How to Upload a Video to YouTube (Audio Production and Loudness Guide)

Creators often fix color, pacing, and thumbnails, then hit publish with a track that sounds small, hissy, or fatiguing. This guide shows a complete audio workflow for YouTube: from recording chain and cleanup to export, loudness, true-peak control, captions, QC, and final upload.

What you'll get out of this

• A reliable recording chain that avoids avoidable problems
• A repeatable cleanup and mix template for speech clarity
• A controlled loudness and true-peak outcome that survives transcoding
• A simple export and upload checklist that prevents "why is it quieter" surprises

1) Recording chain that gives you headroom later

Microphone and placement

• Large-diaphragm condenser or broadcast dynamic, 15–25 cm from mouth, slightly off-axis to reduce plosives.
• Pop filter or foam. Aim the mic so breath passes the capsule, not into it.
• If you must use a lav, place it off the sternum to avoid the chest boom; add a small fur for clothing noise.

Preamp and gain staging

• Target live peaks around −12 dBFS on your meter with normal speech, not shouting. This leaves room for laughs and emphasis.
• Turn off auto gain on cameras and recorders. Lock a manual level.

Room control

• Kill fan noise. Close a closet door. Throw a duvet on the echo wall opposite the mic.
• Record 10 seconds of room tone before the take. You will use this later for noise prints and fill.

2) Cleanup and repair before you mix

Work on a duplicate track. Keep the original hidden and untouched.

Hum and buzz

• Narrow notch at 50 or 60 Hz and first 1–2 harmonics. Start with Q ≈ 10.
• If a ground loop causes wide hum, notch fundamental and apply light broadband denoise (3–6 dB reduction).

Broadband hiss and HVAC

• Set denoise threshold only a few dB below the noise floor. Reduction 3–6 dB is often enough.
• Prefer multiband or spectral denoise so the sibilant band stays lively.

Clicks, lip smacks, and mouth noises

• Run a click detector with sensitivity low enough to skip consonants.
• For stubborn clicks, manual spectral attenuate 4–9 kHz with a soft brush.

De-reverb (if you recorded in a bright room)

• Early reflection reduction 15–35%. Tail reduction modest first.
• Stop when speech stops "breathing." If artifacts appear, back off and prioritize EQ and mic distance next time.

Breaths and silences

• Reduce breaths by 6–10 dB rather than deleting them.
• Use strip silence with a 100–150 ms padding so words aren't clipped.

3) Make dialogue carry without harshness

High-pass and basic EQ

• High-pass at 70–100 Hz for most voices. Back off if it thins the tone.
• Pull 250–400 Hz gently if boxy. Small wide cuts beat big narrow ones for room build-up.
• If dull after denoise, lift 2–3 kHz by 1–2 dB and 8–10 kHz by 1 dB for air. Avoid adding hiss back in.

Dynamics and consistency

• Gentle wideband compression, ratio 2:1–3:1, attack 10–30 ms, release 60–120 ms.
• Aim 3–6 dB gain reduction on excited peaks. Use make-up gain to restore average level.
• A leveler or dialogue rider can smooth sentence-to-sentence swings without pumping.

De-essing that doesn't lisp

• Start with a split-band de-esser around 5–8 kHz. Threshold so it only hits sharp S/T bursts.
• Keep gain reduction under ~5 dB on normal lines. If more is needed, fix mic placement first.

Music and SFX under voice

• Duck music under speech by 6–12 dB.
• Carve a small dip in the music at 2–4 kHz so consonants cut through.
• Check the sum in mono to catch phasey pads masking your voice.

4) Loudness, true peak, and headroom that survives YouTube

Platforms transcode. Protect the transcode by controlling integrated loudness and true peak.

Targets that translate

• Set a consistent dialogue anchor, then normalize the full mix.
• A practical target for mixed speech content is around the mid-teens LUFS integrated with true-peak ceiling a little below full scale.
• Keep short-term loudness swings natural. Excessive look-ahead limiting sounds hot on laptop speakers and brittle on phones.

