| 1 | import asyncio |
| 2 | import base64 |
| 3 | import json |
| 4 | import logging |
| 5 | import math |
| 6 | import os |
| 7 | import re |
| 8 | import shutil |
| 9 | import subprocess |
| 10 | import sys |
| 11 | import tempfile |
| 12 | from html import escape |
| 13 | from pathlib import Path |
| 14 | from typing import Any, Iterable, Optional |
| 15 | |
| 16 | from PIL import Image, ImageDraw, ImageFont |
| 17 | |
| 18 | from config import BASE_DIR |
| 19 | |
| 20 | logger = logging.getLogger(__name__) |
| 21 | |
| 22 | TEMPLATE_FIELD_DEFAULTS = { |
| 23 | "title": "山河入梦", |
| 24 | "text": "心之所向，素履而往", |
| 25 | "author": "HITsz-TMG", |
| 26 | "describe": "开源视频生成智能体", |
| 27 | "brand": "Video-Claw", |
| 28 | "signature": "HITsz-TMG", |
| 29 | "subtitle": "这是一个副标题", |
| 30 | } |
| 31 | CUSTOM_TEMPLATE_FIELDS = ("author", "describe", "brand", "signature", "subtitle") |
| 32 | |
| 33 | |
| 34 | def write_text(path: str, content: str) -> str: |
| 35 | os.makedirs(os.path.dirname(path), exist_ok=True) |
| 36 | with open(path, "w", encoding="utf-8") as f: |
| 37 | f.write(content) |
| 38 | return path |
| 39 | |
| 40 | |
| 41 | def write_json(path: str, data: Any) -> str: |
| 42 | os.makedirs(os.path.dirname(path), exist_ok=True) |
| 43 | with open(path, "w", encoding="utf-8") as f: |
| 44 | json.dump(data, f, ensure_ascii=False, indent=2) |
| 45 | return path |
| 46 | |
| 47 | |
| 48 | def artifact(path: str, kind: str, name: Optional[str] = None) -> dict[str, Any]: |
| 49 | return { |
| 50 | "kind": kind, |
| 51 | "name": name or os.path.basename(path), |
| 52 | "path": path, |
| 53 | "exists": os.path.exists(path), |
| 54 | } |
| 55 | |
| 56 | |
| 57 | def extract_json_array(text: str) -> list[Any]: |
| 58 | text = text.strip() |
| 59 | try: |
| 60 | data = json.loads(text) |
| 61 | if isinstance(data, list): |
| 62 | return data |
| 63 | except Exception: |
| 64 | pass |
| 65 | |
| 66 | match = re.search(r"```(?:json)?\s*(\[.*?\])\s*```", text, re.S) |
| 67 | if not match: |
| 68 | match = re.search(r"(\[.*\])", text, re.S) |
| 69 | if match: |
| 70 | data = json.loads(match.group(1)) |
| 71 | if isinstance(data, list): |
| 72 | return data |
| 73 | raise ValueError("Model response did not contain a JSON array.") |
| 74 | |
| 75 | |
| 76 | def split_script(text: str, split_mode: str = "paragraph") -> list[str]: |
| 77 | if split_mode == "line": |
| 78 | parts = [line.strip() for line in text.splitlines()] |
| 79 | elif split_mode == "sentence": |
| 80 | parts = [part.strip() for part in re.split(r"(?<=[。！？.!?])\s*", text)] |
| 81 | else: |
| 82 | parts = [part.strip() for part in re.split(r"\n\s*\n", text)] |
| 83 | return [part for part in parts if part] |
| 84 | |
| 85 | |
| 86 | def copy_input_file(path: str, output_dir: str, prefix: str) -> str: |
| 87 | if path.startswith(("http://", "https://", "file://", "data:")): |
| 88 | return path |
| 89 | src = Path(path) |
| 90 | if not src.exists(): |
| 91 | raise FileNotFoundError(f"Input file not found: {path}") |
| 92 | dst = Path(output_dir) / f"{prefix}{src.suffix.lower()}" |
| 93 | shutil.copy2(src, dst) |
| 94 | return str(dst) |
| 95 | |
| 96 | |
| 97 | async def run_blocking(func, *args, **kwargs): |
| 98 | return await asyncio.to_thread(func, *args, **kwargs) |
| 99 | |
| 100 | |
| 101 | def concat_videos(video_paths: Iterable[str], output_path: str) -> Optional[str]: |
| 102 | paths = [path for path in video_paths if path and os.path.exists(path)] |
| 103 | if not paths: |
| 104 | return None |
| 105 | |
| 106 | ffmpeg = shutil.which("ffmpeg") |
| 107 | if not ffmpeg: |
| 108 | logger.warning("ffmpeg not found; cannot concatenate videos") |
| 109 | return None |
| 110 | |
| 111 | list_path = os.path.join(os.path.dirname(output_path), "concat.txt") |
| 112 | with open(list_path, "w", encoding="utf-8") as f: |
| 113 | for path in paths: |
| 114 | f.write(f"file '{os.path.abspath(path)}'\n") |
| 115 | |
| 116 | cmd = [ffmpeg, "-y", "-f", "concat", "-safe", "0", "-i", list_path, "-c", "copy", output_path] |
| 117 | logger.info("Concatenating %d videos -> %s", len(paths), output_path) |
| 118 | subprocess.run(cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE) |
| 119 | logger.info("Video concatenation complete: %s", output_path) |
| 120 | return output_path |
| 121 | |
| 122 | |
| 123 | def create_static_image_clip( |
| 124 | image_path: str, |
| 125 | audio_path: str, |
| 126 | output_path: str, |
| 127 | *, |
| 128 | video_ratio: str = "9:16", |
| 129 | duration: Optional[float] = None, |
| 130 | ) -> str: |
| 131 | ffmpeg = shutil.which("ffmpeg") |
| 132 | if not ffmpeg: |
| 133 | raise RuntimeError("ffmpeg is required to create static short-video clips.") |
| 134 | if not os.path.exists(image_path): |
