Guides › Editing speech

How to change a word in an audio recording, in your own voice

A course lesson says “at home” where it should say “with your phone”. A voice-over has last year's product name. A presentation recording has one wrong number. The rest of the recording is fine. You only need to change a word or two. Here are three ways to do it, and why it is usually better to replace the whole sentence than just the word.

Change the words, keep your voice: try Respeak free →

Why changing a single word is harder than it sounds

When we speak, every word is shaped by the words around it. Pitch rises and falls across the sentence, sounds run into each other, and the last word of a sentence is spoken differently from the same word in the middle. That is why a word cut from somewhere else, or recorded on its own, often sounds pasted in even when the voice is the same. The most natural fix is usually to replace the whole sentence, from one pause to the next.

Method 1: Splice the word from elsewhere in the recording

If you say the correct word somewhere else in the same recording, you can copy it and paste it over the wrong one. Cut in the gaps between words, use short crossfades, and adjust the volume to match.

Works when: the word appears elsewhere in a similar position in the sentence. Fails when: the word never appears, or appears with a different intonation, such as at the end of a question.

Method 2: Re-record the sentence and punch it in

Record the whole sentence again and replace the old one. For it to blend in, match the microphone, room, distance and speaking energy of the original. Listen to the sentences around it first and read them aloud to get back into the same rhythm.

Works when: you still have the same setup. Fails when: the recording was made in a different place or a long time ago, or you are not near a mic.

Method 3: Regenerate the sentence in your own voice

A text-based editor like Respeak lets you change the sentence as text and reads the new version in a voice made from your recording:

  1. Upload the recording. Respeak transcribes it into sentences (Chinese, English, Japanese, Korean, Spanish, French, German or Portuguese).
  2. Click the sentence and change the words in the text. You see exactly what changes.
  3. Generate. The sentence is spoken in the voice from that recording, and its tone and room sound are matched to the audio around it.
  4. Listen to the new and the original sentence, try another version if needed (3 free generations per sentence), then replace it.
  5. Export the whole recording as WAV or MP3.
Listen: “at home” in the second sentence was changed to “with your phone”. Sample recording made with AI

Try it on your recording. 10 free credits when you sign up →

Works when: it is your own voice and the recording is reasonably clean. Limits: loud background noise, music under the voice or several people talking over each other make it harder. Respeak only works with your own voice. You confirm that before generating.

Which method should you use?

Tips for a natural-sounding change

FAQ

Why does a single replaced word sound pasted in?

The pitch and rhythm of a word depend on the sentence around it. A word taken from another sentence, or recorded on its own, usually carries the wrong intonation. Replacing the whole sentence avoids this.

Can I change a word in someone else's recording?

Respeak only works with your own voice. Changing what another person said in their voice requires their consent, and Respeak does not offer it.

Which languages are supported?

Respeak transcribes and regenerates Chinese (Mandarin), English, Japanese, Korean, Spanish, French, German and Portuguese.

What does it cost?

Transcribing, previewing and exporting are free. You pay 1 credit for each sentence you replace, and you get 10 free credits when you sign up. See Pricing.

Try it on your own recording. Upload, change the words, listen before you replace. 10 free credits when you sign up.