- Documentation
- »User Guide
- »Enhancing Audio
Enhancing Audio
Audio Audit doesn't just tell you what's wrong with your audio — it can fix the most common problems too. Enhancement takes an episode, runs your chosen clean-up steps over it in a single pass, and gives you back a new file ready to publish: quieter background noise, more even speech levels, loudness matched to platform standards, and clean tags and cover art.
Enhancement is deliberately a simple tool with opinionated defaults, not a mixing suite. Every setting has an Auto option chosen from established podcast production practice, and Auto is always a safe choice. Where you do want control — a different loudness target, a stronger clean-up, a specific export format — the options cover most real-world cases without asking you to learn audio engineering first.
What Enhancement Can Do
| Enhancement | What it does |
|---|---|
| Noise reduction | Removes background hiss, hum and room noise using a noise profile measured from your own recording, without dulling speech. |
| Compression | Evens out the gap between loud and quiet speech so levels stay comfortable throughout. |
| Volume normalisation | Brings the whole episode to a consistent target loudness (e.g. -16 LUFS for podcasts) with safe true-peak limiting. |
| Metadata | Writes clean title, artist, album and other tags plus cover art onto the file. |
Pick any combination — everything you tick runs together as one job on one file.
If You Just Want a Good Sounding Show
You don't need to understand any of the settings to get a good result:
- Analyse your episode first — create a report as usual.
- On the finished report, find the "Enhance this episode" panel. Enhancements that would fix problems the report found are pre-ticked and marked "suggested" — a failed noise check suggests noise reduction, a loudness problem suggests normalisation, and so on.
- Click Enhance, leave every option on its default, and click Start enhancement.
- Wait for processing to finish, listen to the result in your browser, and download the enhanced file.
No report? You can also drop a file straight onto the "Enhance Audio" card on your Dashboard and enhance it directly. Either way, we recommend starting from a lossless WAV or FLAC file if you have one — it gives the clean-up the best possible material to work with. An MP3 still works; lossless works better.
💡 Tip: The defaults are conservative on purpose. Auto noise reduction measures your room before deciding how much to remove, Auto compression uses a gentle speech-friendly setting, and the default loudness preset matches what podcast platforms expect. If you never touch an option, you'll still get a properly levelled, cleaner-sounding episode.
Starting an Enhancement
There are two ways to start an enhancement:
From a Report
The "Enhance this episode" panel appears on every completed report. This is the recommended route — the report has already measured your audio, so Audio Audit can suggest which enhancements will actually help, and your episode's existing tags are pre-filled into the metadata fields for you to tidy up rather than retype.
From the Dashboard
The "Enhance Audio" card on your Dashboard lets you upload a file directly, without running a report first. Supported formats:
- MP3 — most common podcast format
- M4A — Apple's preferred format
- WAV — uncompressed audio
- FLAC — lossless compressed audio
Getting the Best Results
- Upload lossless if you can. WAV or FLAC gives the processing the most information to work with. An MP3 works fine too — but every generation of lossy encoding costs a little quality, so start from the best copy you have. (If you're only writing metadata, format doesn't matter.)
- The deepest clean-up happens before the mix. Noise reduction works best on a voice track before music and stings are mixed in — music can mask the difference between "noise" and "content". A finished episode still benefits; a raw voice track benefits more.
- Level voice and music separately — and keep compression off the music. Compression here is tuned for speech; on music it can audibly "pump" and flatten dynamics your intro was mixed to have. If music is a big part of your show, enhance the voice track alone (noise reduction and compression), balance the music against the treated voice in your editor, then run one last normalisation-only enhancement on the finished episode to hit the loudness target. Normalisation applies a single constant gain change, so it's safe on music.
- Enhance, then re-analyse. Running a fresh report on the enhanced file is the easiest way to confirm the problems are gone — and to see your score improve.
The Enhancements in Detail
This section explains what each enhancement actually does and walks through its options. You don't need any of it to use the feature — it's here so you can understand, and trust, what's happening to your audio.
