How to Clean Up Audio for TikTok and Instagram Reels (Practical Workflow)

Why Reels and TikTok audio needs a different cleanup pass

On YouTube, viewers might forgive a little room noise. On TikTok and Instagram Reels, they decide in under two seconds. Your voice has to cut through phone speakers, background music, and whatever noise is in the viewer's environment. The platform also re-encodes and normalizes your audio, so a recording that sounds fine in headphones can turn harsh, muffled, or quiet after upload.

This workflow is for real-world recordings: phone voice memos, Zoom calls, gaming clips, podcast tracks, and camera audio. You do not need a treated studio. You need to fix the right problems in the right order.

Step 1: Record the cleanest source you can

Cleanup works best when it has less to clean. Before you open an editor:

• Get the mic closer. Phone or lav 6-12 inches from your mouth beats a laptop mic across the room.
• Kill constant noise at the source. Turn off AC, fans, humidifiers, and nearby traffic. A $20 blanket behind you can reduce echo.
• Record separate audio if possible. For Zoom, enable Record a separate audio file for each participant. For gaming, capture mic on a separate track from game sound.
• Leave 5 seconds of room tone. Noise removal tools learn the noise profile from silence. Five seconds of just the room is gold.
• Watch levels. Aim for peaks around -12 dB to -6 dB, never touching 0 dB. Clipping cannot be fixed later.

Step 2: Diagnose the noise before you treat it

Different problems need different tools. Solo each track and listen for:

• Broadband hiss: steady shhh from cheap preamps, phone mics, or fans. Use noise reduction or AI denoise.
• Hum: low buzz at 50 Hz (UK/EU) or 60 Hz (US), plus harmonics at 120/180 Hz or 100/150 Hz. Use a de-hum or notch filter.
• Echo/reverb: hollow room sound. Noise reduction will not fix this well. Use a de-reverb lightly or re-record.
• Clicks, lip smacks, plosives: manual edits or a de-click/de-plosive tool.
• Sibilance: harsh s and t sounds. Use a de-esser, not a broad EQ cut.

If you cannot name the noise, you cannot choose the right fix. A hum plugin on hiss does nothing; noise reduction on echo makes your voice sound underwater.

Step 3: Build a cleanup chain in this order

Use this order in CapCut, Premiere, Audacity, Descript, or any DAW. Settings below are starting points, not laws.

1. High-pass filter. Cut below 80-100 Hz for deep male voices, 100-120 Hz for female or higher voices. This removes rumble, desk thumps, and AC noise without thinning the voice.
2. De-hum. Set 50 Hz or 60 Hz depending on your region, then harmonics. Reduce by 12-20 dB if the buzz is obvious. Bypass often to make sure you are not hollowing out the low end.
3. Noise reduction/denoise. Start with 6-10 dB reduction, not 30 dB. If the tool has a noise floor or sensitivity control, keep it moderate. The goal is to lower the noise floor, not erase all room sound. If your voice sounds like a robot or has watery tails, reduce the amount.
4. Subtractive EQ. Cut muddy boxiness at 200-400 Hz by 2-3 dB with a wide bell. Cut harshness at 2.5-4 kHz only if it hurts. Do not boost yet.
5. Compression. Ratio 3:1, threshold -18 to -12 dB, attack 5-10 ms, release 80-150 ms. Aim for 3-6 dB of gain reduction on loud parts. This evens out Zoom calls and phone recordings.
6. De-esser. If s sounds pierce, set 5-8 kHz and reduce 3-6 dB. Use a split-band de-esser if available.
7. Additive EQ. Boost presence at 2-5 kHz by 1-3 dB for phone-speaker clarity. Add a gentle high shelf at 8-10 kHz by 1-2 dB if it sounds dull.
8. Loudness and limiting. Target -14 LUFS integrated for TikTok/Reels, with true peak at -1 dBTP. Some creators push -12 LUFS for louder playback, but overly crushed audio gets turned down and sounds worse.

Step 4: Use a free online cleanup tool for stubborn noise

If your recording has constant hiss, fan noise, or city rumble, an AI noise remover can save you 20 minutes of manual work. Upload a 10-30 second test to a free online option like ClearVoice, download the cleaned file, then bring it back into your editor for EQ, compression, and loudness. It is not a full mix replacement, but it is a fast first pass.

For a no-install route, try a free tool like ClearVoiceHQ (clearvoicehq.com). Upload your video or audio, run noise removal, then compare the processed file with the original in headphones. If the voice sounds processed or the room disappears unnaturally, dial back the reduction or use the original for quiet sections. Always keep the original file.

Step 5: Mix voice and music for vertical video

Music is half the experience, but it should never fight your voice. In CapCut, Reels, or your editor:

• Duck the music. Sidechain the music to your voice or manually lower it by 6-12 dB whenever you speak. Music under voice should sit around -18 to -24 dB peak, while voice peaks around -6 to -3 dB.
• Use the right music. Avoid tracks with busy vocals or sharp cymbals. Lo-fi, ambient, and simple beats are easier to mix under speech.
• Cut music for punchlines. A half-second of silence before a key line makes the next sentence hit harder. This is a free editing trick.
• Check on phone speakers. Laptop speakers lie. Phone speakers have almost no bass, so if your voice sounds thin there, you cut too much low end. If it sounds muddy, cut 200-400 Hz a little more.

Step 6: Export settings that survive the upload

Your platform will re-encode your file. Give it a clean master so it has less to ruin.

• Format: MP4 for video, H.264 or HEVC. For audio-only, WAV 48 kHz/24-bit into your editor, then export AAC 320 kbps.
• Resolution: 1080x1920 for Reels/TikTok, 30 or 60 fps depending on source.
• Bitrate: 10-20 Mbps for 1080p vertical video. Higher is fine if file size is not a problem.
• Audio: 48 kHz sample rate, 320 kbps AAC, stereo or mono as recorded. Mono is fine for voice and can sound more focused on phone speakers.
• Loudness: -14 LUFS integrated, -1 dBTP true peak. Do not normalize to 0 dB.

Final pre-post checklist

1. Listen to the first 3 seconds on your phone speaker. Is the voice clear and loud enough?
2. Check the noise floor: pause between words should be quiet, not hissy or humming.
3. Listen for artifacts: watery tails, robotic voice, or cut-off room tone mean you over-processed.
4. Confirm music ducks under speech and never masks consonants.
5. Check true peak. If it hits 0 dB, add a limiter at -1 dBTP.
6. Upload a private test to TikTok or Reels. Platforms change the sound, so verify before publishing.

Clean audio is not about buying expensive gear. It is about removing noise, taming dynamics, and hitting a loudness target that sounds good on a phone. Do the steps in order, test on the worst speaker you own, and your Reels and TikToks will feel more professional before the first sentence ends.