Fix Transcript & Speaker Errors

What to expect from AI transcription

AI transcription is very good, but it isn’t perfect. Words that sound alike, specialist terms, and names of people and places are the most common mistakes.

Correct a word or name

1. Open the meeting’s Transcript tab.
2. Highlight the word or name, up to four words.
3. Select Replace and enter the correct text.

Every whole-word match is corrected, and the fix also shows in the summary, action items, minutes and exports.

Beyond that, transcript text can’t be edited. Filler words and background noise can’t be removed, but Echo can write a cleaned-up version for you in chat.

Fix speaker labels

Speaker separation can struggle when voices sound similar or overlap. One person may get two labels, or two people may share one. You can rename a speaker and merge labels that belong to the same person, but one label can’t be split into two people. See how to review speaker labels.

Check the language

Choosing the right language matters more than anything else you can do. Choose the main language of the meeting, or use Auto-detect if you’re not sure. See how languages work.

Help NoteWave with specialist words

Add up to 100 names, acronyms and terms to your Custom Vocabulary before a meeting. For meetings full of technical, medical or legal language, plan time to review the transcript.

When to try again

• The wrong language was used: upload the recording again with the right language selected.
• The audio is poor: use a better copy if you have one. See how to fix poor audio.
• Processing failed: upload the recording again.
• It keeps happening: contact support with the meeting title and date.

What’s hardest to transcribe

• Very strong background noise or poor audio
• Whispered or very quiet speech
• Very short phrases
• Several languages mixed within one sentence
• Songs and music