Use Case

Clean Up a Podcast Recording

Take a raw interview with room echo, long pauses and a lot of 'um' and make it something people can listen to.

Difficulty:beginner

Time saved:2-4 hours per episode

Our pick

Best tool for this job

Descript

Edit video by editing text. AI voices, transcription, and studio sound.

Descript turns the recording into a transcript and you edit the audio by editing the text. Cutting a tangent means deleting a paragraph. Studio Sound handles a lot of echo and background hiss, and the filler word removal saves you from scrubbing through the waveform. For a one-hour interview that's the difference between an evening and a few hours.

Visit Descript →See full review

Pricing

Pricing changes often: check the official site for current plans.

Setup

easy

Coding required?

🥈 Runner-up

Also worth considering

ElevenLabs

Ultra-realistic AI voice generation and cloning.

ElevenLabs has a voice isolator that pulls speech out of noisy audio, which is useful if the recording is badly damaged (a cafe, a windy street). It doesn't edit anything though, so you'd still need an editor for the cuts.

Visit ElevenLabs See review

Getting Started

✓ Do this

→Import each speaker's track separately if you have them. The transcript is more accurate and you can fix levels per person.
→Run Studio Sound on a short clip first and listen with headphones. If it sounds watery, turn the strength down.
→Fix the transcript's misheard names and jargon before you edit. It's quicker than fixing them later.
→Remove filler words in a batch, then skim the result. Leave some in, since a podcast with none of them sounds odd.
→Shorten long silences instead of deleting them, so the pacing still feels like a conversation.
→Export a short test file and listen on your phone speaker and in a car before you publish.

✗ Avoid this

→Applying noise removal to everything at full strength. Voices end up sounding underwater.
→Trusting the auto-removed fillers without listening. It sometimes cuts the start of a real word.
→Editing the transcript on a recording where two people talk over each other. Those parts need a manual fix.
→Forgetting to normalize loudness before export, so your episode is quieter than everything else in the feed.

Looking for something else?

Browse all use cases New search

Getting Started

✓ Do this

→Import each speaker's track separately if you have them. The transcript is more accurate and you can fix levels per person.
→Run Studio Sound on a short clip first and listen with headphones. If it sounds watery, turn the strength down.
→Fix the transcript's misheard names and jargon before you edit. It's quicker than fixing them later.
→Remove filler words in a batch, then skim the result. Leave some in, since a podcast with none of them sounds odd.
→Shorten long silences instead of deleting them, so the pacing still feels like a conversation.
→Export a short test file and listen on your phone speaker and in a car before you publish.

✗ Avoid this

→Applying noise removal to everything at full strength. Voices end up sounding underwater.
→Trusting the auto-removed fillers without listening. It sometimes cuts the start of a real word.
→Editing the transcript on a recording where two people talk over each other. Those parts need a manual fix.
→Forgetting to normalize loudness before export, so your episode is quieter than everything else in the feed.