| 135 | raise FileNotFoundError(f"Image not found: {image_path}") |
| 136 | if not os.path.exists(audio_path): |
| 137 | raise FileNotFoundError(f"Audio not found: {audio_path}") |
| 138 | |
| 139 | width, height = _resolution_from_ratio(video_ratio) |
| 140 | clip_duration = duration or media_duration_seconds(audio_path) or 3.0 |
| 141 | vf = ( |
| 142 | f"scale={width}:{height}:force_original_aspect_ratio=decrease," |
| 143 | f"pad={width}:{height}:(ow-iw)/2:(oh-ih)/2:color=black," |
| 144 | "format=yuv420p" |
| 145 | ) |
| 146 | os.makedirs(os.path.dirname(output_path), exist_ok=True) |
| 147 | logger.info( |
| 148 | "Creating static image clip: image=%s audio=%s duration=%.2fs -> %s", |
| 149 | image_path, |
| 150 | audio_path, |
| 151 | clip_duration, |
| 152 | output_path, |
| 153 | ) |
| 154 | cmd = [ |
| 155 | ffmpeg, |
| 156 | "-y", |
| 157 | "-loop", |
| 158 | "1", |
| 159 | "-framerate", |
| 160 | "30", |
| 161 | "-i", |
| 162 | image_path, |
| 163 | "-i", |
| 164 | audio_path, |
| 165 | "-t", |
| 166 | f"{clip_duration:.3f}", |
| 167 | "-vf", |
| 168 | vf, |
| 169 | "-c:v", |
| 170 | "libx264", |
| 171 | "-preset", |
| 172 | "veryfast", |
| 173 | "-crf", |
| 174 | "18", |
| 175 | "-c:a", |
| 176 | "aac", |
| 177 | "-shortest", |
| 178 | "-movflags", |
| 179 | "+faststart", |
| 180 | output_path, |
| 181 | ] |
| 182 | subprocess.run(cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE) |
| 183 | return output_path |
| 184 | |
| 185 | |
| 186 | def _load_font(size: int) -> ImageFont.ImageFont: |
| 187 | candidates = [ |
| 188 | "/System/Library/Fonts/PingFang.ttc", |
| 189 | "/System/Library/Fonts/STHeiti Light.ttc", |
| 190 | "/System/Library/Fonts/Supplemental/Arial Unicode.ttf", |
| 191 | "/usr/share/fonts/opentype/noto/NotoSansCJK-Regular.ttc", |
| 192 | "/usr/share/fonts/truetype/dejavu/DejaVuSans.ttf", |
| 193 | ] |
| 194 | for path in candidates: |
| 195 | if os.path.exists(path): |
| 196 | try: |
| 197 | return ImageFont.truetype(path, size) |
| 198 | except Exception: |
| 199 | continue |
| 200 | return ImageFont.load_default() |
| 201 | |
| 202 | |
| 203 | def _text_width(draw: ImageDraw.ImageDraw, text: str, font: ImageFont.ImageFont) -> int: |
| 204 | bbox = draw.textbbox((0, 0), text, font=font) |
| 205 | return bbox[2] - bbox[0] |
| 206 | |
| 207 | |
| 208 | def _wrap_text(draw: ImageDraw.ImageDraw, text: str, font: ImageFont.ImageFont, max_width: int) -> list[str]: |
| 209 | lines: list[str] = [] |
| 210 | for paragraph in str(text or "").splitlines() or [""]: |
| 211 | current = "" |
| 212 | for char in paragraph.strip(): |
| 213 | candidate = current + char |
| 214 | if current and _text_width(draw, candidate, font) > max_width: |
| 215 | lines.append(current) |
| 216 | current = char |
| 217 | else: |
| 218 | current = candidate |
| 219 | if current: |
| 220 | lines.append(current) |
| 221 | return lines or [""] |
| 222 | |
| 223 | |
| 224 | def _draw_centered_lines( |
| 225 | draw: ImageDraw.ImageDraw, |
| 226 | lines: list[str], |
| 227 | *, |
| 228 | center_x: int, |
| 229 | y: int, |
| 230 | font: ImageFont.ImageFont, |
| 231 | fill: tuple[int, int, int, int], |
| 232 | stroke_width: int, |
| 233 | stroke_fill: tuple[int, int, int, int], |
| 234 | line_gap: int, |
| 235 | ) -> int: |
| 236 | cursor = y |
| 237 | for line in lines: |
| 238 | bbox = draw.textbbox((0, 0), line, font=font, stroke_width=stroke_width) |
| 239 | width = bbox[2] - bbox[0] |
| 240 | height = bbox[3] - bbox[1] |
| 241 | draw.text( |
| 242 | (center_x - width / 2, cursor), |
| 243 | line, |
| 244 | font=font, |
| 245 | fill=fill, |
| 246 | stroke_width=stroke_width, |
| 247 | stroke_fill=stroke_fill, |
| 248 | ) |
| 249 | cursor += height + line_gap |
| 250 | return cursor |
| 251 | |
| 252 | |
| 253 | def render_static_text_image( |
| 254 | image_path: str, |
| 255 | output_path: str, |
| 256 | *, |
| 257 | subtitle: str, |
| 258 | title: Optional[str] = None, |
| 259 | video_ratio: str = "9:16", |
| 260 | ) -> str: |
| 261 | if not os.path.exists(image_path): |
| 262 | raise FileNotFoundError(f"Image not found: {image_path}") |
| 263 | |
| 264 | width, height = _resolution_from_ratio(video_ratio) |
| 265 | with Image.open(image_path) as source: |
| 266 | source = source.convert("RGB") |
| 267 | source.thumbnail((width, height), Image.Resampling.LANCZOS) |
| 268 | canvas = Image.new("RGB", (width, height), (0, 0, 0)) |
| 269 | canvas.paste(source, ((width - source.width) // 2, (height - source.height) // 2)) |
| 270 | |
| 271 | image = canvas.convert("RGBA") |
| 272 | overlay = Image.new("RGBA", image.size, (0, 0, 0, 0)) |
| 273 | draw = ImageDraw.Draw(overlay) |
| 274 | |
| 275 | title_font = _load_font(max(48, int(height * 0.06))) |
| 276 | subtitle_font = _load_font(max(32, int(height * 0.035))) |
| 277 | margin_x = max(48, int(width * 0.07)) |
