Skip to content

利用 AI 大模型和自动化工作流,根据主题或关键词一键生成高清短视频。Generate HD short videos from a topic or keyword with an automated AI workflow.

fix(subtitle): preserve body timecodes across caption consumers

Rudy Celekli6007797295f7
diff --git a/app/services/subtitle.py b/app/services/subtitle.py
index 71663877..5bb48b67 100644
--- a/app/services/subtitle.py
+++ b/app/services/subtitle.py
@@ -1,6 +1,5 @@
import json
import os.path
-import re
import tempfile
import threading
from timeit import default_timer as timer
@@ -14,6 +13,7 @@ from loguru import logger
from app.config import config
from app.models import const
from app.utils import utils
+from app.utils.subtitle_reader import file_to_subtitles as file_to_subtitles
from app.utils.subtitle_writer import write_subtitle_file
model_size = config.whisper.get("model_size", "large-v3")
@@ -183,37 +183,6 @@ def create(
logger.info(f"subtitle file created: {subtitle_file}")
-def file_to_subtitles(filename):
- if not filename or not os.path.isfile(filename):
- return []
-
- times_texts = []
- current_times = None
- current_text = ""
- index = 0
- with open(filename, "r", encoding="utf-8") as f:
- for line in f:
- times = re.findall("([0-9]*:[0-9]*:[0-9]*,[0-9]*)", line)
- # Once a cue has started, timestamps belong to its text until the
- # blank-line separator; they must not overwrite the cue's timing.
- if times and current_times is None:
- current_times = line
- elif line.strip() == "" and current_times:
- index += 1
- times_texts.append((index, current_times.strip(), current_text.strip()))
- current_times, current_text = None, ""
- elif current_times:
- current_text += line
-
- # Flush the final block. SRT files whose last subtitle is not followed by a
- # trailing blank line never hit the blank-line branch above, so without this
- # the last subtitle would be silently dropped.
- if current_times:
- index += 1
- times_texts.append((index, current_times.strip(), current_text.strip()))
- return times_texts
-
-
def transcribe_audio_bytes(audio_bytes: bytes) -> str:
"""Transcribe in-memory WAV audio and remove every temporary artifact."""
if not isinstance(audio_bytes, bytes) or not audio_bytes:
diff --git a/app/services/video.py b/app/services/video.py
index 9b4a5b05..7371f195 100644
--- a/app/services/video.py
+++ b/app/services/video.py
@@ -43,6 +43,7 @@ from app.models.schema import (
from app.services import bgm as bgm_service
from app.services.utils import video_effects
from app.utils import file_security, logging_utils, utils
+from app.utils.subtitle_reader import moviepy_subtitles
class SubClippedVideoClip:
def __init__(
@@ -1779,7 +1780,7 @@ def generate_video(
if params.subtitle_enabled and subtitle_path and os.path.exists(subtitle_path):
sub = clip_stack.enter_context(
SubtitlesClip(
- subtitles=subtitle_path,
+ subtitles=moviepy_subtitles(subtitle_path),
encoding="utf-8",
make_textclip=make_textclip,
)
diff --git a/app/services/voice.py b/app/services/voice.py
index f69c0b97..0e12c486 100644
--- a/app/services/voice.py
+++ b/app/services/voice.py
@@ -24,13 +24,13 @@ import requests
from edge_tts import SubMaker
from edge_tts.srt_composer import Subtitle
from loguru import logger
-from moviepy.video.tools import subtitles
from moviepy.audio.io.AudioFileClip import AudioFileClip
from openai import OpenAI
from app.config import config
from app.services import bgm
from app.utils import utils
+from app.utils.subtitle_reader import moviepy_subtitles
from app.utils.subtitle_writer import staged_subtitle_file
_DEFAULT_EDGE_TTS_TIMEOUT_SECONDS = 30.0
@@ -3458,7 +3458,7 @@ def _write_subtitle_items(sub_items: list[str], subtitle_file: str) -> bool:
with staged_subtitle_file(subtitle_file) as staged:
with open(staged, "w", encoding="utf-8") as file:
file.write("\n".join(sub_items) + "\n")
- sbs = subtitles.file_to_subtitles(staged, encoding="utf-8")
+ sbs = moviepy_subtitles(staged)
if not sbs:
raise ValueError("subtitle output contains no cues")
duration = max(tb for ((ta, tb), txt) in sbs)
diff --git a/app/utils/subtitle_reader.py b/app/utils/subtitle_reader.py
new file mode 100644
index 00000000..c5f95a31
--- /dev/null
+++ b/app/utils/subtitle_reader.py
@@ -0,0 +1,45 @@
+"""Read SRT cue boundaries before adapting them to MoviePy timing tuples."""
+import os
+import re
+
+
+def file_to_subtitles(filename):
+ if not filename or not os.path.isfile(filename):
+ return []
+
+ times_texts = []
+ current_times = None
+ current_text = ""
+ index = 0
+ with open(filename, "r", encoding="utf-8") as f:
+ for line in f:
+ times = re.findall("([0-9]*:[0-9]*:[0-9]*,[0-9]*)", line)
+ # Once a cue has started, timestamps belong to its text until the
