How Chinese English mixed transcription works
In many workplaces, universities and families, Mandarin speakers mix in English all the time. A product manager says a sentence in Chinese with "deadline", "KPI" and "follow up" in the middle. A student describes a "paper" and a "presentation". A podcast host switches to English for a quote and back again. Tools that expect one language at a time tend to mangle these words or force them into Chinese characters.
RocketScribe handles this with the language set to Chinese, which this page does for you. The recognition model writes the Chinese parts in Chinese characters, normally Simplified, and keeps English words, product names and acronyms spoken inside Chinese sentences in Latin letters in most cases. You get one transcript in the order things were said.
Where mixed recordings get harder
Switching back and forth is harder for any speech model than staying in one language. Be ready to check these spots:
- Short English words said quickly with a Chinese accent can be misheard or written as Chinese characters with a similar sound.
- Long stretches of pure English inside a Chinese recording often come out fine, but occasionally the model writes a passage in the wrong language or skips part of it. Listen to these sections.
- Capitalisation and spacing around English words may be inconsistent. A quick find in your exported document helps tidy brand names.
- Chinese homophones still apply: a person's Chinese name can appear with the right sound but the wrong characters.
Choosing Chinese, not auto-detect
Auto-detect decides the language from the start of the file. A meeting that opens with "OK, let's start" can be detected as English, and then the Chinese parts suffer. Setting Chinese up front avoids that. If a recording is mostly English with only a few Chinese phrases, set English instead and fix the Chinese phrases by hand; the language should match whatever covers most of the speech.
There is no separate mixed-language mode and no custom vocabulary list. The language you pick is a hint for the whole file, and the model does the rest.
For a recording where whole sections are in one language, such as a Chinese talk followed by an English Q&A, you get better results by cutting it into two files and running each with its own language. For a conversation that switches every few sentences, keep it as one file with Chinese selected.
Typical recordings
Mixed transcription is common for team meetings at companies with offices in China and abroad, interviews with Chinese engineers or researchers, lectures at universities where slides are in English and explanation is in Mandarin, and Chinese-language podcasts about tech, finance or study abroad. It also suits family video calls where children answer in English and parents speak Mandarin.
For meetings and interviews with several voices, speaker identification on paid plans or with a Minute Pack labels who said what. On the Free plan and the no-sign-up trial you get the text and timestamps without speaker labels.
Exports for bilingual transcripts
TXT and DOCX keep both Chinese characters and English text for meeting notes or quotes. SRT and VTT work for subtitles where the on-screen text mixes both languages, which many bilingual viewers prefer to a full translation. CSV gives one row per segment if you want to tag topics. The direct PDF supports Western European characters only, so with Chinese text the PDF option opens a print-ready page that you save as PDF from the browser.
To share the content with colleagues who only read English, open the Google Translate link from the transcript, or use the AI prompt export to paste it into ChatGPT or Claude and ask for an English summary of the decisions and action items.