| 278 | max_text_width = width - margin_x * 2 |
| 279 | |
| 280 | if title: |
| 281 | title_lines = _wrap_text(draw, title, title_font, max_text_width) |
| 282 | title_line_height = max(1, draw.textbbox((0, 0), "国", font=title_font)[3]) |
| 283 | title_height = len(title_lines) * title_line_height + max(0, len(title_lines) - 1) * 10 |
| 284 | title_y = max(48, int(height * 0.055)) |
| 285 | _draw_centered_lines( |
| 286 | draw, |
| 287 | title_lines, |
| 288 | center_x=width // 2, |
| 289 | y=title_y, |
| 290 | font=title_font, |
| 291 | fill=(255, 255, 255, 255), |
| 292 | stroke_width=3, |
| 293 | stroke_fill=(0, 0, 0, 210), |
| 294 | line_gap=10, |
| 295 | ) |
| 296 | |
| 297 | subtitle_lines = _wrap_text(draw, subtitle, subtitle_font, max_text_width) |
| 298 | subtitle_line_height = max(1, draw.textbbox((0, 0), "国", font=subtitle_font)[3]) |
| 299 | subtitle_height = len(subtitle_lines) * subtitle_line_height + max(0, len(subtitle_lines) - 1) * 10 |
| 300 | subtitle_y = height - max(96, int(height * 0.08)) - subtitle_height |
| 301 | _draw_centered_lines( |
| 302 | draw, |
| 303 | subtitle_lines, |
| 304 | center_x=width // 2, |
| 305 | y=subtitle_y, |
| 306 | font=subtitle_font, |
| 307 | fill=(255, 255, 255, 255), |
| 308 | stroke_width=3, |
| 309 | stroke_fill=(0, 0, 0, 230), |
| 310 | line_gap=10, |
| 311 | ) |
| 312 | |
| 313 | os.makedirs(os.path.dirname(output_path), exist_ok=True) |
| 314 | Image.alpha_composite(image, overlay).convert("RGB").save(output_path, quality=95) |
| 315 | logger.info("Rendered static title/subtitle image: %s -> %s", image_path, output_path) |
| 316 | return output_path |
| 317 | |
| 318 | |
| 319 | def ratio_from_size(width: int, height: int) -> str: |
| 320 | if width <= 0 or height <= 0: |
| 321 | return "1:1" |
| 322 | divisor = max(1, math.gcd(width, height)) |
| 323 | return f"{width // divisor}:{height // divisor}" |
| 324 | |
| 325 | |
| 326 | def _template_size_from_id(template_id: str) -> tuple[str, int, int]: |
| 327 | size = str(template_id or "").split("/", 1)[0] |
| 328 | if size == "1920x1080": |
| 329 | return size, 1920, 1080 |
| 330 | if size == "1080x1080": |
| 331 | return size, 1080, 1080 |
| 332 | if size == "1080x1920": |
| 333 | return size, 1080, 1920 |
| 334 | raise ValueError(f"Unsupported subtitle template size: {template_id}") |
| 335 | |
| 336 | |
| 337 | def _resolve_template_path(template_id: str, video_ratio: str) -> tuple[str, int, int]: |
| 338 | if "/" in str(template_id): |
| 339 | size, filename = str(template_id).split("/", 1) |
| 340 | _, width, height = _template_size_from_id(template_id) |
| 341 | else: |
| 342 | width, height = _resolution_from_ratio(video_ratio) |
| 343 | size = f"{width}x{height}" |
| 344 | filename = str(template_id) |
| 345 | if "/" in filename or "\\" in filename or not filename.endswith(".html"): |
| 346 | raise ValueError(f"Invalid subtitle template: {template_id}") |
| 347 | path = (BASE_DIR / "templates" / size / filename).resolve() |
| 348 | root = (BASE_DIR / "templates" / size).resolve() |
| 349 | if not str(path).startswith(str(root) + os.sep) or not path.exists(): |
| 350 | raise FileNotFoundError(f"Subtitle template not found: {template_id}") |
| 351 | return str(path), width, height |
| 352 | |
| 353 | |
| 354 | def template_media_spec(template_id: str, video_ratio: str = "9:16") -> dict[str, Any]: |
| 355 | template_path, _, _ = _resolve_template_path(template_id, video_ratio) |
| 356 | with open(template_path, "r", encoding="utf-8") as f: |
| 357 | raw = f.read() |
| 358 | width_match = re.search( |
| 359 | r'<meta\s+name=["\']template:media-width["\']\s+content=["\'](\d+)["\']', |
| 360 | raw, |
| 361 | re.I, |
| 362 | ) |
| 363 | height_match = re.search( |
| 364 | r'<meta\s+name=["\']template:media-height["\']\s+content=["\'](\d+)["\']', |
| 365 | raw, |
| 366 | re.I, |
| 367 | ) |
| 368 | if not width_match or not height_match: |
| 369 | raise ValueError(f"Subtitle template missing media size metadata: {template_id}") |
| 370 | width = int(width_match.group(1)) |
| 371 | height = int(height_match.group(1)) |
| 372 | return { |
| 373 | "media_width": width, |
| 374 | "media_height": height, |
| 375 | "media_ratio": ratio_from_size(width, height), |
| 376 | "media_resolution": f"{width}*{height}", |
| 377 | "supports_video": template_supports_video_media(raw), |
| 378 | } |
| 379 | |
| 380 | |
| 381 | def template_supports_video_media(raw: str) -> bool: |
| 382 | visible_raw = re.sub(r"<!--.*?-->", "", raw, flags=re.S) |
| 383 | if re.search(r"\{\{\s*media(?:[:=][^{}]*)?\s*\}\}", visible_raw): |
| 384 | return True |
| 385 | return bool( |
| 386 | re.search( |
| 387 | r"<img\b[^>]*\bsrc=[\"']\{\{\s*image\s*\}\}[\"'][^>]*>", |
| 388 | visible_raw, |
| 389 | re.I, |
