Update core/visual_engine.py
Browse files- core/visual_engine.py +202 -393
core/visual_engine.py
CHANGED
@@ -1,376 +1,117 @@
|
|
1 |
# core/visual_engine.py
|
2 |
-
|
3 |
-
|
4 |
-
import mimetypes # For Data URI
|
5 |
-
import numpy as np
|
6 |
-
import os
|
7 |
-
import openai # OpenAI v1.x.x+
|
8 |
-
import requests
|
9 |
-
import io
|
10 |
-
import time
|
11 |
-
import random
|
12 |
-
import logging
|
13 |
-
|
14 |
-
# --- MoviePy Imports ---
|
15 |
-
from moviepy.editor import (ImageClip, VideoFileClip, concatenate_videoclips, TextClip,
|
16 |
-
CompositeVideoClip, AudioFileClip)
|
17 |
-
import moviepy.video.fx.all as vfx
|
18 |
-
|
19 |
-
# --- MONKEY PATCH for Pillow/MoviePy compatibility ---
|
20 |
-
try:
|
21 |
-
if hasattr(Image, 'Resampling') and hasattr(Image.Resampling, 'LANCZOS'): # Pillow 9+
|
22 |
-
if not hasattr(Image, 'ANTIALIAS'): Image.ANTIALIAS = Image.Resampling.LANCZOS
|
23 |
-
elif hasattr(Image, 'LANCZOS'): # Pillow 8
|
24 |
-
if not hasattr(Image, 'ANTIALIAS'): Image.ANTIALIAS = Image.LANCZOS
|
25 |
-
elif not hasattr(Image, 'ANTIALIAS'):
|
26 |
-
print("WARNING: Pillow version lacks common Resampling attributes or ANTIALIAS. MoviePy effects might fail or look different.")
|
27 |
-
except Exception as e_monkey_patch:
|
28 |
-
print(f"WARNING: An unexpected error occurred during Pillow ANTIALIAS monkey-patch: {e_monkey_patch}")
|
29 |
-
|
30 |
-
logger = logging.getLogger(__name__)
|
31 |
-
# logger.setLevel(logging.DEBUG) # Uncomment for verbose debugging during development
|
32 |
-
|
33 |
-
# --- External Service Client Imports ---
|
34 |
-
ELEVENLABS_CLIENT_IMPORTED = False; ElevenLabsAPIClient = None; Voice = None; VoiceSettings = None
|
35 |
-
try:
|
36 |
-
from elevenlabs.client import ElevenLabs as ImportedElevenLabsClient
|
37 |
-
from elevenlabs import Voice as ImportedVoice, VoiceSettings as ImportedVoiceSettings
|
38 |
-
ElevenLabsAPIClient = ImportedElevenLabsClient; Voice = ImportedVoice; VoiceSettings = ImportedVoiceSettings
|
39 |
-
ELEVENLABS_CLIENT_IMPORTED = True; logger.info("ElevenLabs client components (SDK v1.x.x pattern) imported successfully.")
|
40 |
-
except ImportError: logger.warning("ElevenLabs SDK not found (expected 'pip install elevenlabs>=1.0.0'). Audio generation will be disabled.")
|
41 |
-
except Exception as e_eleven_import_general: logger.warning(f"General error importing ElevenLabs client components: {e_eleven_import_general}. Audio generation disabled.")
|
42 |
-
|
43 |
-
RUNWAYML_SDK_IMPORTED = False; RunwayMLAPIClientClass = None
|
44 |
-
try:
|
45 |
-
from runwayml import RunwayML as ImportedRunwayMLAPIClientClass
|
46 |
-
RunwayMLAPIClientClass = ImportedRunwayMLAPIClientClass; RUNWAYML_SDK_IMPORTED = True
|
47 |
-
logger.info("RunwayML SDK (runwayml) imported successfully.")
|
48 |
-
except ImportError: logger.warning("RunwayML SDK not found (pip install runwayml). RunwayML video generation will be disabled.")
|
49 |
-
except Exception as e_runway_sdk_import_general: logger.warning(f"General error importing RunwayML SDK: {e_runway_sdk_import_general}. RunwayML features disabled.")
|
50 |
-
|
51 |
|
52 |
class VisualEngine:
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
self.resolved_font_path_pil = next((p for p in font_paths if os.path.exists(p)), None)
|
62 |
-
self.active_font_pil = ImageFont.load_default(); self.active_font_size_pil = self.DEFAULT_FONT_SIZE_PIL; self.active_moviepy_font_name = self.DEFAULT_MOVIEPY_FONT
|
63 |
-
if self.resolved_font_path_pil:
|
64 |
-
try: self.active_font_pil = ImageFont.truetype(self.resolved_font_path_pil, self.PREFERRED_FONT_SIZE_PIL); self.active_font_size_pil = self.PREFERRED_FONT_SIZE_PIL; logger.info(f"Pillow font: {self.resolved_font_path_pil} sz {self.active_font_size_pil}."); self.active_moviepy_font_name = 'DejaVu-Sans-Bold' if "dejavu" in self.resolved_font_path_pil.lower() else ('Liberation-Sans-Bold' if "liberation" in self.resolved_font_path_pil.lower() else self.DEFAULT_MOVIEPY_FONT)
|
65 |
-
except IOError as e_font: logger.error(f"Pillow font IOError '{self.resolved_font_path_pil}': {e_font}. Default.")
|
66 |
-
else: logger.warning("Preferred Pillow font not found. Default.")
|
67 |
-
self.openai_api_key = None; self.USE_AI_IMAGE_GENERATION = False; self.dalle_model = "dall-e-3"; self.image_size_dalle3 = "1792x1024"
|
68 |
-
self.video_frame_size = (1280, 720)
|
69 |
-
self.elevenlabs_api_key = None; self.USE_ELEVENLABS = False; self.elevenlabs_client_instance = None; self.elevenlabs_voice_id = default_elevenlabs_voice_id
|
70 |
-
if VoiceSettings and ELEVENLABS_CLIENT_IMPORTED: self.elevenlabs_voice_settings_obj = VoiceSettings(stability=0.60, similarity_boost=0.80, style=0.15, use_speaker_boost=True)
|
71 |
-
else: self.elevenlabs_voice_settings_obj = None
|
72 |
-
self.pexels_api_key = None; self.USE_PEXELS = False
|
73 |
-
self.runway_api_key = None; self.USE_RUNWAYML = False; self.runway_ml_sdk_client_instance = None
|
74 |
-
if RUNWAYML_SDK_IMPORTED and RunwayMLAPIClientClass and os.getenv("RUNWAYML_API_SECRET"):
|
75 |
-
try: self.runway_ml_sdk_client_instance = RunwayMLAPIClientClass(); self.USE_RUNWAYML = True; logger.info("RunwayML Client init from env var at startup.")
|
76 |
-
except Exception as e_rwy_init: logger.error(f"Initial RunwayML client init failed: {e_rwy_init}"); self.USE_RUNWAYML = False
|
77 |
-
logger.info("VisualEngine initialized.")
|
78 |
-
|
79 |
-
def set_openai_api_key(self, api_key_value): self.openai_api_key = api_key_value; self.USE_AI_IMAGE_GENERATION = bool(api_key_value); logger.info(f"DALL-E status: {'Ready' if self.USE_AI_IMAGE_GENERATION else 'Disabled'}")
|
80 |
-
def set_elevenlabs_api_key(self, api_key_value, voice_id_from_secret=None):
|
81 |
-
self.elevenlabs_api_key = api_key_value
|
82 |
-
if voice_id_from_secret: self.elevenlabs_voice_id = voice_id_from_secret; logger.info(f"11L Voice ID updated to: {self.elevenlabs_voice_id} via set_elevenlabs_api_key.")
|
83 |
-
if api_key_value and ELEVENLABS_CLIENT_IMPORTED and ElevenLabsAPIClient:
|
84 |
-
try: self.elevenlabs_client_instance = ElevenLabsAPIClient(api_key=api_key_value); self.USE_ELEVENLABS = bool(self.elevenlabs_client_instance); logger.info(f"11L Client: {'Ready' if self.USE_ELEVENLABS else 'Failed'} (Voice: {self.elevenlabs_voice_id})")
|
85 |
-
except Exception as e_11l_setkey_init: logger.error(f"11L client init error: {e_11l_setkey_init}. Disabled.", exc_info=True); self.USE_ELEVENLABS=False; self.elevenlabs_client_instance=None
|
86 |
-
else: self.USE_ELEVENLABS = False; logger.info(f"11L Disabled (key/SDK).")
