How to apply this guide
Switch between English and a supported speech language, keep the original transcript, and verify code-switched names and terms.
Start with
Set the language for the main sentence when possible. For a difficult recording, split it at a natural language change and compare the results to one continuous capture. Use both approaches on a disposable sample before settling on a workflow.
Check before you finish
Store or copy the original when you need a record. Apply formatting separately. The final result should preserve names, technical terms, numbers and the intended language. If it does not, edit the draft or deliver the original.
Evidence to keep: Keep the target field open and verify that the complete reviewed result reached the intended destination. This page’s key limit is: Switch between English and a supported speech language, keep the original transcript, and verify code-switched names and terms.
Start here
Bilingual work often includes names and technical terms from more than one language. A model listing two languages is not proof that it handles every code-switched sentence well. Create a short repeatable test that matches how you actually speak.
Choose the dominant language
Set the language for the main sentence when possible. For a difficult recording, split it at a natural language change and compare the results to one continuous capture. Use both approaches on a disposable sample before settling on a workflow.
Check vocabulary and script
Use dictionary entries for fixed names or phrases where the edition supports them. Review the spelling by comparing the transcript with your reference. Do not replace a word globally if that could alter unrelated contexts.
Keep the source through cleanup
Store or copy the original when you need a record. Apply formatting separately. The final result should preserve names, technical terms, numbers and the intended language. If it does not, edit the draft or deliver the original.