Limiter setup

• Transparent brickwall limiter last in chain.
• Ceiling a touch below 0 dBFS to preempt inter-sample overs after transcode.
• If your limiter shows more than ~4 dB gain reduction often, back off and fix balance upstream.

Normalize with math, not guesswork

• Measure integrated loudness on the full program.
• Compute gain = target − measured.
• Apply gain only (no extra compression) and re-check true peak. If true peak is now too close to full scale, lower ceiling slightly and re-render.

5) Reference checks before you export

Phone speaker reality

• Sum to mono and listen on your phone. Can you hear every word while walking outside?
• If not, turn the music down and revisit the 2–4 kHz pocket.

Earbuds and laptop

• Check low-level listening on cheap earbuds and a laptop.
• Fix any hiss bursts from aggressive high-shelf boosts.

Content pauses

• Trim dead lead-ins. Viewers bounce if the first 3 seconds are silent or low.

6) Export settings that work

Container and codecs

• MP4 container with H.264 video.
• AAC audio at 48 kHz. 256–320 kbps for stereo. Use mono if voice only.

True-peak safety and dither

• If you worked at 24-bit, export 24-bit if possible. If exporting 16-bit WAV stems, apply shaped dither once.
• Keep your limiter ceiling a touch below full scale.

A reference ffmpeg template

ffmpeg -i input.mov \
-c:v libx264 -preset medium -crf 18 -pix_fmt yuv420p -r 30 \
-c:a aac -b:a 320k -ar 48000 -ac 2 \
-af "loudnorm=I=-14:TP=-1.0:LRA=11,aresample=async=1:first_pts=0" \
output_youtube.mp4

Adjust the loudnorm values to your chosen anchor and the limiter ceiling to your comfort.

7) Upload flow that protects your work

1Upload and fill Title, Description, and a thumbnail that states the promise.
2Add chapters with timestamps. Keep labels short.
3Upload your SRT or fix the auto captions. Clean captions improve watch time.
4Add end screens and cards.
5Run a final phone speaker pass on the processed video before you push Public.

8) Troubleshooting by symptom

Voice sounds thin on phone

• Move presence from 8–10 kHz down to 2–4 kHz. Reduce music in the same band.

Voice is loud but fatiguing

• Short attack compression and heavy high-shelf cause grit. Lengthen attack, reduce high shelf, verify de-esser isn't clamping whole words.

"Swirly" metallic artifacts

• Too much denoise or de-reverb. Back off 2–3 dB, re-EQ, and consider re-recording key lines.

Pops on P and B

• Add a gentle low-band compressor keyed to 80–160 Hz or re-record the worst line. Use a pop filter next time.

Checklists you can tape to your monitor

Pre-mix

• Record 10 s room tone
• Peak around −12 dBFS
• Pop filter, off-axis mic

Cleanup

• Hum notch 50/60 Hz + harmonics
• Denoise 3–6 dB
• Click repair on spikes
• Light de-reverb only if needed

Mix

• HPF 70–100 Hz
• Comp 2–3:1, 3–6 dB GR
• De-ess on peaks
• Music duck 6–12 dB

Final

• Normalize to target
• Limiter ceiling just below full scale
• AAC 48 kHz, 256–320 kbps
• Captions, chapters, phone check

Need Professional Audio Help?

Get a Quote Try Our LUFS Meter

FAQ

Why does YouTube sometimes sound quieter after upload?

Transcoding and playback normalization affect perceived loudness. If dialogue is clearly above music and your true peaks are controlled, the upload will translate. The QC steps here minimize surprises.

Should I export mono or stereo?

If it is voice only, mono is fine and saves bitrate for clarity. If there is music, keep stereo but ensure dialogue remains centered and strong in mono.

What is a safe limiter attack and release?

Start with attack around 1–5 ms and release 50–100 ms. Avoid constant heavy limiting; if your meter shows big gain reduction all the time, fix balance upstream.