Audio Cleaning Workflow for Creators: Mic to Export in 6 Steps

By AudioClean Pro team · February 6, 2026

Updated: May 2026

TL;DR: A repeatable 6-step audio cleaning workflow for podcasters, YouTubers, and video creators working on a Mac. Start with how you record, end with a platform-ready file. Total active time for a 30-minute episode is about 15 minutes once the workflow is dialled in.

  1. Capture clean source audio
  2. Trim and prep the file
  3. Run noise reduction
  4. Remove filler words and stutters
  5. Normalize loudness
  6. Export for your platform

Step 1 — Capture clean source audio

The single biggest determinant of how much work the rest of the workflow has to do is your source recording. Two minutes of attention here saves twenty later.

  • Gain staging: set input gain so the loudest peaks land between -12 and -6 dBFS. Anything hotter clips; anything quieter forces you to compensate later and amplifies the noise floor with it.
  • Mic distance: a cardioid dynamic mic 6 inches from the mouth (closer for soft talkers) rejects most room sound before it reaches the capsule. Doubling the distance from the mic roughly quadruples the share of room reflections in the recording.
  • Room treatment: a soft-furnished room (rugs, bookshelves, curtains) beats a treated booth that you do not own. Recording in a hard-walled office produces flutter echo that no cleaner fully removes.
  • Recording format: WAV or AIFF at 48 kHz, 24-bit if your interface supports it. MP3 source recordings work but limit your headroom for normalisation.

Step 2 — Trim and prep the file

Drop the raw file into AudioClean Pro. Trim the silence before your first sentence and after your last. Two reasons:

  • The leading silence often contains a button click, throat clear, or "OK starting now" that nobody needs to hear.
  • Trimming first means every subsequent step processes less audio, which means everything runs faster.

Confirm the file format is supported. AudioClean Pro reads WAV, AIFF, FLAC, MP3, M4A for audio and MP4, MOV, M4V for video. Video files have their audio track cleaned in place and the video is re-muxed unchanged.

Step 3 — Run noise reduction

Apply AI noise reduction. AudioClean Pro runs an on-device neural model trained on speech, so it removes:

  • Broadband background noise: HVAC, computer fan, distant traffic, room tone hiss.
  • Tonal noise: mains hum, ground loops, fluorescent buzz.
  • Sporadic noise: mouse clicks, keyboard taps, chair creaks.

The default attenuation works for 90% of recordings. If you have a particularly noisy source, the live preview lets you raise the attenuation while toggling the cleaned signal against the original; you confirm what you are hearing before you commit the change.

If the recording has obvious room reverb, also apply the de-reverb step. It tightens the apparent room without making the voice sound dry or unnatural.

Step 4 — Remove filler words and stutters

AudioClean Pro transcribes your file on-device and flags every "um", "uh", "like", "you know", "so", and stuttered opening with a timestamp. You see them in a transcript editor with the surrounding context and approve which to cut.

Two tactical notes:

  • Do not cut all of them. A speaker who says "um" zero times sounds robotic. Keep 5 to 15 percent of natural ones, especially mid-thought, and cut the rest.
  • Stutters at the start of sentences ("I, I, I think") consolidate cleanly; the same pattern mid-sentence sometimes hides a meaningful pause and is worth a manual listen.

The cuts are applied as audio splices with a 25 ms equal-power crossfade at each boundary, so the result sounds continuous rather than chopped.

Step 5 — Normalize loudness

Loudness normalization sets the perceived volume of the file to a platform-appropriate target. AudioClean Pro measures the integrated loudness across the whole file and applies a single gain correction to hit your target. Targets that matter:

  • Podcast platforms (Apple Podcasts, Spotify): -16 LUFS integrated, -1 dBTP true-peak ceiling.
  • YouTube: -14 LUFS integrated. YouTube re-normalises uploads to this level anyway, so matching it preserves your dynamics.
  • Broadcast / EBU R128: -23 LUFS integrated. Use this if a station or distributor asks for it.

The "Broadcast Ready" toggle in AudioClean Pro applies a final true-peak limiter so nothing exceeds the ceiling regardless of platform.

Step 6 — Export for your platform

Pick the format your downstream pipeline expects:

  • Podcast hosts (Buzzsprout, Transistor, Megaphone): MP3 at 320 kbps, joint stereo or mono if your source is mono.
  • Video editors (Final Cut, Premiere, DaVinci): WAV at 24-bit, 48 kHz. Match your video project sample rate.
  • Direct upload (Riverside, Descript, Substack): M4A at 256 kbps; smaller than WAV, transparent compared to MP3.
  • Archive: FLAC at 24-bit. Lossless, about 60% the size of WAV.

AudioClean Pro can also produce a transcript export alongside the audio (SRT for captions, VTT for web players, plain text for show notes). If you removed fillers in Step 4, the exported transcript reflects only the words that survived.

One-time setup that pays off forever

Save your settings as a preset the first time you complete this workflow. Every subsequent episode is a one-click run: drop the file, hit "Clean", review the filler suggestions, export. The whole sequence for a 30-minute episode runs in about 15 minutes on Apple Silicon.

If you produce in batches, drop a whole folder into the app at once; it applies the same settings to every clip in parallel. The automation guide walks through batch-style workflow in more depth.

AudioClean Pro home · AI features explained · Free trial download · For podcasters · Compare alternatives