Why this happens
By default, Speak AI auto-detects the language of each recording. Detection can land on the wrong language when a file is short, has music or noise at the start, or mixes languages. When that happens, the transcript comes back in the wrong language. You have three ways to control it.
1. Set the language at upload
On the upload screen, each file has a Language setting in the file list. It defaults to Auto-detect. Open it and choose the correct language before you start processing, and Speak AI will transcribe in that language.
2. Save a default language
If you mostly work in one language, save it so every new upload starts there instead of relying on auto-detection:
Set Media Language to your language. This is the default source language for transcription and analysis of your uploads.
3. Re-transcribe a file you already uploaded
If a file already processed in the wrong language, re-transcribe it:
Open the media file and choose Re-transcribe from the actions menu.
Set the Target Language and confirm.
Re-transcribing creates a new version of the media and runs the full pipeline again, so it uses transcription hours just like a new upload.
Tips for accurate detection
Make sure speech starts early in the file. Long musical or silent intros throw detection off.
For mixed-language recordings, set the dominant language manually instead of using Auto-detect.
The fastest way to reach us is the live chat in the app (the chat bubble in the bottom corner). You can also email [email protected].