Voice Input zh-TW still transcribes into Simplified Chinese

Support Bug Reports
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

Speech language is set to Traditional Chinese (Taiwan) / zh-TW, but the transcribed text in the chat input is Simplified Chinese.

Example: speaking Taiwan Mandarin produces 语音 / 翻译 / 时候 / 选 / 繁体 / 怎么 / 写 / 简体 instead of 語音 / 翻譯 / 時候 / 選 / 繁體 / 怎麼 / 寫 / 簡體.

zh-TW appears to be used only as an accent hint. The character set still follows zh-CN / Simplified Chinese. This happens in Agent chat Voice Input (microphone, Ctrl+Shift+Space).

Steps to Reproduce

  1. Open Agent chat in Cursor IDE.
  2. Open Voice settings next to the microphone.
  3. Select Traditional Chinese (Taiwan) / zh-TW.
  4. Press the mic (or Ctrl+Shift+Space) and speak Mandarin as used in Taiwan.
  5. Check the text inserted into the input box — it is Simplified Chinese.

Expected Behavior

Transcription should use Traditional Chinese characters (zh-TW), matching the selected voice language.

Operating System

Windows 10/11

Version Information

Cursor 3.20.17
Windows 11 (10.0.26200)

Additional Information

This is not the AI reply language. The Voice Input box itself already inserts Simplified Chinese before sending.

Workaround today: convert the text manually, or type instead of using the mic.

Does this stop you from using Cursor

No - Cursor works, but with this issue

5

Hey, thanks for the detailed report and the specific character examples, it really helps.

You got it right. Right now, selecting Traditional Chinese Taiwan zh-TW or zh-HK only affects speech recognition. The returned text is always Simplified, no matter that setting.

I’ve passed this to the team. There isn’t a setting that changes this today, so for now the only practical options are the ones you mentioned, manually convert the text or type it. I’ll reply in the thread once there’s an update.

8

Hey @arturwu. There’s an update on this thread.

The Traditional Chinese transcription issue is fixed on our side. Now, when zh-TW or zh-HK is selected, the text is converted to Traditional characters (Taiwan or Hong Kong variants, respectively) before it reaches the input.

This is a server-side change, so you don’t need to update anything on your end. It should roll out soon, if it hasn’t already. Could you please try Voice Input again with zh-TW selected and let us know if you still see Simplified characters (for example, 语音 instead of 語音).

Thanks again for the detailed report and the concrete examples. They really helped.