|
87 |
-
def set_pexels_api_key(self, api_key_value): self.pexels_api_key = api_key_value; self.USE_PEXELS = bool(api_key_value); logger.info(f"Pexels status: {'Ready' if self.USE_PEXELS else 'Disabled'}")
|
88 |
-
def set_runway_api_key(self, api_key_value):
|
89 |
-
self.runway_api_key = api_key_value
|
90 |
-
if api_key_value:
|
91 |
-
if RUNWAYML_SDK_IMPORTED and RunwayMLAPIClientClass:
|
92 |
-
if not self.runway_ml_sdk_client_instance:
|
93 |
-
try:
|
94 |
-
original_env_secret = os.getenv("RUNWAYML_API_SECRET")
|
95 |
-
if not original_env_secret: os.environ["RUNWAYML_API_SECRET"] = api_key_value; logger.info("Temp set RUNWAYML_API_SECRET for SDK.")
|
96 |
-
self.runway_ml_sdk_client_instance = RunwayMLAPIClientClass(); self.USE_RUNWAYML = True; logger.info("RunwayML Client init via set_runway_api_key.")
|
97 |
-
if not original_env_secret: del os.environ["RUNWAYML_API_SECRET"]; logger.info("Cleared temp RUNWAYML_API_SECRET.")
|
98 |
-
except Exception as e_runway_setkey_init: logger.error(f"RunwayML Client init in set_runway_api_key fail: {e_runway_setkey_init}", exc_info=True); self.USE_RUNWAYML=False;self.runway_ml_sdk_client_instance=None
|
99 |
-
else: self.USE_RUNWAYML = True; logger.info("RunwayML Client already init.")
|
100 |
-
else: logger.warning("RunwayML SDK not imported. Service disabled."); self.USE_RUNWAYML = False
|
101 |
-
else: self.USE_RUNWAYML = False; self.runway_ml_sdk_client_instance = None; logger.info("RunwayML Disabled (no API key).")
|
102 |
-
|
103 |
-
def _image_to_data_uri(self, image_path):
|
104 |
-
# <<< CORRECTED METHOD >>>
|
105 |
-
try:
|
106 |
-
mime_type, _ = mimetypes.guess_type(image_path)
|
107 |
-
if not mime_type:
|
108 |
-
ext = os.path.splitext(image_path)[1].lower()
|
109 |
-
mime_map = {".png": "image/png", ".jpg": "image/jpeg", ".jpeg": "image/jpeg", ".webp": "image/webp"}
|
110 |
-
mime_type = mime_map.get(ext, "application/octet-stream")
|
111 |
-
if mime_type == "application/octet-stream":
|
112 |
-
logger.warning(f"Could not determine MIME type for {image_path} from extension '{ext}', using default {mime_type}.")
|
113 |
-
|
114 |
-
with open(image_path, "rb") as image_file_handle:
|
115 |
-
image_binary_data = image_file_handle.read()
|
116 |
-
|
117 |
-
encoded_base64_string = base64.b64encode(image_binary_data).decode('utf-8')
|
118 |
-
|
119 |
-
data_uri_string = f"data:{mime_type};base64,{encoded_base64_string}"
|
120 |
-
logger.debug(f"Generated data URI for {os.path.basename(image_path)} (MIME: {mime_type}). Data URI starts with: {data_uri_string[:100]}...")
|
121 |
-
return data_uri_string
|
122 |
-
except FileNotFoundError:
|
123 |
-
logger.error(f"Image file not found at path: '{image_path}' when trying to create data URI.")
|
124 |
-
return None
|
125 |
-
except Exception as e_data_uri_conversion:
|
126 |
-
logger.error(f"Error converting image '{image_path}' to data URI: {e_data_uri_conversion}", exc_info=True)
|
127 |
-
return None
|
128 |
-
|
129 |
-
def _map_resolution_to_runway_ratio(self, width, height):
|
130 |
-
ratio_str=f"{width}:{height}";supported_ratios_gen4=["1280:720","720:1280","1104:832","832:1104","960:960","1584:672"];
|
131 |
-
if ratio_str in supported_ratios_gen4:return ratio_str
|
132 |
-
logger.warning(f"Res {ratio_str} not in Gen-4 list. Default 1280:720.");return "1280:720"
|
133 |
-
|
134 |
-
def _get_text_dimensions(self, text_content, font_object_pil):
|
135 |
-
dch=getattr(font_object_pil,'size',self.active_font_size_pil);
|
136 |
-
if not text_content:return 0,dch
|
137 |
-
try:
|
138 |
-
if hasattr(font_object_pil,'getbbox'):bb=font_object_pil.getbbox(text_content);w=bb[2]-bb[0];h=bb[3]-bb[1];return w,h if h>0 else dch
|
139 |
-
elif hasattr(font_object_pil,'getsize'):w,h=font_object_pil.getsize(text_content);return w,h if h>0 else dch
|
140 |
-
else:return int(len(text_content)*dch*0.6),int(dch*1.2)
|
141 |
-
except Exception as e_getdim_inner:logger.warning(f"Error in _get_text_dimensions:{e_getdim_inner}");return int(len(text_content)*self.active_font_size_pil*0.6),int(self.active_font_size_pil*1.2)
|
142 |
-
|
143 |
-
def _create_placeholder_image_content(self,text_description,filename,size=None):
|
144 |
-
# (Corrected version from previous response)
|
145 |
-
if size is None: size = self.video_frame_size
|
146 |
-
img = Image.new('RGB', size, color=(20, 20, 40)); d = ImageDraw.Draw(img); padding = 25
|
147 |
-
max_w = size[0] - (2 * padding); lines_for_placeholder = []
|
148 |
-
if not text_description: text_description = "(Placeholder Image)"
|
149 |
-
words_list = text_description.split(); current_line_buffer = ""
|
150 |
-
for word_idx, word_item in enumerate(words_list):
|
151 |
-
prospective_addition = word_item + (" " if word_idx < len(words_list) - 1 else "")
|
152 |
-
test_line_candidate = current_line_buffer + prospective_addition
|
153 |
-
current_w_text, _ = self._get_text_dimensions(test_line_candidate, self.active_font_pil)
|
154 |
-
if current_w_text == 0 and test_line_candidate.strip(): current_w_text = len(test_line_candidate) * (self.active_font_size_pil * 0.6)
|
155 |
-
if current_w_text <= max_w: current_line_buffer = test_line_candidate
|
156 |
-
else:
|
157 |
-
if current_line_buffer.strip(): lines_for_placeholder.append(current_line_buffer.strip())
|
158 |
-
current_line_buffer = prospective_addition
|
159 |
-
if current_line_buffer.strip(): lines_for_placeholder.append(current_line_buffer.strip())
|
160 |
-
if not lines_for_placeholder and text_description:
|
161 |
-
avg_char_w_est, _ = self._get_text_dimensions("W", self.active_font_pil); avg_char_w_est = avg_char_w_est or (self.active_font_size_pil * 0.6)
|
162 |
-
chars_per_line_est = int(max_w / avg_char_w_est) if avg_char_w_est > 0 else 20
|
163 |
-
lines_for_placeholder.append(text_description[:chars_per_line_est] + ("..." if len(text_description) > chars_per_line_est else ""))
|
164 |
-
elif not lines_for_placeholder: lines_for_placeholder.append("(Placeholder Error)")
|
165 |
-
_, single_h = self._get_text_dimensions("Ay", self.active_font_pil); single_h = single_h if single_h > 0 else self.active_font_size_pil + 2
|
166 |
-
max_l = min(len(lines_for_placeholder), (size[1] - (2 * padding)) // (single_h + 2)) if single_h > 0 else 1; max_l = max(1, max_l)
|
167 |
-
y_p = padding + (size[1] - (2 * padding) - max_l * (single_h + 2)) / 2.0
|
168 |
-
for i_line in range(max_l):
|
169 |
-
line_txt_content = lines_for_placeholder[i_line]; line_w_val, _ = self._get_text_dimensions(line_txt_content, self.active_font_pil)