| 390 | ) |
| 391 | ) |
| 392 | |
| 393 | |
| 394 | def parse_template_placeholders(raw: str) -> list[dict[str, str]]: |
| 395 | placeholders = [] |
| 396 | for match in re.finditer(r"\{\{\s*([^{}]+?)\s*\}\}", raw): |
| 397 | token = match.group(1).strip() |
| 398 | key = token.split(":", 1)[0].split("=", 1)[0].strip() |
| 399 | field_type = token.split(":", 1)[1].split("=", 1)[0].strip() if ":" in token else "text" |
| 400 | default = token.split("=", 1)[1].strip() if "=" in token else "" |
| 401 | placeholders.append({ |
| 402 | "key": key, |
| 403 | "type": field_type, |
| 404 | "default": TEMPLATE_FIELD_DEFAULTS.get(key, default), |
| 405 | }) |
| 406 | return placeholders |
| 407 | |
| 408 | |
| 409 | def template_custom_fields(template_id: str, video_ratio: str = "9:16") -> list[dict[str, str]]: |
| 410 | template_path, _, _ = _resolve_template_path(template_id, video_ratio) |
| 411 | with open(template_path, "r", encoding="utf-8") as f: |
| 412 | placeholders = parse_template_placeholders(f.read()) |
| 413 | |
| 414 | fields = [] |
| 415 | seen = set() |
| 416 | for item in placeholders: |
| 417 | key = item["key"] |
| 418 | if key not in CUSTOM_TEMPLATE_FIELDS or key in seen: |
| 419 | continue |
| 420 | seen.add(key) |
| 421 | fields.append(item) |
| 422 | return fields |
| 423 | |
| 424 | |
| 425 | def _image_data_uri(path: str) -> str: |
| 426 | if path.startswith(("http://", "https://", "data:", "file:")): |
| 427 | return path |
| 428 | if not os.path.exists(path): |
| 429 | raise FileNotFoundError(f"Image not found: {path}") |
| 430 | suffix = Path(path).suffix.lower() |
| 431 | mime = { |
| 432 | ".jpg": "image/jpeg", |
| 433 | ".jpeg": "image/jpeg", |
| 434 | ".png": "image/png", |
| 435 | ".webp": "image/webp", |
| 436 | ".gif": "image/gif", |
| 437 | }.get(suffix, "image/png") |
| 438 | with open(path, "rb") as f: |
| 439 | encoded = base64.b64encode(f.read()).decode("ascii") |
| 440 | return f"data:{mime};base64,{encoded}" |
| 441 | |
| 442 | |
| 443 | def _render_template_html(raw: str, values: dict[str, Any], raw_keys: Optional[set[str]] = None) -> str: |
| 444 | raw_keys = raw_keys or set() |
| 445 | |
| 446 | def repl(match: re.Match) -> str: |
| 447 | token = match.group(1).strip() |
| 448 | key = token.split(":", 1)[0].split("=", 1)[0].strip() |
| 449 | if key in values: |
| 450 | if key in raw_keys: |
| 451 | return str(values[key] or "") |
| 452 | return escape(str(values[key] or ""), quote=True) |
| 453 | if "=" in token: |
| 454 | return escape(token.split("=", 1)[1].strip(), quote=True) |
| 455 | return "" |
| 456 | |
| 457 | return re.sub(r"\{\{\s*([^{}]+?)\s*\}\}", repl, raw) |
| 458 | |
| 459 | |
| 460 | def _media_file_uri(path: str) -> str: |
| 461 | if path.startswith(("http://", "https://", "data:", "file:")): |
| 462 | return path |
| 463 | if not os.path.exists(path): |
| 464 | raise FileNotFoundError(f"Media not found: {path}") |
| 465 | return Path(path).resolve().as_uri() |
| 466 | |
| 467 | |
| 468 | def _template_media_element(src: str, media_kind: str) -> str: |
| 469 | attrs = 'class="template-media" style="width:100%;height:100%;object-fit:cover;display:block;"' |
| 470 | if media_kind == "video": |
| 471 | return f'<video {attrs} src="{escape(src, quote=True)}" muted playsinline preload="auto"></video>' |
| 472 | return f'<img {attrs} src="{escape(src, quote=True)}" alt="">' |
| 473 | |
| 474 | |
| 475 | def _inject_template_media_css(raw: str) -> str: |
| 476 | css = ( |
| 477 | "<style>" |
| 478 | ".template-media{width:100%;height:100%;object-fit:cover;display:block;}" |
| 479 | "video.template-media{background:#000;}" |
| 480 | "</style>" |
| 481 | ) |
| 482 | if "</head>" in raw: |
| 483 | return raw.replace("</head>", f"{css}</head>", 1) |
| 484 | return css + raw |
| 485 | |
| 486 | |
| 487 | def _prepare_template_media_html(raw: str, *, media_kind: str) -> str: |
| 488 | prepared = _inject_template_media_css(raw) |
| 489 | if media_kind != "video": |
| 490 | return prepared |
| 491 | if re.search(r"\{\{\s*media(?:[:=][^{}]*)?\s*\}\}", prepared): |
| 492 | return prepared |
| 493 | return re.sub( |
| 494 | r"<img\b[^>]*\bsrc=[\"']\{\{\s*image\s*\}\}[\"'][^>]*>", |
| 495 | "{{media}}", |
| 496 | prepared, |
| 497 | flags=re.I, |
| 498 | ) |
| 499 | |
| 500 | |
| 501 | def _template_values( |
| 502 | *, |
| 503 | image_path: str, |
| 504 | subtitle: str, |
| 505 | title: Optional[str], |
| 506 | template_values: Optional[dict[str, Any]], |
| 507 | index: int, |
| 508 | media_kind: str = "image", |
| 509 | media_path: Optional[str] = None, |
| 510 | ) -> dict[str, Any]: |
| 511 | image_uri = _image_data_uri(image_path) |
| 512 | media_src = _media_file_uri(media_path) if media_kind == "video" and media_path else image_uri |
| 513 | return { |
