单词级字幕重新字幕优化后出现较多时间戳与单词级不一致的情况 #883
Description
问题描述 | Problem Description
我在使用1.4版本利用字幕优化功能使用deepseek api进行英文单词级字幕重新组句（单句最多12词）的时候出现组句后较多时间戳与单词级不一致的情况
日志信息（可选）| Logs (Optional)
正在字幕断句...
根据字数 453,确定断句分段数: 1
开始调用API进行分段,文本长度: 453
2025-11-18 14:52:56 - split_by_llm - WARNING - 断句验证失败，开始反馈循环 (第1次尝试): Length violations:
- Segment 18 'So if you look at this little area here,...': 13 words > 12 limit
- Segment 31 'Remember I was telling you this candlest...': 16 words > 12 limit
- Segment 39 'So once we get below these two lows here...': 17 words > 12 limit
- Segment 44 'Now if all things being equal,that being...': 13 words > 12 limit
- Segment 47 'If those are true,then my stop loss is s...': 16 words > 12 limit
- ... and 4 more segments too long
Split these long segments further with <br>.
2025-11-18 14:53:29 - split_by_llm - WARNING - 断句验证失败，开始反馈循环 (第2次尝试): Length violations:
- Segment 18 'So if you look at this little area here,...': 13 words > 12 limit
- Segment 52 'then my stop loss is safe going just abo...': 13 words > 12 limit
- Segment 53 'So I'm going to put it right inside the ...': 14 words > 12 limit
- Segment 60 'that's a dog whistle or it's a going to ...': 16 words > 12 limit
- Segment 67 'Cause it's so close in proximity to this...': 13 words > 12 limit
- ... and 1 more segments too long
Split these long segments further with <br>.