|
170 |
-
if line_w_val == 0 and line_txt_content.strip(): line_w_val = len(line_txt_content) * (self.active_font_size_pil * 0.6)
|
171 |
-
x_p = (size[0] - line_w_val) / 2.0
|
172 |
-
try: d.text((x_p, y_p), line_txt_content, font=self.active_font_pil, fill=(200, 200, 180))
|
173 |
-
except Exception as e_draw: logger.error(f"Pillow d.text error: {e_draw} for '{line_txt_content}'")
|
174 |
-
y_p += single_h + 2
|
175 |
-
if i_line == 6 and max_l > 7:
|
176 |
-
try: d.text((x_p, y_p), "...", font=self.active_font_pil, fill=(200, 200, 180))
|
177 |
-
except Exception as e_elip: logger.error(f"Pillow d.text ellipsis error: {e_elip}"); break
|
178 |
-
filepath_placeholder = os.path.join(self.output_dir, filename)
|
179 |
-
try: img.save(filepath_placeholder); return filepath_placeholder
|
180 |
-
except Exception as e_save: logger.error(f"Saving placeholder image '{filepath_placeholder}' error: {e_save}", exc_info=True); return None
|
181 |
-
|
182 |
-
def _search_pexels_image(self, query_str, output_fn_base):
|
183 |
-
# (Corrected version from previous response)
|
184 |
-
if not self.USE_PEXELS or not self.pexels_api_key: return None
|
185 |
-
http_headers = {"Authorization": self.pexels_api_key}
|
186 |
-
http_params = {"query": query_str, "per_page": 1, "orientation": "landscape", "size": "large2x"}
|
187 |
-
base_name_px, _ = os.path.splitext(output_fn_base)
|
188 |
-
pexels_fn_str = base_name_px + f"_pexels_{random.randint(1000,9999)}.jpg"
|
189 |
-
file_path_px = os.path.join(self.output_dir, pexels_fn_str)
|
190 |
-
try:
|
191 |
-
logger.info(f"Pexels: Searching for '{query_str}'")
|
192 |
-
eff_query_px = " ".join(query_str.split()[:5])
|
193 |
-
http_params["query"] = eff_query_px
|
194 |
-
response_px = requests.get("https://api.pexels.com/v1/search", headers=http_headers, params=http_params, timeout=20)
|
195 |
-
response_px.raise_for_status()
|
196 |
-
data_px = response_px.json()
|
197 |
-
if data_px.get("photos") and len(data_px["photos"]) > 0:
|
198 |
-
photo_details_px = data_px["photos"][0]
|
199 |
-
photo_url_px = photo_details_px.get("src", {}).get("large2x")
|
200 |
-
if not photo_url_px: logger.warning(f"Pexels: 'large2x' URL missing for '{eff_query_px}'. Details: {photo_details_px}"); return None
|
201 |
-
image_response_px = requests.get(photo_url_px, timeout=60); image_response_px.raise_for_status()
|
202 |
-
img_pil_data_px = Image.open(io.BytesIO(image_response_px.content))
|
203 |
-
if img_pil_data_px.mode != 'RGB': img_pil_data_px = img_pil_data_px.convert('RGB')
|
204 |
-
img_pil_data_px.save(file_path_px); logger.info(f"Pexels: Image saved to {file_path_px}"); return file_path_px
|
205 |
-
else: logger.info(f"Pexels: No photos for '{eff_query_px}'."); return None
|
206 |
-
except requests.exceptions.RequestException as e_req_px: logger.error(f"Pexels: RequestException for '{query_str}': {e_req_px}", exc_info=False); return None
|
207 |
-
except Exception as e_px_gen: logger.error(f"Pexels: General error for '{query_str}': {e_px_gen}", exc_info=True); return None
|
208 |
-
|
209 |
-
def _generate_video_clip_with_runwayml(self, text_prompt_for_motion, input_image_path, scene_identifier_filename_base, target_duration_seconds=5):
|
210 |
-
# (Updated RunwayML integration from before)
|
211 |
-
if not self.USE_RUNWAYML or not self.runway_ml_sdk_client_instance: logger.warning("RunwayML not enabled/client not init. Skip video."); return None
|
212 |
-
if not input_image_path or not os.path.exists(input_image_path): logger.error(f"Runway Gen-4 needs input image. Path invalid: {input_image_path}"); return None
|
213 |
-
image_data_uri_str = self._image_to_data_uri(input_image_path)
|
214 |
-
if not image_data_uri_str: return None
|
215 |
-
runway_dur = 10 if target_duration_seconds >= 8 else 5
|
216 |
-
runway_ratio = self._map_resolution_to_runway_ratio(self.video_frame_size[0], self.video_frame_size[1])
|
217 |
-
base_name_for_runway_vid, _ = os.path.splitext(scene_identifier_filename_base); output_vid_fn = base_name_for_runway_vid + f"_runway_gen4_d{runway_dur}s.mp4"
|
218 |
-
output_vid_fp = os.path.join(self.output_dir, output_vid_fn)
|
219 |
-
logger.info(f"Runway Gen-4 task: motion='{text_prompt_for_motion[:100]}...', img='{os.path.basename(input_image_path)}', dur={runway_dur}s, ratio='{runway_ratio}'")
|
220 |
-
try:
|
221 |
-
task_submitted_runway = self.runway_ml_sdk_client_instance.image_to_video.create(model='gen4_turbo', prompt_image=image_data_uri_str, prompt_text=text_prompt_for_motion, duration=runway_dur, ratio=runway_ratio)
|
222 |
-
task_id_runway = task_submitted_runway.id; logger.info(f"Runway Gen-4 task ID: {task_id_runway}. Polling...")
|
223 |
-
poll_sec=10; max_poll_count=36; poll_start_time = time.time()
|
224 |
-
while time.time() - poll_start_time < max_poll_count * poll_sec:
|
225 |
-
time.sleep(poll_sec); task_details_runway = self.runway_ml_sdk_client_instance.tasks.retrieve(id=task_id_runway)
|
226 |
-
logger.info(f"Runway task {task_id_runway} status: {task_details_runway.status}")
|
227 |
-
if task_details_runway.status == 'SUCCEEDED':
|
228 |
-
output_url_runway = getattr(getattr(task_details_runway,'output',None),'url',None) or \
|
229 |
-
(getattr(task_details_runway,'artifacts',None) and task_details_runway.artifacts and hasattr(task_details_runway.artifacts[0],'url')and task_details_runway.artifacts[0].url) or \
|
230 |
-
(getattr(task_details_runway,'artifacts',None) and task_details_runway.artifacts and hasattr(task_details_runway.artifacts[0],'download_url')and task_details_runway.artifacts[0].download_url)
|
231 |
-
if not output_url_runway: logger.error(f"Runway task {task_id_runway} SUCCEEDED, but no output URL. Details: {vars(task_details_runway) if hasattr(task_details_runway,'__dict__') else task_details_runway}"); return None
|
232 |
-
logger.info(f"Runway task {task_id_runway} SUCCEEDED. Downloading: {output_url_runway}")
|
233 |
-
video_resp_get = requests.get(output_url_runway, stream=True, timeout=300); video_resp_get.raise_for_status()
|
234 |
-
with open(output_vid_fp,'wb') as f_vid:
|
235 |
-
for chunk_data in video_resp_get.iter_content(chunk_size=8192): f_vid.write(chunk_data)
|
236 |
-
logger.info(f"Runway Gen-4 video saved: {output_vid_fp}"); return output_vid_fp
|
237 |
-
elif task_details_runway.status in ['FAILED','ABORTED','ERROR']:
|
238 |
-
err_msg_runway = getattr(task_details_runway,'error_message',None) or getattr(getattr(task_details_runway,'output',None),'error',"Unknown Runway error.")
