
|
313 |
finally:
|
314 |
+
for c_item in processed_clips:
|
315 |
+
if hasattr(c_item, 'close'): c_item.close()
|
316 |
if narration_audio_clip and hasattr(narration_audio_clip, 'close'): narration_audio_clip.close()
|
317 |
if final_video_clip_obj and hasattr(final_video_clip_obj, 'close'): final_video_clip_obj.close()
|