199 lines
8.4 KiB
Python
199 lines
8.4 KiB
Python
# -*- coding: utf-8 -*-
|
||
"""KTV产线交付质检Worker — 4项QA检查"""
|
||
import json, os, subprocess, re
|
||
|
||
def run_ffprobe(filepath, args):
|
||
cmd = ["ffprobe", "-v", "error"] + args + [filepath]
|
||
r = subprocess.run(cmd, capture_output=True, text=True)
|
||
return r.stdout.strip()
|
||
|
||
class VerifyWorker:
|
||
def __init__(self, longtasks):
|
||
self.lt = longtasks
|
||
|
||
async def run(self, task):
|
||
payload = task.payload
|
||
task_id = task.task_id
|
||
|
||
mtv_path = payload.get("mtv_path", "")
|
||
ktv_path = payload.get("ktv_path", "")
|
||
ass_path = payload.get("ass_path", "")
|
||
lyrics_path = payload.get("lyrics_path", "")
|
||
original_duration = float(payload.get("original_duration", 0))
|
||
|
||
# 检查文件是否存在
|
||
missing = []
|
||
for label, p in [("MTV", mtv_path), ("KTV", ktv_path), ("ASS", ass_path), ("歌词", lyrics_path)]:
|
||
if p and not os.path.exists(p):
|
||
missing.append(f"{label}: {p}")
|
||
if missing:
|
||
return {"status": "FAILED", "error": "文件不存在", "missing": missing}
|
||
|
||
all_errors = []
|
||
qa_results = {}
|
||
|
||
# QA-1: 字幕时间精准性（从calibrated数据检查）
|
||
# 如果有calibrated_path，检查时间轴
|
||
calibrated_path = payload.get("calibrated_path", "")
|
||
if calibrated_path and os.path.exists(calibrated_path):
|
||
qa1_errors = self._check_timeline(calibrated_path)
|
||
qa_results["QA1_timeline"] = {"passed": len(qa1_errors) == 0, "errors": qa1_errors}
|
||
all_errors.extend(qa1_errors)
|
||
else:
|
||
qa_results["QA1_timeline"] = {"passed": True, "errors": [], "skipped": "no calibrated_path"}
|
||
|
||
# QA-2: 字幕歌词正确性
|
||
qa2_errors = self._check_subtitles(ass_path, lyrics_path)
|
||
qa_results["QA2_lyrics"] = {"passed": len(qa2_errors) == 0, "errors": qa2_errors}
|
||
all_errors.extend(qa2_errors)
|
||
|
||
# QA-3: MTV使用原音频
|
||
if mtv_path and os.path.exists(mtv_path):
|
||
qa3_errors = self._check_mtv(mtv_path, original_duration)
|
||
qa_results["QA3_mtv"] = {"passed": len(qa3_errors) == 0, "errors": qa3_errors}
|
||
all_errors.extend(qa3_errors)
|
||
else:
|
||
qa_results["QA3_mtv"] = {"passed": True, "errors": [], "skipped": "no MTV file"}
|
||
|
||
# QA-4: KTV双轨音序
|
||
if ktv_path and os.path.exists(ktv_path):
|
||
qa4_errors = self._check_ktv(ktv_path, original_duration)
|
||
qa_results["QA4_ktv"] = {"passed": len(qa4_errors) == 0, "errors": qa4_errors}
|
||
all_errors.extend(qa4_errors)
|
||
else:
|
||
qa_results["QA4_ktv"] = {"passed": True, "errors": [], "skipped": "no KTV file"}
|
||
|
||
# 汇总
|
||
passed = len(all_errors) == 0
|
||
return {
|
||
"status": "PASSED" if passed else "FAILED",
|
||
"task_id": task_id,
|
||
"qa_results": qa_results,
|
||
"total_errors": len(all_errors),
|
||
"errors": all_errors
|
||
}
|
||
|
||
def _check_timeline(self, calibrated_path):
|
||
"""QA-1: 字幕时间精准性 — 时间单调递增、无重叠"""
|
||
errors = []
|
||
try:
|
||
with open(calibrated_path, 'r') as f:
|
||
data = json.load(f)
|
||
lyrics = data.get("lyrics", data.get("segments", data))
|
||
if not isinstance(lyrics, list):
|
||
return ["calibrated数据格式错误，应为list"]
|
||
|
||
for i in range(len(lyrics)):
|
||
seg = lyrics[i]
|
||
start = seg.get("start", 0)
|
||
end = seg.get("end", 0)
|
||
if end <= start:
|
||
errors.append(f"行{i+1}: end({end:.2f})<=start({start:.2f})")
|
||
if i > 0:
|
||
prev_end = lyrics[i-1].get("end", 0)
|
||
if start < prev_end - 0.1:
|
||
errors.append(f"行{i+1}: start({start:.2f})与前一行end({prev_end:.2f})重叠")
|
||
except Exception as e:
|
||
errors.append(f"读取calibrated文件失败: {e}")
|
||
return errors
|
||
|
||
def _check_subtitles(self, ass_path, lyrics_path):