|
239 |
-
logger.error(f"Runway task {task_id_runway} status: {task_details_runway.status}. Error: {err_msg_runway}"); return None
|
240 |
-
logger.warning(f"Runway task {task_id_runway} timed out."); return None
|
241 |
-
except AttributeError as ae_sdk: logger.error(f"RunwayML SDK AttrError: {ae_sdk}. SDK/methods changed?", exc_info=True); return None
|
242 |
-
except Exception as e_runway_gen: logger.error(f"Runway Gen-4 API error: {e_runway_gen}", exc_info=True); return None
|
243 |
-
|
244 |
-
def _create_placeholder_video_content(self, text_desc_ph, filename_ph, duration_ph=4, size_ph=None):
|
245 |
-
# (Corrected from previous response)
|
246 |
-
if size_ph is None: size_ph = self.video_frame_size
|
247 |
-
filepath_ph = os.path.join(self.output_dir, filename_ph)
|
248 |
-
text_clip_ph = None
|
249 |
-
try:
|
250 |
-
text_clip_ph = TextClip(text_desc_ph, fontsize=50, color='white', font=self.video_overlay_font,
|
251 |
-
bg_color='black', size=size_ph, method='caption').set_duration(duration_ph)
|
252 |
-
text_clip_ph.write_videofile(filepath_ph, fps=24, codec='libx264', preset='ultrafast', logger=None, threads=2)
|
253 |
-
logger.info(f"Generic placeholder video created: {filepath_ph}")
|
254 |
-
return filepath_ph
|
255 |
-
except Exception as e_ph_vid:
|
256 |
-
logger.error(f"Failed to create generic placeholder video '{filepath_ph}': {e_ph_vid}", exc_info=True)
|
257 |
-
return None
|
258 |
-
finally:
|
259 |
-
if text_clip_ph and hasattr(text_clip_ph, 'close'):
|
260 |
-
try: text_clip_ph.close()
|
261 |
-
except Exception as e_cl_phv: logger.warning(f"Ignoring error closing placeholder TextClip: {e_cl_phv}")
|
262 |
-
|
263 |
-
def generate_scene_asset(self, image_generation_prompt_text, motion_prompt_text_for_video,
|
264 |
-
scene_data_dict, scene_identifier_fn_base,
|
265 |
-
generate_as_video_clip_flag=False, runway_target_dur_val=5):
|
266 |
-
# (Corrected DALL-E loop from previous response)
|
267 |
-
base_name_asset, _ = os.path.splitext(scene_identifier_fn_base)
|
268 |
-
asset_info_result = {'path': None, 'type': 'none', 'error': True, 'prompt_used': image_generation_prompt_text, 'error_message': 'Asset generation init failed'}
|
269 |
-
path_for_input_image_runway = None
|
270 |
-
fn_for_base_image = base_name_asset + ("_base_for_video.png" if generate_as_video_clip_flag else ".png")
|
271 |
-
fp_for_base_image = os.path.join(self.output_dir, fn_for_base_image)
|
272 |
-
if self.USE_AI_IMAGE_GENERATION and self.openai_api_key:
|
273 |
-
max_r_dalle, attempt_count_dalle = 2,0;
|
274 |
-
for att_n_dalle in range(max_r_dalle):
|
275 |
-
attempt_count_dalle = att_n_dalle + 1
|
276 |
-
try:
|
277 |
-
logger.info(f"Att {attempt_count_dalle} DALL-E (base img): {image_generation_prompt_text[:70]}..."); oai_cl = openai.OpenAI(api_key=self.openai_api_key,timeout=90.0); oai_r = oai_cl.images.generate(model=self.dalle_model,prompt=image_generation_prompt_text,n=1,size=self.image_size_dalle3,quality="hd",response_format="url",style="vivid"); oai_iu = oai_r.data[0].url; oai_rp = getattr(oai_r.data[0],'revised_prompt',None);
|
278 |
-
if oai_rp: logger.info(f"DALL-E revised: {oai_rp[:70]}...")
|
279 |
-
oai_ir = requests.get(oai_iu,timeout=120); oai_ir.raise_for_status(); oai_id = Image.open(io.BytesIO(oai_ir.content));
|
280 |
-
if oai_id.mode!='RGB': oai_id=oai_id.convert('RGB')
|
281 |
-
oai_id.save(fp_for_base_image); logger.info(f"DALL-E base img saved: {fp_for_base_image}"); path_for_input_image_runway=fp_for_base_image; asset_info_result={'path':fp_for_base_image,'type':'image','error':False,'prompt_used':image_generation_prompt_text,'revised_prompt':oai_rp}; break
|
282 |
-
except openai.RateLimitError as e_oai_rl: logger.warning(f"OpenAI RateLimit Att {attempt_count_dalle}:{e_oai_rl}.Retry...");time.sleep(5*attempt_count_dalle);asset_info_result['error_message']=str(e_oai_rl)
|
283 |
-
except openai.APIError as e_oai_api: logger.error(f"OpenAI APIError Att {attempt_count_dalle}:{e_oai_api}");asset_info_result['error_message']=str(e_oai_api);break
|
284 |
-
except requests.exceptions.RequestException as e_oai_req: logger.error(f"Requests Err DALL-E Att {attempt_count_dalle}:{e_oai_req}");asset_info_result['error_message']=str(e_oai_req);break
|
285 |
-
except Exception as e_oai_gen: logger.error(f"General DALL-E Err Att {attempt_count_dalle}:{e_oai_gen}",exc_info=True);asset_info_result['error_message']=str(e_oai_gen);break
|
286 |
-
if asset_info_result['error']: logger.warning(f"DALL-E failed after {attempt_count_dalle} attempts for base img.")
|
287 |
-
if asset_info_result['error'] and self.USE_PEXELS:
|
288 |
-
logger.info("Trying Pexels for base img.");px_qt=scene_data_dict.get('pexels_search_query_감독',f"{scene_data_dict.get('emotional_beat','')} {scene_data_dict.get('setting_description','')}");px_pp=self._search_pexels_image(px_qt,fn_for_base_image);
|
289 |
-
if px_pp:path_for_input_image_runway=px_pp;asset_info_result={'path':px_pp,'type':'image','error':False,'prompt_used':f"Pexels:{px_qt}"}
|
290 |
-
else:current_em_px=asset_info_result.get('error_message',"");asset_info_result['error_message']=(current_em_px+" Pexels failed for base.").strip()
|
291 |
-
if asset_info_result['error']:
|
292 |
-
logger.warning("Base img (DALL-E/Pexels) failed. Using placeholder.");ph_ppt=asset_info_result.get('prompt_used',image_generation_prompt_text);php=self._create_placeholder_image_content(f"[Base Placeholder]{ph_ppt[:70]}...",fn_for_base_image);
|
293 |
-
if php:path_for_input_image_runway=php;asset_info_result={'path':php,'type':'image','error':False,'prompt_used':ph_ppt}
|
294 |
-
else:current_em_ph=asset_info_result.get('error_message',"");asset_info_result['error_message']=(current_em_ph+" Base placeholder failed.").strip()
|
295 |
-
if generate_as_video_clip_flag:
|
296 |
-
if not path_for_input_image_runway:logger.error("RunwayML video: base img failed.");asset_info_result['error']=True;asset_info_result['error_message']=(asset_info_result.get('error_message',"")+" Base img miss, Runway abort.").strip();asset_info_result['type']='none';return asset_info_result
|
297 |
-
if self.USE_RUNWAYML:
|
298 |
-
runway_video_p=self._generate_video_clip_with_runwayml(motion_prompt_text_for_video,path_for_input_image_runway,base_name_asset,runway_target_dur_val)
|
299 |
-
if runway_video_p and os.path.exists(runway_video_p):asset_info_result={'path':runway_video_p,'type':'video','error':False,'prompt_used':motion_prompt_text_for_video,'base_image_path':path_for_input_image_runway}
|
300 |
-
else:logger.warning(f"RunwayML video failed for {base_name_asset}. Fallback to base img.");asset_info_result['error']=True;asset_info_result['error_message']=(asset_info_result.get('error_message',"Base img ok.")+" RunwayML video fail; use base img.").strip();asset_info_result['path']=path_for_input_image_runway;asset_info_result['type']='image';asset_info_result['prompt_used']=image_generation_prompt_text
|
301 |
-
else:logger.warning("RunwayML selected but disabled. Use base img.");asset_info_result['error']=True;asset_info_result['error_message']=(asset_info_result.get('error_message',"Base img ok.")+" RunwayML disabled; use base img.").strip();asset_info_result['path']=path_for_input_image_runway;asset_info_result['type']='image';asset_info_result['prompt_used']=image_generation_prompt_text
|
302 |
-
return asset_info_result
|
303 |
-
|
304 |
-
def generate_narration_audio(self, text_to_narrate, output_filename="narration_overall.mp3"):
|
305 |
-
# (Corrected version from previous response)
|
306 |
-
if not self.USE_ELEVENLABS or not self.elevenlabs_client_instance or not text_to_narrate:
|
307 |
-
logger.info("ElevenLabs conditions not met. Skipping audio generation.")
