Guides › Editing

How to edit audio by editing text

Editing speech on a waveform means listening, scrubbing and zooming to find every word. Text-based editing turns the recording into a transcript first: delete a word in the text and it's cut from the audio, rewrite a sentence and it's spoken again. Here is how it works, what it's good at, where it struggles, and how to do it step by step.

Edit your recording like a document: try Respeak free →

How text-based audio editing works

Speech recognition turns the recording into words, and each word knows where it starts and ends in the audio. Once that link exists, editing the text can edit the sound. There are three kinds of edit:

• Delete: remove a word, a false start or an “um”, and that stretch of audio is cut out.
• Shorten: a long pause becomes a short one.
• Rewrite: change the words of a sentence, and the new sentence is spoken again, either recorded by you or generated in your own voice.

What it's good at, and where it struggles

• Good at: spoken word, such as podcasts, lessons, voice-overs and talks. You find a spot by reading instead of scrubbing, and cutting fillers, repeats and long pauses takes minutes instead of an hour.
• Struggles with: music and sound effects (they have no words to edit), several people talking at once, and cuts in the middle of a word. Edits timed to music are still a job for a waveform editor.
• Always listen: the text shows what was said, not how it sounds. A cut that reads well can still leave half a breath or a clipped syllable, so check every edit by ear.

Step by step with Respeak

1. Upload the recording (MP3, M4A, WAV, MP4, MOV and more). It's transcribed into sentences (Chinese, English, Japanese, Korean, Spanish, French, German or Portuguese). Double-click a sentence to fix a misheard word; that changes the text only, not the sound.
2. Cut words and pauses: open Remove fillers. Ums, repeats, restarts and pauses over a second are marked for you. Click any word to delete it and click again to bring it back, drag across several words to delete them together, and choose how much pause to keep where you cut.
3. Listen, then apply: play from any sentence as it will sound, or review the marked spots one by one, then click Apply cuts.
4. Fix a wrong sentence: click it, type what you meant to say, and generate. It's spoken in the voice from that recording (3 free generations per sentence, 1 credit to replace). Or use Record this line to say it again yourself, which is free.
5. Export the audio as WAV or MP3, or the transcript as TXT or SRT subtitles.

Try it on your recording: transcription and cutting are free →

Free, and limits: transcription is free up to 60 minutes a day, and cutting words and pauses and exporting are free; you need an account. Recordings can be up to 10 minutes on the free trial and 30 minutes after buying any credit pack. Generating a sentence works only with your own voice, and you confirm that before generating.

Tips for edits nobody notices

• Cut whole thoughts. When you restarted a sentence, delete the false start and the breath before it, not just the repeated words.
• Keep a little pause at every cut. Removing every gap makes speech sound breathless and rushed.
• Don't cut every filler. A few “so”s and short pauses keep it sounding like you. Cut the ones that interrupt the point.
• Fix the transcript before you export subtitles. Names and numbers are the words most often misheard.
• Subtitling the original video? Export the subtitles before you cut, because cuts change the timing.

FAQ

Is editing audio by text the same as AI voice cloning?

No. Deleting words and shortening pauses only cuts the original audio. A voice is needed only when you rewrite a sentence and choose to generate it, and you can record the new sentence yourself instead.

Which languages are supported?

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

What does it cost?

Transcription is free up to 60 minutes a day, and cutting and exporting are free. Replacing a sentence with a generated one costs 1 credit, and you get 10 free credits when you sign up.