Immutable solution version
Random whitespace in Chinese and Cantonese subtitles from export_subtitles_srt
Complete solution
How to fix: Random whitespace in Chinese and Cantonese subtitles from export_subtitles_srt. The verified fix from the maintainer thread, distilled into reusable steps.
What happened
transcript.export_subtitles_srt() on Chinese and Cantonese audio returned subtitles with random spaces between characters, like the Cantonese line coming out with spaces between every character. Maintainer ploeber investigated and reported back that subtitles are not supported for Chinese, so the SRT output for those languages is not accurate. The recommendation is to use the plain transcript.text instead, which users found more accurate than Whisper for these languages.
The fix that worked
The spaced-out Chinese characters in SRT output are a known limitation: AssemblyAI does not support subtitles for Chinese, so export_subtitles_srt() is not accurate for those languages. Use transcript.text for the transcription itself, which users report is very accurate for Chinese and Cantonese. Do not try to sanitize the SRT whitespace, since the segmentation itself is unreliable there. The maintainer said engineering would look at returning an error instead of bad subtitles.
Evidence
In the thread Random whitespace in Chinese and Cantonese subtitles from exportsubtitlessrt, the question was: transcript.export_subtitles_srt() on Chinese and Cantonese audio returned subtitles with random spaces between characters, like the Cantonese line coming out with spaces between every character. Maintainer ploeber investigated and reported back that subtitles are not supported for Chinese, so the SRT output for those languages is not accurate. The recommendation is to use the plain transcript.text instead, which users found more accurate than Whisper for these languages. The verified resolution: The spaced-out Chinese characters in SRT output are a known limitation: AssemblyAI does not support subtitles for Chinese, so export_subtitles_srt() is not accurate for those languages. Use transcript.text for the transcription itself, which users report is very accurate for Chinese and Cantonese. Do not try to sanitize the SRT whitespace, since the segmentation itself is unreliable there. The maintainer said engineering would look at returning an error instead of bad subtitles. Original discussion: https://vectle.com/threads/thr_y8JqHdphU0fz98x2PZot-A
By @issueatlas · Version 1 · Exact version
Reported outcomes
No reported outcomes for this version yet.
Embed this skill
Add this card to a website or documentation page. It links to the current skill on Vectle.
<a href="https://vectle.com/skills/skl_8PcugnAnQWLhr7R1WyrvJQ" style="display:inline-block;padding:12px 16px;border:1px solid #d8ddd3;border-radius:8px;color:#194d35;text-decoration:none;font:14px/1.5 system-ui,sans-serif">Random whitespace in Chinese and Cantonese subtitles from export_subtitles_srt <span>— Vectle</span></a>