|
308 |
-
return None
|
309 |
-
audio_filepath_narration = os.path.join(self.output_dir, output_filename)
|
310 |
-
try:
|
311 |
-
logger.info(f"Generating ElevenLabs audio (Voice ID: {self.elevenlabs_voice_id}) for text: \"{text_to_narrate[:70]}...\"")
|
312 |
-
audio_stream_method_11l = None
|
313 |
-
if hasattr(self.elevenlabs_client_instance, 'text_to_speech') and hasattr(self.elevenlabs_client_instance.text_to_speech, 'stream'):
|
314 |
-
audio_stream_method_11l = self.elevenlabs_client_instance.text_to_speech.stream; logger.info("Using ElevenLabs SDK method: client.text_to_speech.stream()")
|
315 |
-
elif hasattr(self.elevenlabs_client_instance, 'generate_stream'):
|
316 |
-
audio_stream_method_11l = self.elevenlabs_client_instance.generate_stream; logger.info("Using ElevenLabs SDK method: client.generate_stream()")
|
317 |
-
elif hasattr(self.elevenlabs_client_instance, 'generate'):
|
318 |
-
logger.info("Using ElevenLabs SDK method: client.generate() (non-streaming).")
|
319 |
-
voice_param_11l = str(self.elevenlabs_voice_id)
|
320 |
-
if Voice and self.elevenlabs_voice_settings_obj: voice_param_11l = Voice(voice_id=str(self.elevenlabs_voice_id), settings=self.elevenlabs_voice_settings_obj)
|
321 |
-
audio_bytes_data = self.elevenlabs_client_instance.generate(text=text_to_narrate, voice=voice_param_11l, model="eleven_multilingual_v2")
|
322 |
-
with open(audio_filepath_narration, "wb") as audio_file_out: audio_file_out.write(audio_bytes_data)
|
323 |
-
logger.info(f"ElevenLabs audio (non-streamed) saved successfully to: {audio_filepath_narration}"); return audio_filepath_narration
|
324 |
-
else: logger.error("No recognized audio generation method found on the ElevenLabs client instance."); return None
|
325 |
-
|
326 |
-
if audio_stream_method_11l:
|
327 |
-
params_for_voice_stream = {"voice_id": str(self.elevenlabs_voice_id)}
|
328 |
-
if self.elevenlabs_voice_settings_obj:
|
329 |
-
if hasattr(self.elevenlabs_voice_settings_obj, 'model_dump'): params_for_voice_stream["voice_settings"] = self.elevenlabs_voice_settings_obj.model_dump()
|
330 |
-
elif hasattr(self.elevenlabs_voice_settings_obj, 'dict'): params_for_voice_stream["voice_settings"] = self.elevenlabs_voice_settings_obj.dict()
|
331 |
-
else: params_for_voice_stream["voice_settings"] = self.elevenlabs_voice_settings_obj
|
332 |
-
audio_data_iterator_11l = audio_stream_method_11l(text=text_to_narrate, model_id="eleven_multilingual_v2", **params_for_voice_stream)
|
333 |
-
with open(audio_filepath_narration, "wb") as audio_file_out_stream:
|
334 |
-
for audio_chunk_data in audio_data_iterator_11l:
|
335 |
-
if audio_chunk_data: audio_file_out_stream.write(audio_chunk_data)
|
336 |
-
logger.info(f"ElevenLabs audio (streamed) saved successfully to: {audio_filepath_narration}"); return audio_filepath_narration
|
337 |
-
except AttributeError as ae_11l_sdk: logger.error(f"AttributeError with ElevenLabs SDK client: {ae_11l_sdk}. SDK version/methods might differ.", exc_info=True); return None
|
338 |
-
except Exception as e_11l_general_audio: logger.error(f"General error during ElevenLabs audio generation: {e_11l_general_audio}", exc_info=True); return None
|
339 |
-
|
340 |
def assemble_animatic_from_assets(self, asset_data_list, overall_narration_path=None, output_filename="final_video.mp4", fps=24):
|
341 |
-
# (Keep as in the version with robust image processing, C-contiguous array, debug saves, and pix_fmt)
|
342 |
if not asset_data_list: logger.warning("No assets for animatic."); return None
|
343 |
processed_moviepy_clips_list = []; narration_audio_clip_mvpy = None; final_video_output_clip = None
|
344 |
logger.info(f"Assembling from {len(asset_data_list)} assets. Target Frame: {self.video_frame_size}.")
|
|
|
345 |
for i_asset, asset_info_item_loop in enumerate(asset_data_list):
|
346 |
path_of_asset, type_of_asset, duration_for_scene = asset_info_item_loop.get('path'), asset_info_item_loop.get('type'), asset_info_item_loop.get('duration', 4.5)
|
347 |
num_of_scene, action_in_key = asset_info_item_loop.get('scene_num', i_asset + 1), asset_info_item_loop.get('key_action', '')
|
348 |
logger.info(f"S{num_of_scene}: Path='{path_of_asset}', Type='{type_of_asset}', Dur='{duration_for_scene}'s")
|
|
|
349 |
if not (path_of_asset and os.path.exists(path_of_asset)): logger.warning(f"S{num_of_scene}: Not found '{path_of_asset}'. Skip."); continue
|
350 |
if duration_for_scene <= 0: logger.warning(f"S{num_of_scene}: Invalid duration ({duration_for_scene}s). Skip."); continue
|
|
|
351 |
active_scene_clip = None
|
352 |
try:
|
353 |
if type_of_asset == 'image':
|
354 |
-
|
355 |
-
|
356 |
-
|
357 |
-
|
358 |
-
|
359 |
-
|
360 |
-
|
361 |
-
|
362 |
-
|
363 |
-
logger.debug(f"S{num_of_scene}:
|
364 |
-
|
365 |
-
|
366 |
-
|
367 |
-
|
368 |
-
|
369 |
-
|
370 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
371 |
elif type_of_asset == 'video':
|
|
|
|
|
372 |
source_video_clip_obj=None
|
373 |
try:
|
|
|
374 |
source_video_clip_obj=VideoFileClip(path_of_asset,target_resolution=(self.video_frame_size[1],self.video_frame_size[0])if self.video_frame_size else None, audio=False)
|
375 |
temp_video_clip_obj_loop=source_video_clip_obj
|
376 |
if source_video_clip_obj.duration!=duration_for_scene:
|
@@ -378,51 +119,119 @@ class VisualEngine:
|
|
378 |
else:
|
379 |
if duration_for_scene/source_video_clip_obj.duration > 1.5 and source_video_clip_obj.duration>0.1:temp_video_clip_obj_loop=source_video_clip_obj.loop(duration=duration_for_scene)
|
380 |
else:temp_video_clip_obj_loop=source_video_clip_obj.set_duration(source_video_clip_obj.duration);logger.info(f"S{num_of_scene} Video clip ({source_video_clip_obj.duration:.2f}s) shorter than target ({duration_for_scene:.2f}s).")