| 514 | **TEMPLATE_FIELD_DEFAULTS, |
| 515 | **(template_values or {}), |
| 516 | "title": title or TEMPLATE_FIELD_DEFAULTS["title"], |
| 517 | "text": subtitle or "", |
| 518 | "image": image_uri, |
| 519 | "media": _template_media_element(media_src, media_kind), |
| 520 | "index": index, |
| 521 | } |
| 522 | |
| 523 | |
| 524 | def _install_playwright_chromium() -> None: |
| 525 | cmd = [sys.executable, "-m", "playwright", "install", "chromium"] |
| 526 | logger.info("Installing Playwright Chromium for HTML subtitle templates") |
| 527 | try: |
| 528 | subprocess.run(cmd, check=True) |
| 529 | except subprocess.CalledProcessError as exc: |
| 530 | raise RuntimeError( |
| 531 | "Playwright is installed, but Chromium is missing and automatic installation failed. " |
| 532 | "Run `python -m playwright install chromium` manually and retry." |
| 533 | ) from exc |
| 534 | |
| 535 | |
| 536 | def _launch_playwright_chromium(playwright): |
| 537 | try: |
| 538 | return playwright.chromium.launch(headless=True) |
| 539 | except Exception as exc: |
| 540 | message = str(exc) |
| 541 | if "Executable doesn't exist" not in message and "playwright install" not in message: |
| 542 | raise |
| 543 | _install_playwright_chromium() |
| 544 | return playwright.chromium.launch(headless=True) |
| 545 | |
| 546 | |
| 547 | def render_template_text_image( |
| 548 | image_path: str, |
| 549 | output_path: str, |
| 550 | *, |
| 551 | subtitle: str, |
| 552 | title: Optional[str] = None, |
| 553 | video_ratio: str = "9:16", |
| 554 | template_id: str, |
| 555 | template_values: Optional[dict[str, Any]] = None, |
| 556 | index: int = 1, |
| 557 | ) -> str: |
| 558 | template_path, width, height = _resolve_template_path(template_id, video_ratio) |
| 559 | with open(template_path, "r", encoding="utf-8") as f: |
| 560 | raw = _prepare_template_media_html(f.read(), media_kind="image") |
| 561 | html = _render_template_html( |
| 562 | raw, |
| 563 | _template_values( |
| 564 | image_path=image_path, |
| 565 | subtitle=subtitle, |
| 566 | title=title, |
| 567 | template_values=template_values, |
| 568 | index=index, |
| 569 | ), |
| 570 | raw_keys={"media"}, |
| 571 | ) |
| 572 | |
| 573 | try: |
| 574 | from playwright.sync_api import sync_playwright |
| 575 | except ImportError as exc: |
| 576 | raise RuntimeError( |
| 577 | "HTML subtitle templates require Playwright. " |
| 578 | "Install backend dependencies again or run `pip install playwright`." |
| 579 | ) from exc |
| 580 | |
| 581 | os.makedirs(os.path.dirname(output_path), exist_ok=True) |
| 582 | with sync_playwright() as playwright: |
| 583 | browser = _launch_playwright_chromium(playwright) |
| 584 | try: |
| 585 | page = browser.new_page(viewport={"width": width, "height": height}, device_scale_factor=1) |
| 586 | page.set_content(html, wait_until="networkidle", timeout=30000) |
| 587 | page.screenshot(path=output_path, full_page=False, type="jpeg", quality=95) |
| 588 | finally: |
| 589 | browser.close() |
| 590 | logger.info("Rendered HTML template subtitle image: template=%s image=%s -> %s", template_id, image_path, output_path) |
| 591 | return output_path |
| 592 | |
| 593 | |
| 594 | def render_template_media_video( |
| 595 | media_video_path: str, |
| 596 | output_path: str, |
| 597 | *, |
| 598 | poster_image_path: str, |
| 599 | subtitle: str, |
| 600 | title: Optional[str] = None, |
| 601 | video_ratio: str = "9:16", |
| 602 | template_id: str, |
| 603 | template_values: Optional[dict[str, Any]] = None, |
| 604 | index: int = 1, |
| 605 | duration: Optional[float] = None, |
| 606 | fps: int = 24, |
| 607 | ) -> str: |
| 608 | ffmpeg = shutil.which("ffmpeg") |
| 609 | if not ffmpeg: |
| 610 | raise RuntimeError("ffmpeg is required to render HTML template videos.") |
| 611 | if not os.path.exists(media_video_path): |
| 612 | raise FileNotFoundError(f"Template media video not found: {media_video_path}") |
| 613 | |
| 614 | template_path, width, height = _resolve_template_path(template_id, video_ratio) |
| 615 | with open(template_path, "r", encoding="utf-8") as f: |
| 616 | raw = f.read() |
| 617 | if not template_supports_video_media(raw): |
| 618 | raise ValueError(f"Template does not support video media: {template_id}") |
| 619 | |
| 620 | raw = _prepare_template_media_html(raw, media_kind="video") |
| 621 | html = _render_template_html( |
| 622 | raw, |
| 623 | _template_values( |
| 624 | image_path=poster_image_path, |
| 625 | media_kind="image", |
| 626 | subtitle=subtitle, |
| 627 | title=title, |
| 628 | template_values=template_values, |
| 629 | index=index, |
| 630 | ), |
| 631 | raw_keys={"media"}, |
| 632 | ) |
| 633 | |
| 634 | try: |
| 635 | from playwright.sync_api import sync_playwright |
| 636 | except ImportError as exc: |
| 637 | raise RuntimeError( |