|
||
"""QA-2: 字幕歌词正确性 — ASS中歌词与原始歌词逐行比对"""
|
||
errors = []
|
||
if not ass_path or not lyrics_path:
|
||
return errors
|
||
if not os.path.exists(ass_path):
|
||
return [f"ASS文件不存在: {ass_path}"]
|
||
if not os.path.exists(lyrics_path):
|
||
return [f"原始歌词文件不存在: {lyrics_path}"]
|
||
|
||
with open(lyrics_path, 'r') as f:
|
||
original_lines = [l.strip() for l in f.readlines() if l.strip()]
|
||
|
||
with open(ass_path, 'r') as f:
|
||
ass_content = f.read()
|
||
|
||
karaoke_lines = []
|
||
for line in ass_content.split('\n'):
|
||
if 'Karaoke' in line and 'Dialogue' in line:
|
||
parts = line.split(',', 9)
|
||
if len(parts) >= 10:
|
||
text = parts[9]
|
||
clean = re.sub(r'\{\\kf\d+\}', '', text).strip()
|
||
if clean:
|
||
karaoke_lines.append(clean)
|
||
|
||
if len(karaoke_lines) != len(original_lines):
|
||
errors.append(f"歌词行数不匹配: ASS={len(karaoke_lines)}行, 原始={len(original_lines)}行")
|
||
|
||
mismatch_count = 0
|
||
for i in range(min(len(karaoke_lines), len(original_lines))):
|
||
kl = karaoke_lines[i].replace(' ', '')
|
||
ol = original_lines[i].replace(' ', '')
|
||
if kl != ol:
|
||
mismatch_count += 1
|
||
if mismatch_count <= 5:
|
||
errors.append(f"行{i+1}: ASS='{kl}' vs 原始='{ol}'")
|
||
if mismatch_count > 5:
|
||
errors.append(f"... 还有{mismatch_count-5}处不匹配")
|
||
|
||
return errors
|
||
|
||
def _check_mtv(self, path, original_duration):
|
||
"""QA-3: MTV使用原音频（单轨，duration匹配）"""
|
||
errors = []
|
||
try:
|
||
streams = run_ffprobe(path, [
|
||
"-show_entries", "stream=index,codec_type,duration",
|
||
"-show_entries", "stream_tags=handler_name",
|
||
"-of", "json"
|
||
])
|
||
data = json.loads(streams)
|
||
audio_streams = [s for s in data.get("streams", []) if s.get("codec_type") == "audio"]
|
||
if len(audio_streams) != 1:
|
||
errors.append(f"MTV应有1条音轨，实际{len(audio_streams)}条")
|
||
if audio_streams and original_duration > 0:
|
||
dur = float(audio_streams[0].get("duration", 0))
|
||
if abs(dur - original_duration) > 2:
|
||
errors.append(f"音频时长{dur:.1f}s与原曲{original_duration:.1f}s差距>2s")
|
||
|
||
fmt = run_ffprobe(path, ["-show_entries", "format=duration", "-of", "csv=p=0"])
|
||
if fmt and original_duration > 0:
|
||
video_dur = float(fmt)
|
||
if abs(video_dur - original_duration) > 3:
|
||
errors.append(f"视频时长{video_dur:.1f}s与原曲{original_duration:.1f}s差距>3s")
|
||
except Exception as e:
|
||
errors.append(f"ffprobe失败: {e}")
|
||
return errors
|
||
|
||
def _check_ktv(self, path, original_duration):
|
||
"""QA-4: KTV双轨音序 — Track1=伴奏, Track2=原唱"""
|
||
errors = []
|
||
try:
|
||
streams = run_ffprobe(path, [
|
||
"-show_entries", "stream=index,codec_type,duration",
|
||
"-show_entries", "stream_tags=handler_name",
|
||
"-of", "json"
|
||
])
|
||
data = json.loads(streams)
|
||
audio_streams = [s for s in data.get("streams", []) if s.get("codec_type") == "audio"]
|
||
if len(audio_streams) != 2:
|
||
errors.append(f"KTV应有2条音轨，实际{len(audio_streams)}条")
|
||
return errors
|
||
|
||
t1 = audio_streams[0].get("tags", {}).get("handler_name", "")
|
||
if "伴奏" not in t1 and "Accompaniment" not in t1:
|
||
errors.append(f"Track1应为伴奏，实际标签: '{t1}'")
|
||
t2 = audio_streams[1].get("tags", {}).get("handler_name", "")
|
||
if "原唱" not in t2 and "Original" not in t2:
|
||
errors.append(f"Track2应为原唱，实际标签: '{t2}'")
|
||
|
||
if original_duration > 0:
|
||
for i, a in enumerate(audio_streams):
|
||
dur = float(a.get("duration", 0))
|
||
if abs(dur - original_duration) > 2:
|
||
errors.append(f"音轨{i+1}时长{dur:.1f}s与原曲{original_duration:.1f}s差距>2s")
|
||
except Exception as e:
|
||
errors.append(f"ffprobe失败: {e}")
|
||
return errors
|