|
381 |
-
active_scene_clip=temp_video_clip_obj_loop.set_duration(duration_for_scene)
|
382 |
if active_scene_clip.size!=list(self.video_frame_size):active_scene_clip=active_scene_clip.resize(self.video_frame_size)
|
383 |
-
|
384 |
-
|
385 |
-
|
386 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
387 |
if active_scene_clip and action_in_key:
|
388 |
try:
|
389 |
-
|
390 |
-
|
391 |
-
|
392 |
-
|
393 |
-
|
394 |
-
|
395 |
-
|
396 |
-
|
397 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
398 |
if active_scene_clip and hasattr(active_scene_clip,'close'):
|
399 |
try: active_scene_clip.close()
|
400 |
-
except:
|
401 |
-
|
|
|
|
|
|
|
|
|
|
|
402 |
transition_duration_val=0.75
|
403 |
try:
|
404 |
-
logger.info(f"Concatenating {len(processed_moviepy_clips_list)} clips for final animatic.");
|
405 |
-
if len(processed_moviepy_clips_list)>1:
|
406 |
-
|
407 |
-
|
408 |
-
|
409 |
-
|
410 |
-
|
411 |
-
|
412 |
-
if
|
413 |
-
|
414 |
-
|
415 |
-
|
416 |
-
if
|
417 |
-
|
418 |
-
|
419 |
-
|
420 |
-
|
421 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
422 |
finally:
|
423 |
logger.debug("Closing all MoviePy clips in `assemble_animatic_from_assets` main finally block.")
|
424 |
-
|
425 |
-
|
426 |
-
|
427 |
-
|
428 |
-
|
|
|
|
|
|
|
|
|
|
1 |
# core/visual_engine.py
|
2 |
+
# ... (all imports and class setup as in the previous "expertly crafted" version) ...
|
3 |
+
# ... (methods __init__ through generate_narration_audio also as in the previous full version) ...
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
|
5 |
class VisualEngine:
|
6 |
+
# ... (previous __init__ and set_api_key methods, _image_to_data_uri, _map_resolution_to_runway_ratio,
|
7 |
+
# _get_text_dimensions, _create_placeholder_image_content, _search_pexels_image,
|
8 |
+
# _generate_video_clip_with_runwayml, _create_placeholder_video_content,
|
9 |
+
# generate_scene_asset, generate_narration_audio - KEEP THESE AS THEY WERE in the last full version) ...
|
10 |
+
|
11 |
+
# =========================================================================
|
12 |
+
# ASSEMBLE ANIMATIC - EXTREME DEBUGGING FOR IMAGE ASSETS
|
13 |
+
# =========================================================================
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
def assemble_animatic_from_assets(self, asset_data_list, overall_narration_path=None, output_filename="final_video.mp4", fps=24):
|
|
|
15 |
if not asset_data_list: logger.warning("No assets for animatic."); return None
|
16 |
processed_moviepy_clips_list = []; narration_audio_clip_mvpy = None; final_video_output_clip = None
|
17 |
logger.info(f"Assembling from {len(asset_data_list)} assets. Target Frame: {self.video_frame_size}.")
|
18 |
+
|
19 |
for i_asset, asset_info_item_loop in enumerate(asset_data_list):
|
20 |
path_of_asset, type_of_asset, duration_for_scene = asset_info_item_loop.get('path'), asset_info_item_loop.get('type'), asset_info_item_loop.get('duration', 4.5)
|
21 |
num_of_scene, action_in_key = asset_info_item_loop.get('scene_num', i_asset + 1), asset_info_item_loop.get('key_action', '')
|
22 |
logger.info(f"S{num_of_scene}: Path='{path_of_asset}', Type='{type_of_asset}', Dur='{duration_for_scene}'s")
|
23 |
+
|
24 |
if not (path_of_asset and os.path.exists(path_of_asset)): logger.warning(f"S{num_of_scene}: Not found '{path_of_asset}'. Skip."); continue
|
25 |
if duration_for_scene <= 0: logger.warning(f"S{num_of_scene}: Invalid duration ({duration_for_scene}s). Skip."); continue
|
26 |
+
|
27 |
active_scene_clip = None
|
28 |
try:
|
29 |
if type_of_asset == 'image':
|
30 |
+
logger.info(f"S{num_of_scene}: Processing IMAGE asset: {path_of_asset}")
|
31 |
+
# 0. Load original image
|
32 |
+
pil_img_original = Image.open(path_of_asset)
|
33 |
+
logger.debug(f"S{num_of_scene} (0-Load): Original loaded. Mode:{pil_img_original.mode}, Size:{pil_img_original.size}")
|
34 |
+
pil_img_original.save(os.path.join(self.output_dir,f"debug_0_ORIGINAL_S{num_of_scene}.png"))
|
35 |
+
|
36 |
+
|
37 |
+
# 1. Convert to RGBA for consistent alpha handling (even if it's already RGB)
|
38 |
+
img_rgba_intermediate = pil_img_original.convert('RGBA') if pil_img_original.mode != 'RGBA' else pil_img_original.copy().convert('RGBA') # Ensure copy if already RGBA
|
39 |
+
logger.debug(f"S{num_of_scene} (1-ToRGBA): Converted to RGBA. Mode:{img_rgba_intermediate.mode}, Size:{img_rgba_intermediate.size}")
|
40 |
+
img_rgba_intermediate.save(os.path.join(self.output_dir,f"debug_1_AS_RGBA_S{num_of_scene}.png"))
|
41 |
+
|
42 |
+
# 2. Thumbnail the RGBA image
|
43 |
+
thumbnailed_img_rgba = img_rgba_intermediate.copy() # Work on a copy for thumbnailing
|
44 |
+
resample_filter_pil = Image.Resampling.LANCZOS if hasattr(Image.Resampling,'LANCZOS') else Image.BILINEAR
|
45 |
+
thumbnailed_img_rgba.thumbnail(self.video_frame_size, resample_filter_pil)
|
46 |
+
logger.debug(f"S{num_of_scene} (2-Thumbnail): Thumbnailed RGBA. Mode:{thumbnailed_img_rgba.mode}, Size:{thumbnailed_img_rgba.size}")
|
47 |
+
thumbnailed_img_rgba.save(os.path.join(self.output_dir,f"debug_2_THUMBNAIL_RGBA_S{num_of_scene}.png"))
|
48 |
+
|
49 |
+
# 3. Create a target-sized RGBA canvas (fully transparent for true alpha blending)
|
50 |
+
canvas_for_compositing_rgba = Image.new('RGBA', self.video_frame_size, (0,0,0,0))
|
51 |
+
pos_x_paste = (self.video_frame_size[0] - thumbnailed_img_rgba.width) // 2
|
52 |
+
pos_y_paste = (self.video_frame_size[1] - thumbnailed_img_rgba.height) // 2
|
53 |
+
# Paste the (potentially smaller) thumbnailed RGBA image onto the transparent RGBA canvas, using its own alpha
|
54 |
+
canvas_for_compositing_rgba.paste(thumbnailed_img_rgba, (pos_x_paste, pos_y_paste), thumbnailed_img_rgba)
|
55 |
+
logger.debug(f"S{num_of_scene} (3-PasteOnRGBA): Image pasted onto transparent RGBA canvas. Mode:{canvas_for_compositing_rgba.mode}, Size:{canvas_for_compositing_rgba.size}")
|
56 |
+
canvas_for_compositing_rgba.save(os.path.join(self.output_dir,f"debug_3_COMPOSITED_RGBA_S{num_of_scene}.png"))
|
57 |
+
|
58 |
+
# 4. Create a final RGB image by pasting the composited RGBA canvas onto an opaque background
|
59 |
+
# This flattens all transparency and ensures a 3-channel RGB image for MoviePy.
|
60 |
+
final_rgb_image_for_pil = Image.new("RGB", self.video_frame_size, (5, 5, 15)) # Dark opaque background (e.g., dark blue)
|
61 |
+
# Paste canvas_for_compositing_rgba using its alpha channel as the mask
|
62 |
+
if canvas_for_compositing_rgba.mode == 'RGBA':
|
63 |
+
final_rgb_image_for_pil.paste(canvas_for_compositing_rgba, mask=canvas_for_compositing_rgba.split()[3])
|
64 |
+
else: # Should not happen if step 1 & 3 are correct, but as a fallback
|
65 |
+
final_rgb_image_for_pil.paste(canvas_for_compositing_rgba) # Paste without mask if not RGBA
|
66 |
+
logger.debug(f"S{num_of_scene} (4-ToRGB): Final RGB image created. Mode:{final_rgb_image_for_pil.mode}, Size:{final_rgb_image_for_pil.size}")
