中文的文本生成的字幕，词语分割点总是在奇怪的位置，本来是一个词语应该在一条字幕内显示，而现在生成的字幕会把词语拆成两条字幕显示，不知道--write-subtitles命令中文本是如何分割的，我把中文标点符号替换成英文标点符号，结果还是一样

[xiaomielaidi]

中文的文本生成的字幕，词语分割点总是在奇怪的位置，本来是一个词语应该在一条字幕内显示，而现在生成的字幕会把词语拆成两条字幕显示，不知道--write-subtitles命令中文本是如何分割的，我把中文标点符号替换成英文标点符号，结果还是一样

The subtitles generated from Chinese text always have word breaks at strange positions. Originally, a single word should appear in one subtitle line, but now the generated subtitles split the words into two lines. I'm not sure how the text is segmented in the --write-subtitles command. Even after replacing Chinese punctuation marks with English ones, the issue remains the same.

[rany2]

Hello, can you try this branch for me? https://github.com/rany2/edge-tts/tree/wip-subtitles

[xiaomielaidi]

Hello, can you try this branch for me? https://github.com/rany2/edge-tts/tree/wip-subtitles

Do I need to reinstall this branch of edge-tts? I tried the command in the cmd window, but the result was unsatisfactory:edge-tts --voice zh-CN-YunxiNeural --text "萧漪气得粉拳挥舞,二师兄,你的嘴巴也太讨厌了.谁痴呆,谁脑残了.真是的,二师兄,你这样子,没有女孩子会喜欢你.吕少卿满脸不屑,爱情什么的,有灵石重要吗?谁要女孩子喜欢了?麻烦!" --write-media D:\hello.mp3 --write-subtitles D:\hello.vtt
The subtitles I generated are as follows .I've extracted the first few sentences:
WEBVTT

00:00:00.100 --> 00:00:03.000

萧漪 气 得 粉 拳 挥舞 二师兄 你 的 嘴巴

00:00:03.025 --> 00:00:06.638

也 太 讨厌 了 谁 痴呆 谁 脑残 了 真是的

00:00:06.912 --> 00:00:11.062

二师兄 你 这 样子 没有 女孩子 会 喜欢 你 吕少卿

00:00:11.062 --> 00:00:14.325

满 脸 不屑 爱情 什么的 有 灵石 重要 吗 谁

00:00:14.325 --> 00:00:15.900

要 女孩子 喜欢 了 麻烦

========================================
If the generated subtitles are segmented as shown below, it's considered quite good:

1
00:00:00,100 --> 00:00:01,525
萧漪气得粉拳挥舞

2
00:00:01,825 --> 00:00:02,437
二师兄

3
00:00:02,799 --> 00:00:04,299
你的嘴巴也太讨厌了

4
00:00:04,862 --> 00:00:05,587
谁痴呆

5
00:00:05,900 --> 00:00:06,862
谁脑残了

6
00:00:07,412 --> 00:00:08,087
真是的

7
00:00:08,349 --> 00:00:08,900
二师兄

8
00:00:09,262 --> 00:00:10,175
你这样子

9
00:00:10,500 --> 00:00:11,637
没有女孩子会喜欢你

10
00:00:12,199 --> 00:00:13,449
吕少卿满脸不屑

11
00:00:13,750 --> 00:00:15,937
爱情什么的有灵石重要吗

12
00:00:16,500 --> 00:00:17,699
谁要女孩子喜欢了

13
00:00:18,262 --> 00:00:18,837
麻烦

[rany2]

You need to reinstall from that branch, I believe you didn't do so which is why the same behaviour remained.

[xiaomielaidi]

You need to reinstall from that branch, I believe you didn't do so which is why the same behaviour remained.

After testing, I found that the subtitle format did change, and it even retained punctuation marks. However, I'm not sure where the problem lies because the subtitles didn't start a new line at punctuation marks. Generally, it's more reasonable for subtitles to move to the next line at pauses in the audio to avoid making a subtitle too long and cluttered. Was it supposed to be like this originally?

The following is the subtitle I obtained using the command (edge-tts --voice zh-CN-YunxiNeural --text "萧漪气得粉拳挥舞，二师兄，你的嘴巴也太讨厌了。谁痴呆，谁脑残了。真是的，二师兄，你这样子，没有女孩子会喜欢你。吕少卿满脸不屑，爱情什么的，有灵石重要吗？谁要女孩子喜欢了？麻烦！" --write-media D:\hello.mp3 --write-subtitles D:\hello.vtt):

1
00:00:00,100 --> 00:00:03,525
萧漪气得粉拳挥舞，二师兄，你的嘴巴

2
00:00:03,550 --> 00:00:08,250
也太讨厌了。谁痴呆，谁脑残了。真是的

3
00:00:08,550 --> 00:00:13,112
二师兄，你这样子，没有女孩子会喜欢你。吕少卿

4
00:00:13,137 --> 00:00:17,212
满脸不屑，爱情什么的，有灵石重要吗？谁

5
00:00:17,212 --> 00:00:19,300
要女孩子喜欢了？麻烦

[rany2]

Makes sense, thanks for the feedback.

[kaa99]

Hi rany2, IIUC, you are generating the subtitle using results with wordBoundaryEnabled. I wonder have you tried getting subtitle with sentenceBoundaryEnabled. that looks like a promising way to solve the issue listed above.

[rany2]

Unfortunately sentence boundary has many bugs if proper punctuation isn't provided and is deprecated by Microsoft themselves.