Noise Reduction
Background noise — air conditioning rumble, computer fan hiss, electrical hum, the general "sound of the room" — sits underneath your voice through the whole episode. Noise reduction estimates what that constant background sounds like and subtracts it, leaving speech intact.
How it works. Audio Audit first analyses your file to build a noise profile: it measures the level of the quietest moments (the noise floor) and how it differs from your speech. It then applies two complementary techniques used throughout the audio industry — a gentle broadband smoothing pass, and spectral noise reduction, which works frequency-by-frequency to subtract the measured noise profile from the signal. Because the profile comes from your recording rather than a generic assumption, it removes your room's particular noise rather than guessing.
When your recording has audible room tone between phrases, a soft speech gate is also applied: it gently lowers the level in the gaps between sentences, rather than hard-muting them, so pauses sound natural instead of unnervingly dead.
Why there's a ceiling. All noise reduction is a trade-off: remove too much and speech starts to sound underwater, with a swirly artefact engineers call "musical noise". Audio Audit caps reduction at 12 dB — a deliberate limit below the point where those artefacts appear. If your recording is noisier than that, the better fix is at the source (a quieter room, a closer microphone).
Options:
| Strength | Reduction | When to use it |
|---|---|---|
| Auto (default) | Up to 12 dB, guided by your measured room tone | Almost always — it measures the room and picks the safe amount |
| Light | Up to 6 dB | Already-clean recordings that need just a touch less hiss |
| Medium | Up to 9 dB | Typical home-studio background noise |
| Strong | Up to 12 dB, with a deeper gate between phrases | Noisy rooms — the most reduction that stays artefact-free |
💡 Tip: If your recorder or software already mutes the gaps between sentences (a "voice-activated" or gated recording), Audio Audit detects this and automatically skips the steps that need natural room tone to work — they'd misfire on artificial silence.
Compression
If you've never met a compressor: it's an automatic volume control. When the sound gets louder than a set level (the threshold), the compressor turns it down; how firmly it turns it down is the ratio — at 3:1, sound that goes 3 dB over the threshold comes out only 1 dB over. The effect is like a careful engineer riding the volume fader through your whole episode: loud bursts are tamed, so the quiet moments no longer force your listener to reach for their own volume control. This matters most where podcasts are actually heard — cars, kitchens, earbuds on a train — where background noise swallows quiet speech.
How it works. Audio Audit measures your episode's average speech level first and sets the threshold relative to your recording, not a fixed number: the threshold sits at your measured average level, so it's the louder-than-average moments that get levelled. (On very hot material it's capped so it never sits above -6 dBFS.) The attack and release times — how quickly the compressor reacts and lets go — are fixed values tuned for speech: a 20 ms attack catches a laugh or an outburst while still letting the crisp onset of consonants through, and a 250 ms release recovers smoothly between phrases rather than audibly "pumping" on every syllable. On stereo files both channels are controlled together by whichever is louder, so the stereo image never wanders. The Amount setting changes only the ratio — threshold, attack and release stay measurement-driven and speech-tuned at every setting.
One thing our compressor deliberately does not do is make-up gain — the volume boost most compressors apply afterwards to compensate for the level they took away. Levelling the result back up is a job for normalisation, which does it accurately against a measured target. This is why ticking Compression automatically adds Volume normalisation (you'll see it locked with a "required for compression" badge): compression alone would simply make your file quieter.
Options:
| Amount | Ratio | When to use it |
|---|---|---|
| Auto (default) | 3:1 | A moderate, broadcast-style setting that suits most spoken word |
| Light | 2:1 | Gentle smoothing that preserves most of the natural dynamics |
| Medium | 3:1 | Same as Auto |
| Strong | 4:1 | Very uneven recordings — multiple speakers at different distances, big laughs |
| Off | — | Skip compression entirely while keeping other enhancements |
Volume Normalisation
Different platforms, and different listeners, expect audio at a consistent loudness. LUFS (Loudness Units relative to Full Scale) is the industry's standard way of measuring perceived loudness — how loud a whole programme feels to a human ear, not just how big its waveform is. Podcast platforms and apps have converged on around -16 LUFS for stereo shows; audio that's far off target gets turned up or down by the platform, or worse, arrives noticeably quieter than every other show in the listener's feed.