|
67 |
+
|
68 |
+
# THIS IS THE CRITICAL DEBUG IMAGE - what does it look like?
|
69 |
+
debug_path_img_pre_numpy = os.path.join(self.output_dir,f"debug_4_PRE_NUMPY_RGB_S{num_of_scene}.png");
|
70 |
+
final_rgb_image_for_pil.save(debug_path_img_pre_numpy);
|
71 |
+
logger.info(f"CRITICAL DEBUG: Saved PRE_NUMPY_RGB_S{num_of_scene} (image fed to NumPy) to {debug_path_img_pre_numpy}")
|
72 |
+
|
73 |
+
# 5. Convert to C-contiguous NumPy array, dtype uint8
|
74 |
+
numpy_frame_arr = np.array(final_rgb_image_for_pil, dtype=np.uint8)
|
75 |
+
if not numpy_frame_arr.flags['C_CONTIGUOUS']:
|
76 |
+
numpy_frame_arr = np.ascontiguousarray(numpy_frame_arr, dtype=np.uint8) # Ensure C-order
|
77 |
+
logger.debug(f"S{num_of_scene} (5-NumPy): Ensured NumPy array is C-contiguous.")
|
78 |
+
|
79 |
+
logger.debug(f"S{num_of_scene} (5-NumPy): Final NumPy array for MoviePy. Shape:{numpy_frame_arr.shape}, DType:{numpy_frame_arr.dtype}, Flags:{numpy_frame_arr.flags}")
|
80 |
+
|
81 |
+
if numpy_frame_arr.size == 0 or numpy_frame_arr.ndim != 3 or numpy_frame_arr.shape[2] != 3:
|
82 |
+
logger.error(f"S{num_of_scene}: Invalid NumPy array shape/size ({numpy_frame_arr.shape}) for ImageClip. Skipping this asset."); continue
|
83 |
+
|
84 |
+
# 6. Create MoviePy ImageClip
|
85 |
+
base_image_clip_mvpy = ImageClip(numpy_frame_arr, transparent=False, ismask=False).set_duration(duration_for_scene)
|
86 |
+
logger.debug(f"S{num_of_scene} (6-ImageClip): Base ImageClip created. Duration: {base_image_clip_mvpy.duration}")
|
87 |
+
|
88 |
+
# 7. DEBUG: Save a frame directly FROM the MoviePy ImageClip object
|
89 |
+
debug_path_moviepy_frame = os.path.join(self.output_dir,f"debug_7_MOVIEPY_FRAME_S{num_of_scene}.png")
|
90 |
+
try:
|
91 |
+
base_image_clip_mvpy.save_frame(debug_path_moviepy_frame, t=min(0.1, base_image_clip_mvpy.duration / 2 if base_image_clip_mvpy.duration > 0 else 0.1)) # Save frame at 0.1s or mid-point
|
92 |
+
logger.info(f"CRITICAL DEBUG: Saved frame FROM MOVIEPY ImageClip for S{num_of_scene} to {debug_path_moviepy_frame}")
|
93 |
+
except Exception as e_save_mvpy_frame:
|
94 |
+
logger.error(f"DEBUG: Error saving frame FROM MOVIEPY ImageClip for S{num_of_scene}: {e_save_mvpy_frame}", exc_info=True)
|
95 |
+
|
96 |
+
# 8. Apply Ken Burns effect (optional, can be commented out for further isolation)
|
97 |
+
fx_image_clip_mvpy = base_image_clip_mvpy
|
98 |
+
try:
|
99 |
+
scale_end_kb_val = random.uniform(1.03, 1.08)
|
100 |
+
if duration_for_scene > 0: # Avoid division by zero
|
101 |
+
fx_image_clip_mvpy = base_image_clip_mvpy.fx(vfx.resize, lambda t_val: 1 + (scale_end_kb_val - 1) * (t_val / duration_for_scene)).set_position('center')
|
102 |
+
logger.debug(f"S{num_of_scene} (8-KenBurns): Ken Burns effect applied.")
|
103 |
+
else:
|
104 |
+
logger.warning(f"S{num_of_scene}: Duration is zero, skipping Ken Burns.")
|
105 |
+
except Exception as e_kb_fx_loop: logger.error(f"S{num_of_scene} Ken Burns effect error: {e_kb_fx_loop}", exc_info=False) # exc_info=False for brevity
|
106 |
+
|
107 |
+
active_scene_clip = fx_image_clip_mvpy
|
108 |
+
|
109 |
elif type_of_asset == 'video':
|
110 |
+
# ... (Video processing logic from the previous full, corrected version) ...
|
111 |
+
# Ensure this part also handles clip closing diligently.
|
112 |
source_video_clip_obj=None
|
113 |
try:
|
114 |
+
logger.debug(f"S{num_of_scene}: Loading VIDEO asset: {path_of_asset}")
|
115 |
source_video_clip_obj=VideoFileClip(path_of_asset,target_resolution=(self.video_frame_size[1],self.video_frame_size[0])if self.video_frame_size else None, audio=False)
|
116 |
temp_video_clip_obj_loop=source_video_clip_obj
|
117 |
if source_video_clip_obj.duration!=duration_for_scene:
|
|
|
119 |
else:
|
120 |
if duration_for_scene/source_video_clip_obj.duration > 1.5 and source_video_clip_obj.duration>0.1:temp_video_clip_obj_loop=source_video_clip_obj.loop(duration=duration_for_scene)
|
121 |
else:temp_video_clip_obj_loop=source_video_clip_obj.set_duration(source_video_clip_obj.duration);logger.info(f"S{num_of_scene} Video clip ({source_video_clip_obj.duration:.2f}s) shorter than target ({duration_for_scene:.2f}s).")
|
122 |
+
active_scene_clip=temp_video_clip_obj_loop.set_duration(duration_for_scene) # Ensure final clip has target duration
|
123 |
if active_scene_clip.size!=list(self.video_frame_size):active_scene_clip=active_scene_clip.resize(self.video_frame_size)
|
124 |
+
logger.debug(f"S{num_of_scene}: Video asset processed. Final duration for scene: {active_scene_clip.duration:.2f}s")
|
125 |
+
except Exception as e_vid_load_loop:logger.error(f"S{num_of_scene} Video load error '{path_of_asset}':{e_vid_load_loop}",exc_info=True);continue # Skip this broken video asset
|
126 |
+
finally: # Close the original source_video_clip_obj if it's different from active_scene_clip
|
127 |
+
if source_video_clip_obj and source_video_clip_obj is not active_scene_clip and hasattr(source_video_clip_obj,'close'):
|
128 |
+
try: source_video_clip_obj.close()
|
129 |
+
except Exception as e_close_src_vid: logger.warning(f"S{num_of_scene}: Error closing source VideoFileClip: {e_close_src_vid}")
|
130 |
+
else:
|
131 |
+
logger.warning(f"S{num_of_scene} Unknown asset type '{type_of_asset}'. Skipping."); continue
|
132 |
+
|
133 |
+
# Add text overlay (common to both image and video assets)
|
134 |
if active_scene_clip and action_in_key:
|
135 |
try:
|
136 |
+
dur_text_overlay_val=min(active_scene_clip.duration-0.5,active_scene_clip.duration*0.8)if active_scene_clip.duration>0.5 else active_scene_clip.duration
|
137 |
+
start_text_overlay_val=0.25 # Start text a bit into the clip
|
138 |
+
if dur_text_overlay_val > 0:
|
139 |
+
text_clip_for_overlay_obj=TextClip(f"Scene {num_of_scene}\n{action_in_key}",fontsize=self.VIDEO_OVERLAY_FONT_SIZE,color=self.VIDEO_OVERLAY_FONT_COLOR,font=self.active_moviepy_font_name,bg_color='rgba(10,10,20,0.7)',method='caption',align='West',size=(self.video_frame_size[0]*0.9,None),kerning=-1,stroke_color='black',stroke_width=1.5).set_duration(dur_text_overlay_val).set_start(start_text_overlay_val).set_position(('center',0.92),relative=True)
|
140 |
+
active_scene_clip=CompositeVideoClip([active_scene_clip,text_clip_for_overlay_obj],size=self.video_frame_size,use_bgclip=True) # Ensure use_bgclip=True
|
141 |
+
logger.debug(f"S{num_of_scene}: Text overlay composited.")