| 638 | "HTML subtitle templates require Playwright. " |
| 639 | "Install backend dependencies again or run `pip install playwright`." |
| 640 | ) from exc |
| 641 | |
| 642 | clip_duration = float(duration or media_duration_seconds(media_video_path) or 3.0) |
| 643 | frame_count = max(1, int(math.ceil(clip_duration * fps))) |
| 644 | os.makedirs(os.path.dirname(output_path), exist_ok=True) |
| 645 | frame_dir = tempfile.mkdtemp(prefix="template_frames_", dir=os.path.dirname(output_path)) |
| 646 | media_frame_dir = tempfile.mkdtemp(prefix="template_media_frames_", dir=os.path.dirname(output_path)) |
| 647 | frame_pattern = os.path.join(frame_dir, "frame_%05d.jpg") |
| 648 | media_frame_pattern = os.path.join(media_frame_dir, "media_%05d.jpg") |
| 649 | |
| 650 | logger.info( |
| 651 | "Rendering HTML template video: template=%s media=%s duration=%.2fs fps=%d -> %s", |
| 652 | template_id, |
| 653 | media_video_path, |
| 654 | clip_duration, |
| 655 | fps, |
| 656 | output_path, |
| 657 | ) |
| 658 | try: |
| 659 | extract_cmd = [ |
| 660 | ffmpeg, |
| 661 | "-y", |
| 662 | "-i", |
| 663 | media_video_path, |
| 664 | "-vf", |
| 665 | f"fps={fps}", |
| 666 | "-q:v", |
| 667 | "2", |
| 668 | media_frame_pattern, |
| 669 | ] |
| 670 | subprocess.run(extract_cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE) |
| 671 | media_frames = sorted(str(path) for path in Path(media_frame_dir).glob("media_*.jpg")) |
| 672 | if not media_frames: |
| 673 | media_frames = [poster_image_path] |
| 674 | |
| 675 | with sync_playwright() as playwright: |
| 676 | browser = _launch_playwright_chromium(playwright) |
| 677 | try: |
| 678 | page = browser.new_page(viewport={"width": width, "height": height}, device_scale_factor=1) |
| 679 | page.set_content(html, wait_until="networkidle", timeout=30000) |
| 680 | for frame_index in range(frame_count): |
| 681 | media_frame_uri = _image_data_uri(media_frames[frame_index % len(media_frames)]) |
| 682 | page.evaluate( |
| 683 | """async (src) => { |
| 684 | const images = Array.from(document.querySelectorAll('img.template-media')); |
| 685 | await Promise.all(images.map(async (image) => { |
| 686 | image.src = src; |
| 687 | if (image.decode) { |
| 688 | await image.decode().catch(() => {}); |
| 689 | } |
| 690 | })); |
| 691 | await new Promise(resolve => requestAnimationFrame(resolve)); |
| 692 | }""", |
| 693 | media_frame_uri, |
| 694 | ) |
| 695 | page.screenshot( |
| 696 | path=os.path.join(frame_dir, f"frame_{frame_index + 1:05d}.jpg"), |
| 697 | full_page=False, |
| 698 | type="jpeg", |
| 699 | quality=92, |
| 700 | ) |
| 701 | finally: |
| 702 | browser.close() |
| 703 | |
| 704 | cmd = [ |
| 705 | ffmpeg, |
| 706 | "-y", |
| 707 | "-framerate", |
| 708 | str(fps), |
| 709 | "-i", |
| 710 | frame_pattern, |
| 711 | "-t", |
| 712 | f"{clip_duration:.3f}", |
| 713 | "-vf", |
| 714 | "format=yuv420p", |
| 715 | "-c:v", |
| 716 | "libx264", |
| 717 | "-preset", |
| 718 | "veryfast", |
| 719 | "-crf", |
| 720 | "18", |
| 721 | "-movflags", |
| 722 | "+faststart", |
| 723 | output_path, |
| 724 | ] |
| 725 | subprocess.run(cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE) |
| 726 | finally: |
| 727 | shutil.rmtree(frame_dir, ignore_errors=True) |
| 728 | shutil.rmtree(media_frame_dir, ignore_errors=True) |
| 729 | return output_path |
| 730 | |
| 731 | |
| 732 | def concat_audios(audio_paths: Iterable[str], output_path: str) -> Optional[str]: |
| 733 | paths = [path for path in audio_paths if path and os.path.exists(path)] |
| 734 | if not paths: |
| 735 | return None |
| 736 | if len(paths) == 1: |
| 737 | return paths[0] |
| 738 | |
| 739 | ffmpeg = shutil.which("ffmpeg") |
| 740 | if not ffmpeg: |
| 741 | logger.warning("ffmpeg not found; cannot concatenate audios") |
| 742 | return None |
| 743 | |
| 744 | list_path = os.path.join(os.path.dirname(output_path), "concat_audio.txt") |
| 745 | with open(list_path, "w", encoding="utf-8") as f: |
| 746 | for path in paths: |
| 747 | f.write(f"file '{os.path.abspath(path)}'\n") |
| 748 | |
| 749 | cmd = [ffmpeg, "-y", "-f", "concat", "-safe", "0", "-i", list_path, "-c", "copy", output_path] |
| 750 | logger.info("Concatenating %d audios -> %s", len(paths), output_path) |
| 751 | subprocess.run(cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE) |
| 752 | return output_path |
| 753 | |
| 754 | |
| 755 | def replace_video_audio(video_path: str, audio_path: str, output_path: str) -> str: |
| 756 | ffmpeg = shutil.which("ffmpeg") |
| 757 | if not ffmpeg: |
| 758 | raise RuntimeError("ffmpeg is required to replace digital-human video audio.") |
| 759 | if not os.path.exists(video_path): |
| 760 | raise FileNotFoundError(f"Video not found: {video_path}") |
| 761 | if not os.path.exists(audio_path): |
