Audacity Editing for ElevenLabs: The 6-Step Workflow I Use (Free, Exact Settings)

New to ElevenLabs? Get 50% off your first month* – $11 instead of $22. Same creator tools, half the price.

Raw audio from even the best mic has background noise, inconsistent volume, long pauses, and mouth sounds. The AI trains on whatever you give it. Feed it unedited audio, the clone sounds unedited.

This is the 6-step Audacity workflow I used. My first voice (no proper editing) peaked at €50/month and died. My second voice (this workflow) earned €263 in 3 months. The editing is a big part of why.

Before you start

Set up Audacity first: Edit > Preferences > Quality. Set Default Sample Rate: 48000 Hz and Default Sample Format: 32-bit float. Make sure your recording device is set to your audio interface (not your laptop mic). This only needs to be done once.

The 6-Step Workflow

Noise Reduction

Removes the constant background hum your mic picks up.

Even quiet rooms have a noise floor – hum from your computer, electricity, traffic. You can’t hear it when talking, but the mic picks it up. First, find a 1-2 second silent gap in your recording (where you weren’t speaking) and select it.

Select the silent section > Effect > Noise Reduction and Repair > Noise Reduction
Click “Get Noise Profile” – Audacity learns what your background noise sounds like
Select your entire track (Ctrl+A)
Go back to Effect > Noise Reduction and Repair > Noise Reduction and apply:

          Noise reduction: 12 dB

          Sensitivity: 6.00

          Frequency smoothing: 3 bands

Don’t go above 12 dB. Higher values make your voice sound robotic. If 12 dB isn’t enough, the problem is your room, not this setting.

Truncate Silence

Removes long dead-air gaps between sentences.

You’ve got pauses where you were thinking or catching your breath. ElevenLabs doesn’t need 3-second gaps. This compresses them down.

          Select entire track (Ctrl+A)

          Effect > Truncate Silence

          Minimum silence duration: 1.0 seconds

          Truncate to: 0.5 seconds

Your 2-hour recording might become 1 hour 45 minutes. That’s good – no wasted space.

Amplify (Measure Noise Floor)

A measurement step – you don’t actually apply anything here.

Find another silent section (after noise reduction). Select it, then:

          Effect > Amplify

          Note the suggested amplification value

          DON’T click OK – just note the number, then Cancel

This tells you how quiet your noise floor is. Higher number = cleaner recording. It’s just a check.

Normalize

Centers the waveform and brings audio to a consistent level.

          Select entire track (Ctrl+A)

          Effect > Normalize

          Check: Remove DC offset

          Peak amplitude: -1 dB (or -3 dB for more headroom)

Your audio now sits in a sensible range. Not clipping, not too quiet.

Compressor

The step most people skip – and the one that makes the biggest difference.

Compression brings loud parts down and quiet parts up, so your voice has consistent volume throughout. This is what makes audio sound like a podcast instead of a bedroom recording.

          Select entire track (Ctrl+A)

          Effect > Compressor

          Threshold: -18 dB

          Ratio: 3:1 (or 4:1 for more compression)

          Attack Time: 0.2 seconds

          Release Time: 1.0 second

          Check: Make-up gain for 0 dB after compressing

Listen to the result. Your voice should sound fuller, more polished. Don’t skip this step.

Final Amplify

Push levels up to finished, professional volume.

          Select entire track (Ctrl+A)

          Effect > Amplify

          Make sure “Allow clipping” is OFF

          Audacity suggests a safe value – accept it

          Click OK

Your audio is now at finished, professional level. Loud enough, not distorted.

Export and Check Loudness

Export as WAV, not MP3. File > Export > Export as WAV. MP3 is lossy – you lose quality every time you edit and re-export. WAV is lossless. ElevenLabs handles the file size fine.

Check loudness before uploading. Use Youlean Online Loudness Meter (free, browser-based). Drag your WAV file in and check these targets:

Metric	Target	Why
Integrated Loudness	-18 to -23 LUFS	Enough headroom for AI processing
True Peak	-3.0 dB or lower	Prevents digital clipping
Loudness Range (LRA)	5-7 LUFS or less	Consistent volume = stable clone

I ran every session through Youlean before uploading. It takes a few minutes for short files, longer for full sessions – my two 1-hour files took about 20 minutes total. Worth it.

Common Mistakes

Skipping the compressor

This is the step that separates amateur from professional. Without it, some words are loud and some are quiet. The AI clone reproduces that inconsistency. Don’t skip it.

Cranking noise reduction above 12 dB

More noise reduction doesn’t mean cleaner audio – it means more robotic audio. If 12 dB isn’t enough, the problem is your room, not the setting. Fix the room first.

Not checking loudness before uploading

Audio that’s too loud clips during AI processing. Audio that’s too quiet produces a weak clone. Two minutes in Youlean catches both problems before you waste an upload.

Want the full process?

The complete guide covers equipment setup, Windows configuration, this editing workflow with screenshots, ElevenLabs upload, publishing, and monetization.

See the guides →

*This article contains affiliate links. If you sign up through one of these links, I may receive a commission at no extra cost to you. The price you pay stays the same. I only recommend tools I actually use.

Andy from KindredView

I test creator monetization strategies and write about what actually works. No hype – just the numbers.

The 6-Step Audacity Editing Workflow I Use for ElevenLabs Voice Clones