When translating your own voice to text we recommend wearing headphones and when translating a few people’s conversations to text we recommend using an interface microphone with less interference between the direct sound and the first reflection. Speak slowly and clearly in quiet situations. If there is noise such as TV or family conversations, the microphone will pick up the sound and the accuracy of the transcription will decrease.

The human ear can only discern necessary sounds even in noise but speech recognition software doesn’t work that way. Try to speak as slowly and clearly as possible while capturing audio. Don’t speak in words. Speak in sentences. Speaking word for word may reduce the accuracy of transcription.

This is because there are many homonyms in Japanese that speech recognition software judges sentence by sentence rather than word by word. Also if you speak verbatim even the same speaker may be identified as a different speaker. When transcribing, understand the context as much as possible and speak sentence by sentence.