|
142 |
+
else:
|
143 |
+
logger.warning(f"S{num_of_scene}: Text overlay duration is zero or negative ({dur_text_overlay_val}). Skipping text overlay.")
|
144 |
+
except Exception as e_txt_comp_loop:logger.error(f"S{num_of_scene} TextClip compositing error:{e_txt_comp_loop}. Proceeding without text for this scene.",exc_info=True) # Log full error but continue
|
145 |
+
|
146 |
+
if active_scene_clip:
|
147 |
+
processed_moviepy_clips_list.append(active_scene_clip)
|
148 |
+
logger.info(f"S{num_of_scene}: Asset successfully processed. Clip duration: {active_scene_clip.duration:.2f}s. Added to final list for concatenation.")
|
149 |
+
|
150 |
+
except Exception as e_asset_loop_main_exc: # Catch errors during the processing of a single asset
|
151 |
+
logger.error(f"MAJOR UNHANDLED ERROR processing asset for S{num_of_scene} (Path: {path_of_asset}): {e_asset_loop_main_exc}", exc_info=True)
|
152 |
+
# Ensure any partially created clip for this iteration is closed
|
153 |
if active_scene_clip and hasattr(active_scene_clip,'close'):
|
154 |
try: active_scene_clip.close()
|
155 |
+
except Exception as e_close_active_err: logger.warning(f"S{num_of_scene}: Error closing active_scene_clip in error handler: {e_close_active_err}")
|
156 |
+
# Continue to the next asset
|
157 |
+
continue
|
158 |
+
|
159 |
+
if not processed_moviepy_clips_list:
|
160 |
+
logger.warning("No MoviePy clips were successfully processed. Aborting animatic assembly before concatenation."); return None
|
161 |
+
|
162 |
transition_duration_val=0.75
|
163 |
try:
|
164 |
+
logger.info(f"Concatenating {len(processed_moviepy_clips_list)} processed clips for final animatic.");
|
165 |
+
if len(processed_moviepy_clips_list)>1:
|
166 |
+
final_video_output_clip=concatenate_videoclips(processed_moviepy_clips_list,
|
167 |
+
padding=-transition_duration_val if transition_duration_val > 0 else 0,
|
168 |
+
method="compose") # "compose" is often more robust for mixed content
|
169 |
+
elif processed_moviepy_clips_list:
|
170 |
+
final_video_output_clip=processed_moviepy_clips_list[0] # Single clip, no concatenation needed
|
171 |
+
|
172 |
+
if not final_video_output_clip: logger.error("Concatenation resulted in a None clip. Aborting."); return None
|
173 |
+
logger.info(f"Concatenated animatic base duration:{final_video_output_clip.duration:.2f}s")
|
174 |
+
|
175 |
+
# Apply fade effects if duration allows
|
176 |
+
if transition_duration_val > 0 and final_video_output_clip.duration > 0:
|
177 |
+
if final_video_output_clip.duration > transition_duration_val * 2:
|
178 |
+
final_video_output_clip=final_video_output_clip.fx(vfx.fadein,transition_duration_val).fx(vfx.fadeout,transition_duration_val)
|
179 |
+
else: # Shorter clip, just fade in
|
180 |
+
final_video_output_clip=final_video_output_clip.fx(vfx.fadein,min(transition_duration_val,final_video_output_clip.duration/2.0))
|
181 |
+
logger.debug("Applied fade in/out effects to final composite clip.")
|
182 |
+
|
183 |
+
# Add overall narration audio
|
184 |
+
if overall_narration_path and os.path.exists(overall_narration_path) and final_video_output_clip.duration > 0:
|
185 |
+
try:
|
186 |
+
narration_audio_clip_mvpy=AudioFileClip(overall_narration_path)
|
187 |
+
logger.info(f"Adding overall narration. Video duration: {final_video_output_clip.duration:.2f}s, Narration duration: {narration_audio_clip_mvpy.duration:.2f}s")
|
188 |
+
# MoviePy will cut the audio if it's longer than the video, or pad with silence if shorter (when using set_audio)
|
189 |
+
final_video_output_clip=final_video_output_clip.set_audio(narration_audio_clip_mvpy)
|
190 |
+
logger.info("Overall narration successfully added to animatic.")
|
191 |
+
except Exception as e_narr_add_final:logger.error(f"Error adding overall narration to animatic:{e_narr_add_final}",exc_info=True)
|
192 |
+
elif final_video_output_clip.duration <= 0:
|
193 |
+
logger.warning("Animatic has zero or negative duration before adding audio. Audio will not be added.")
|
194 |
+
|
195 |
+
# Write the final video file
|
196 |
+
if final_video_output_clip and final_video_output_clip.duration > 0:
|
197 |
+
final_output_path_str=os.path.join(self.output_dir,output_filename)
|
198 |
+
logger.info(f"Writing final animatic video to: {final_output_path_str} (Target Duration: {final_video_output_clip.duration:.2f}s)")
|
199 |
+
|
200 |
+
# Ensure threads is at least 1, common os.cpu_count() can be None in some restricted envs
|
201 |
+
num_threads = os.cpu_count()
|
202 |
+
if not isinstance(num_threads, int) or num_threads < 1:
|
203 |
+
num_threads = 2 # Fallback to 2 threads
|
204 |
+
logger.warning(f"os.cpu_count() returned invalid, defaulting to {num_threads} threads for ffmpeg.")
|
205 |
+
|
206 |
+
|
207 |
+
final_video_output_clip.write_videofile(
|
208 |
+
final_output_path_str,
|
209 |
+
fps=fps,
|
210 |
+
codec='libx264', # Standard H.264 codec
|
211 |
+
preset='medium', # Good balance of speed and quality. 'ultrafast' for speed, 'slower' for quality.
|
212 |
+
audio_codec='aac', # Standard audio codec
|
213 |
+
temp_audiofile=os.path.join(self.output_dir,f'temp-audio-{os.urandom(4).hex()}.m4a'), # Temporary audio file
|
214 |
+
remove_temp=True, # Clean up temp audio
|
215 |
+
threads=num_threads,
|
216 |
+
logger='bar', # Show progress bar
|
217 |
+
bitrate="5000k", # Decent quality bitrate for 720p
|
218 |
+
ffmpeg_params=["-pix_fmt", "yuv420p"] # Crucial for compatibility and color accuracy
|
219 |
+
)
|
220 |
+
logger.info(f"Animatic video created successfully: {final_output_path_str}")
|
221 |
+
return final_output_path_str
|
222 |
+
else:
|
223 |
+
logger.error("Final animatic clip is invalid or has zero duration. Cannot write video file."); return None
|
224 |
+
except Exception as e_vid_write_final_op: # Renamed
|
225 |
+
logger.error(f"Error during final animatic video file writing or composition stage: {e_vid_write_final_op}", exc_info=True)
|
226 |
+
return None
|
227 |
finally:
|
228 |
logger.debug("Closing all MoviePy clips in `assemble_animatic_from_assets` main finally block.")
|
229 |
+
# Consolidate list of all clips that might need closing
|
230 |
+
all_clips_for_closure = processed_moviepy_clips_list[:] # Start with a copy
|
231 |
+
if narration_audio_clip_mvpy: all_clips_for_closure.append(narration_audio_clip_mvpy)
|
232 |
+
if final_video_output_clip: all_clips_for_closure.append(final_video_output_clip)
|
233 |
+
|
234 |
+
for clip_to_close_item_final in all_clips_for_closure: # Renamed
|
235 |
+
if clip_to_close_item_final and hasattr(clip_to_close_item_final, 'close'):
|
236 |
+
try: clip_to_close_item_final.close()
|
237 |
+
except Exception as e_final_clip_close_op: logger.warning(f"Ignoring error while closing a MoviePy clip ({type(clip_to_close_item_final).__name__}): {e_final_clip_close_op}")
|