+ # blank-line separator; they must not overwrite the cue's timing.
+ if times and current_times is None:
+ current_times = line
+ elif line.strip() == "" and current_times:
+ index += 1
+ times_texts.append((index, current_times.strip(), current_text.strip()))
+ current_times, current_text = None, ""
+ elif current_times:
+ current_text += line
+
+ # Flush the final block. SRT files whose last subtitle is not followed by a
+ # trailing blank line never hit the blank-line branch above, so without this
+ # the last subtitle would be silently dropped.
+ if current_times:
+ index += 1
+ times_texts.append((index, current_times.strip(), current_text.strip()))
+ return times_texts
+
+
+def moviepy_subtitles(filename):
+ """Convert only cue headers; timestamp-bearing text remains caption text."""
+ from moviepy.tools import convert_to_seconds
+
+ cues = []
+ for _index, timing, text in file_to_subtitles(filename):
+ start, end = re.findall("([0-9]*:[0-9]*:[0-9]*,[0-9]*)", timing)
+ cues.append(((convert_to_seconds(start), convert_to_seconds(end)), text))
+ return cues
diff --git a/test/services/test_caption_timecodes.py b/test/services/test_caption_timecodes.py
new file mode 100644
index 00000000..719c9223
--- /dev/null
+++ b/test/services/test_caption_timecodes.py
@@ -0,0 +1,29 @@
+"""Native caption publication and final rendering must retain spoken timecodes."""
+
+import pytest
+from edge_tts import SubMaker
+
+from app.services import subtitle, voice
+
+
+CAPTIONS = [
+ "Jump to the recording.",
+ "Jump to 00:01:23,456 in the recording.",
+ "Show 00:01:23,456 --> 00:01:25,456 in the log.",
+]
+
+
[email protected]("text", CAPTIONS)
+def test_speech_caption_publication_preserves_timecodes(tmp_path, text):
+ maker = SubMaker()
+ maker.feed({
+ "type": "SentenceBoundary", "offset": 0,
+ "duration": 8000000, "text": text,
+ })
+ destination = tmp_path / "speech.srt"
+ voice.create_subtitle(maker, text, str(destination), word_level=True)
+
+ assert destination.is_file()
+ assert subtitle.file_to_subtitles(str(destination)) == [
+ (1, "00:00:00,000 --> 00:00:00,800", text)
+ ]
diff --git a/test/services/test_video_project.py b/test/services/test_video_project.py
index 2989066b..a6cdde8a 100644
--- a/test/services/test_video_project.py
+++ b/test/services/test_video_project.py
@@ -582,3 +582,61 @@ def test_dash_external_segments_cannot_escape_source_snapshot(native):
project.render()
assert project.status()["failed_stage"] == "opening:scene"
assert project.metadata()["last_successful_export"] is None
+
+
+ "Jump to the recording.",
+ "Jump to 00:01:23,456 in the recording.",
+ "Show 00:01:23,456 --> 00:01:25,456 in the log.",
+])
+def test_final_caption_render_keeps_body_timecodes_and_duration(tmp_path, text):
+ from moviepy import VideoFileClip
+
+ from app.config import config
+ from app.models.schema import VideoAspect, VideoParams
+ from app.services import video
+ from app.utils.subtitle_writer import write_subtitle_file
+
+ ffmpeg = shutil.which("ffmpeg")
+ if not ffmpeg:
+ pytest.skip("native FFmpeg executable is unavailable")
+ font = Path(config.root_dir) / "resource/fonts/Charm-Regular.ttf"
+ if not font.is_file():
+ pytest.skip("native caption font is unavailable")
+
+ base = [ffmpeg, "-nostdin", "-v", "error", "-y", "-threads", "1"]
+ media = tmp_path / "white.mp4"
+ audio = tmp_path / "narration.wav"
+ subprocess.run([
+ *base, "-f", "lavfi", "-i", "color=c=white:s=1080x1080:r=10:d=0.8",
+ "-c:v", "libx264", "-pix_fmt", "yuv420p", str(media),
+ ], check=True, timeout=30)
+ subprocess.run([
+ *base, "-f", "lavfi", "-i", "sine=frequency=220:duration=0.8",
+ "-ar", "44100", str(audio),
+ ], check=True, timeout=30)
+ artifact = tmp_path / "accepted.srt"
+ assert write_subtitle_file(
+ str(artifact), "1\n00:00:00,100 --> 00:00:00,700\n" + text + "\n\n"
+ )
+ output = tmp_path / "captioned.mp4"
+ params = VideoParams(
+ video_subject="owned-caption", video_aspect=VideoAspect.square,
+ subtitle_enabled=True, subtitle_position="center",
+ font_name="Charm-Regular.ttf", font_size=36,
+ text_fore_color="#000000", stroke_width=0, text_background_color=False,
+ bgm_type="", bgm_volume=0, voice_volume=1, n_threads=1,
+ )
+ previous = config.app.copy()
+ try:
+ config.app.update({"video_codec": "libx264", "video_clip_concurrency": 1})
+ assert video.generate_video(
+ str(media), str(audio), str(artifact), str(output), params
+ )
+ with VideoFileClip(str(output), audio=False) as rendered:
+ assert 0.7 <= rendered.duration <= 0.9
+ frame = rendered.get_frame(0.4)
+ assert int((frame.mean(axis=2) < 180).sum()) > 100
+ finally:
+ config.app.clear()
+ config.app.update(previous)