| 762 | raise FileNotFoundError(f"Audio not found: {audio_path}") |
| 763 | |
| 764 | os.makedirs(os.path.dirname(output_path), exist_ok=True) |
| 765 | logger.info("Replacing video audio: video=%s audio=%s -> %s", video_path, audio_path, output_path) |
| 766 | cmd = [ |
| 767 | ffmpeg, |
| 768 | "-y", |
| 769 | "-i", |
| 770 | video_path, |
| 771 | "-i", |
| 772 | audio_path, |
| 773 | "-map", |
| 774 | "0:v:0", |
| 775 | "-map", |
| 776 | "1:a:0", |
| 777 | "-c:v", |
| 778 | "copy", |
| 779 | "-c:a", |
| 780 | "aac", |
| 781 | "-shortest", |
| 782 | output_path, |
| 783 | ] |
| 784 | subprocess.run(cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE) |
| 785 | return output_path |
| 786 | |
| 787 | |
| 788 | def media_duration_seconds(path: str) -> Optional[float]: |
| 789 | ffprobe = shutil.which("ffprobe") |
| 790 | if not ffprobe or not os.path.exists(path): |
| 791 | return None |
| 792 | |
| 793 | cmd = [ |
| 794 | ffprobe, |
| 795 | "-v", |
| 796 | "error", |
| 797 | "-show_entries", |
| 798 | "format=duration", |
| 799 | "-of", |
| 800 | "default=noprint_wrappers=1:nokey=1", |
| 801 | path, |
| 802 | ] |
| 803 | try: |
| 804 | result = subprocess.run(cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True) |
| 805 | duration = float(result.stdout.strip()) |
| 806 | logger.debug("Media duration: %s = %.3fs", path, duration) |
| 807 | return duration |
| 808 | except Exception as exc: |
| 809 | logger.warning("Failed to probe media duration: %s (%s)", path, exc) |
| 810 | return None |
| 811 | |
| 812 | |
| 813 | def _resolution_from_ratio(video_ratio: str) -> tuple[int, int]: |
| 814 | ratio = (video_ratio or "9:16").strip() |
| 815 | if ratio in {"16:9", "landscape"}: |
| 816 | return 1920, 1080 |
| 817 | if ratio in {"1:1", "square"}: |
| 818 | return 1080, 1080 |
| 819 | return 1080, 1920 |
| 820 | |
| 821 | |
| 822 | def _ass_time(seconds: float) -> str: |
| 823 | total_centiseconds = max(0, int(round(seconds * 100))) |
| 824 | centiseconds = total_centiseconds % 100 |
| 825 | total_seconds = total_centiseconds // 100 |
| 826 | hours = total_seconds // 3600 |
| 827 | minutes = (total_seconds % 3600) // 60 |
| 828 | secs = total_seconds % 60 |
| 829 | return f"{hours}:{minutes:02d}:{secs:02d}.{centiseconds:02d}" |
| 830 | |
| 831 | |
| 832 | def _ass_text(text: str) -> str: |
| 833 | cleaned = str(text or "").strip() |
| 834 | cleaned = cleaned.replace("{", "｛").replace("}", "｝") |
| 835 | return cleaned.replace("\r\n", "\\N").replace("\n", "\\N") |
| 836 | |
| 837 | |
| 838 | def write_ass_subtitles( |
| 839 | path: str, |
| 840 | *, |
| 841 | subtitles: Iterable[tuple[str, float]], |
| 842 | title: Optional[str] = None, |
| 843 | video_ratio: str = "9:16", |
| 844 | ) -> str: |
| 845 | width, height = _resolution_from_ratio(video_ratio) |
| 846 | title_size = max(36, int(height * 0.038)) |
| 847 | subtitle_size = max(34, int(height * 0.032)) |
| 848 | title_margin_v = max(70, int(height * 0.06)) |
| 849 | subtitle_margin_v = max(80, int(height * 0.085)) |
| 850 | subtitle_margin_h = max(70, int(width * 0.07)) |
| 851 | |
| 852 | events: list[str] = [] |
| 853 | cursor = 0.0 |
| 854 | normalized: list[tuple[str, float, float]] = [] |
| 855 | for text, duration in subtitles: |
| 856 | duration_seconds = max(0.1, float(duration or 0)) |
| 857 | start = cursor |
| 858 | end = cursor + duration_seconds |
| 859 | normalized.append((text, start, end)) |
| 860 | cursor = end |
| 861 | |
| 862 | total_duration = cursor or 0.1 |
| 863 | if title: |
| 864 | events.append( |
| 865 | f"Dialogue: 0,{_ass_time(0)},{_ass_time(total_duration)},Title,,0,0,0,,{_ass_text(title)}" |
| 866 | ) |
| 867 | for text, start, end in normalized: |
| 868 | events.append( |
| 869 | f"Dialogue: 0,{_ass_time(start)},{_ass_time(end)},Subtitle,,0,0,0,,{_ass_text(text)}" |
| 870 | ) |
| 871 | |
| 872 | content = "\n".join( |
| 873 | [ |
| 874 | "[Script Info]", |
| 875 | "ScriptType: v4.00+", |
| 876 | "Collisions: Normal", |
| 877 | f"PlayResX: {width}", |
| 878 | f"PlayResY: {height}", |
| 879 | "WrapStyle: 0", |
| 880 | "ScaledBorderAndShadow: yes", |
| 881 | "", |
| 882 | "[V4+ Styles]", |
| 883 | "Format: Name, Fontname, Fontsize, PrimaryColour, SecondaryColour, OutlineColour, BackColour, Bold, Italic, Underline, StrikeOut, ScaleX, ScaleY, Spacing, Angle, BorderStyle, Outline, Shadow, Alignment, MarginL, MarginR, MarginV, Encoding", |
| 884 | f"Style: Title,Arial,{title_size},&H00FFFFFF,&H000000FF,&H99000000,&H66000000,-1,0,0,0,100,100,0,0,1,4,0,8,{subtitle_margin_h},{subtitle_margin_h},{title_margin_v},1", |
| 885 | f"Style: Subtitle,Arial,{subtitle_size},&H00FFFFFF,&H000000FF,&HAA000000,&H66000000,0,0,0,0,100,100,0,0,1,4,0,2,{subtitle_margin_h},{subtitle_margin_h},{subtitle_margin_v},1", |
| 886 | "", |
| 887 | "[Events]", |
