If your video or audio mixes two languages (for example, English and Cantonese, or Arabic and English alternating), you may find after uploading that only one of the languages made it into the subtitles. Here is how this currently works, and what you can do about it.
File-upload modes transcribe using one "source language" by default
In the file-upload-based modes ([Subtitle generation], [Transcript], [Meeting Note]), the system asks you to choose a [Source language] before starting and by default transcribes the entire content according to that one language — parts not in the selected source language may not be transcribed accurately. If the content mixes Chinese and English or alternates between several languages, choose your source language and then click [Add a second language] to turn on Multiple languages mode, which lets you select up to 10 languages. Multiple languages mode is available on paid plans only (the Free plan shows [Upgrade required]); it automatically uses the Enhanced engine and deducts minutes at 1.5 times the standard rate, and going back to one language restores the Standard engine.
"Translation" is not the same as "bilingual verbatim transcription"
Note that the [Add additional translation languages] / [Translate] feature translates the already-transcribed original text into another language — it does not transcribe each of the two languages mixed in the audio as actually spoken. If what you want is to faithfully preserve the two languages the speaker actually used, the translation feature won't achieve that; use Multiple languages mode (above) instead.
What you can do today
If you stay with one language, choose the one that makes up more of the content as the source language. Selecting the language with the larger share of the video as the [Source language] usually gives you a more complete transcription.
Fix the mixed-language parts by hand in the editor. After transcription, you can edit the incorrectly transcribed mixed-language passages line by line in the Subanana editor, without re-uploading the file.
If the content splits cleanly by language, consider splitting the file yourself first, then uploading the parts separately with the matching source language for each. This is a situational suggestion, not a required step; if the languages alternate frequently and are hard to split cleanly, using Multiple languages mode, or uploading with the main language and fixing things in the editor, is usually more practical.
Mixed-language speech (code-switching) is itself one of the audio characteristics most likely to cause gaps in transcription. If the transcription has obvious gaps, you can also go through the rating flow in that video's editor and choose [Not ideal] → [Subtitles are missing a lot of content]. The system keeps your existing subtitle track and re-transcribes a new subtitle track for that video using more suitable settings; this reprocessing does not use any of your minutes.
Live caption translation mode can follow language switches automatically
If what you need is live captions for an in-person event (rather than an uploaded file), the [Event Captions] mode — with the source language set to auto-detect — can automatically recognize and follow the language when a fluent speaker switches languages mid-sentence, with no manual action needed. For multilingual events with clearly separated segments, we recommend the operator switch the source language manually at each segment boundary, which is more stable than relying on auto-detection for long stretches. This auto-follow capability applies only to the live captions mode, not to file-upload transcription (use Multiple languages mode there).
If you still have questions, reach us via the in-app live chat or by email at hello@subanana.com — we usually reply within one business day (Monday to Friday, Hong Kong time).