| 888 | "Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text", |
| 889 | *events, |
| 890 | "", |
| 891 | ] |
| 892 | ) |
| 893 | logger.info("Writing ASS subtitles: %s lines=%d title=%s", path, len(normalized), bool(title)) |
| 894 | return write_text(path, content) |
| 895 | |
| 896 | |
| 897 | def _ffmpeg_filter_path(path: str) -> str: |
| 898 | escaped = os.path.abspath(path).replace("\\", "/") |
| 899 | return escaped.replace(":", "\\:").replace("'", "\\'") |
| 900 | |
| 901 | |
| 902 | def _ffmpeg_has_filter(ffmpeg: str, filter_name: str) -> bool: |
| 903 | try: |
| 904 | result = subprocess.run( |
| 905 | [ffmpeg, "-hide_banner", "-filters"], |
| 906 | check=True, |
| 907 | stdout=subprocess.PIPE, |
| 908 | stderr=subprocess.PIPE, |
| 909 | text=True, |
| 910 | ) |
| 911 | except Exception as exc: |
| 912 | logger.warning("Failed to inspect ffmpeg filters: %s", exc) |
| 913 | return False |
| 914 | return any(line.split()[1:2] == [filter_name] for line in result.stdout.splitlines() if line.strip()) |
| 915 | |
| 916 | |
| 917 | def burn_ass_subtitles(video_path: str, subtitle_path: str, output_path: str) -> str: |
| 918 | ffmpeg = shutil.which("ffmpeg") |
| 919 | if not ffmpeg: |
| 920 | raise RuntimeError("ffmpeg is required to burn subtitles into video.") |
| 921 | if not _ffmpeg_has_filter(ffmpeg, "ass"): |
| 922 | raise RuntimeError( |
| 923 | "ffmpeg was found, but its libass/ass filter is unavailable. " |
| 924 | "Install an ffmpeg build with libass support before burning ASS subtitles." |
| 925 | ) |
| 926 | if not os.path.exists(video_path): |
| 927 | raise FileNotFoundError(f"Video not found: {video_path}") |
| 928 | if not os.path.exists(subtitle_path): |
| 929 | raise FileNotFoundError(f"Subtitle not found: {subtitle_path}") |
| 930 | |
| 931 | os.makedirs(os.path.dirname(output_path), exist_ok=True) |
| 932 | logger.info("Burning ASS subtitles: video=%s subtitles=%s -> %s", video_path, subtitle_path, output_path) |
| 933 | cmd = [ |
| 934 | ffmpeg, |
| 935 | "-y", |
| 936 | "-i", |
| 937 | video_path, |
| 938 | "-vf", |
| 939 | f"ass=filename='{_ffmpeg_filter_path(subtitle_path)}'", |
| 940 | "-c:v", |
| 941 | "libx264", |
| 942 | "-crf", |
| 943 | "18", |
| 944 | "-preset", |
| 945 | "veryfast", |
| 946 | "-c:a", |
| 947 | "copy", |
| 948 | output_path, |
| 949 | ] |
| 950 | try: |
| 951 | subprocess.run(cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True) |
| 952 | except subprocess.CalledProcessError as exc: |
| 953 | logger.error("Failed to burn ASS subtitles: %s", exc.stderr) |
| 954 | raise |
| 955 | return output_path |
| 956 | |
| 957 | |
| 958 | def speed_audio_to_duration(audio_path: str, output_path: str, target_seconds: int) -> str: |
| 959 | ffmpeg = shutil.which("ffmpeg") |
| 960 | if not ffmpeg: |
| 961 | raise RuntimeError("ffmpeg is required to speed up long digital-human narration audio.") |
| 962 | |
| 963 | duration = media_duration_seconds(audio_path) |
| 964 | if not duration: |
| 965 | return audio_path |
| 966 | if duration <= target_seconds: |
| 967 | return audio_path |
| 968 | |
| 969 | speed = duration / float(target_seconds) |
| 970 | filters = [] |
| 971 | remaining = speed |
| 972 | while remaining > 2.0: |
| 973 | filters.append("atempo=2.0") |
| 974 | remaining /= 2.0 |
| 975 | filters.append(f"atempo={remaining:.6f}") |
| 976 | |
| 977 | os.makedirs(os.path.dirname(output_path), exist_ok=True) |
| 978 | logger.info( |
| 979 | "Speeding audio to fit video duration: %s duration=%.2fs target=%ss speed=%.3fx -> %s", |
| 980 | audio_path, |
| 981 | duration, |
| 982 | target_seconds, |
| 983 | speed, |
| 984 | output_path, |
| 985 | ) |
| 986 | cmd = [ |
| 987 | ffmpeg, |
| 988 | "-y", |
| 989 | "-i", |
| 990 | audio_path, |
| 991 | "-filter:a", |
| 992 | ",".join(filters), |
| 993 | "-vn", |
| 994 | "-acodec", |
| 995 | "libmp3lame", |
| 996 | output_path, |
| 997 | ] |
| 998 | subprocess.run(cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE) |
| 999 | return output_path |
| 1000 | |
| 1001 | |
| 1002 | def extract_last_frame(video_path: str, output_path: str) -> str: |
| 1003 | ffmpeg = shutil.which("ffmpeg") |
| 1004 | if not ffmpeg: |
| 1005 | raise RuntimeError("ffmpeg is required to extract the previous video tail frame.") |
| 1006 | |
| 1007 | os.makedirs(os.path.dirname(output_path), exist_ok=True) |
| 1008 | logger.info("Extracting tail frame: %s -> %s", video_path, output_path) |
| 1009 | cmd = [ |
| 1010 | ffmpeg, |
| 1011 | "-y", |
| 1012 | "-sseof", |
| 1013 | "-0.1", |
| 1014 | "-i", |
| 1015 | video_path, |
| 1016 | "-frames:v", |
| 1017 | "1", |
| 1018 | "-q:v", |
| 1019 | "2", |
| 1020 | output_path, |
| 1021 | ] |
| 1022 | subprocess.run(cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE) |
| 1023 | return output_path |
| 1024 |