Alongside the loudness target there's a true peak ceiling. True peak measures the highest instantaneous level the audio will actually reach on playback — including "inter-sample" peaks that can exceed what the stored samples show. Keeping true peaks at or below -1 dBTP leaves headroom so the file won't distort when platforms transcode it to lossy formats like AAC or MP3.
How it works. Audio Audit uses two-pass loudness normalisation built on the same international loudness measurement standard used in broadcast (the ITU-R BS.1770 family, the basis of EBU R128). The first pass measures your episode's integrated loudness, true peak and loudness range; the second applies a single, constant gain adjustment to hit the target. Because the gain is constant — linear, in engineering terms — your dynamics are untouched: normalisation changes how loud the episode is, never how it breathes. A safety limiter at the true-peak ceiling then catches only the occasional stray peak; it is a backstop, not a levelling tool, and does nothing at all when nothing exceeds the ceiling.
Options:
| Preset | Target | True peak | When to use it |
|---|---|---|---|
| Podcast (default) | -16 LUFS | -1 dBTP | The podcast industry standard — right for almost every show |
| YouTube | -14 LUFS | -1 dBTP | Matches the louder normalisation used by YouTube and music streaming platforms |
| ACX audiobook | -19 LUFS | -3 dB | Aimed at Audible/ACX audiobook submission requirements |
| Custom | Your choice | Your choice | Set your own target LUFS and true-peak ceiling |
💡 Tip: For mono files, the Podcast preset automatically targets -19 LUFS instead of -16. This is standard practice: a mono file plays from both speakers at once, so -19 LUFS mono sounds as loud to the listener as -16 LUFS stereo.
Metadata & Cover Art
Clean tags are how podcast apps, car stereos and file managers identify your episode. The metadata enhancement writes standard tags — Title, Artist, Album, Track, Year, Genre, Copyright and Comment — in the right native format for your file type, plus embedded cover art.
- When you start from a report, the fields are pre-filled with the episode's existing tags, so you're correcting rather than retyping.
- Fill in a field to write it; clear a pre-filled field to remove that tag from the file.
- Cover art is resized and cropped to a square 800×800 JPEG by default. That matches what podcast directories recommend for embedded art and keeps the image comfortably under the size at which some apps and players start refusing it. Untick "Resize & crop" to embed your image exactly as uploaded.
💡 Tip: If metadata is the only enhancement you select (and you leave the output format on "Same as uploaded"), your audio is not re-encoded at all — the sound data is copied bit-for-bit and only the tags change. There is zero quality loss.
Output Format
By default the enhanced file comes back in the same format you uploaded — same codec family, same sample rate, same mono/stereo layout. If you'd rather get a ready-to-publish file in one step, choose an export format:
| Option | Best for |
|---|---|
| Same as uploaded (default) | Keeping your existing workflow and format |
| MP3 · 320 kbps | Highest MP3 quality |
| MP3 · 256 kbps | High quality |
| MP3 · 192 kbps | Great for publishing — the sweet spot for spoken word |
| MP3 · 128 kbps | Smallest file |
| FLAC | A lossless master copy for your archive |
However many enhancements you select, your audio is decoded once, processed once, and encoded once — the enhancements run as a single chain at full resolution, so there's no generational quality loss between steps.
Following Progress and Downloading
Starting an enhancement takes you to its progress page, which updates automatically. Recent jobs also appear under "Latest Enhancements" on the Dashboard and on the Enhancements page.
When the job completes you'll see a Results summary of what was actually done and measured — for example, the loudness target, the measured loudness before and after, the final true peak, and the noise floor that was detected — along with the credits charged. You can play the enhanced file directly in your browser to check it, then download it with "Download enhanced file".
💡 Tip: Don't take our word for it — run a fresh report on the enhanced file. The before/after numbers on the results page come from the same measurements the reports use, so the two will agree.
For Engineers: How the Pipeline Works
If audio is your job, here's the part you'll want to verify before trusting a one-button tool with your programme material.
Processing order is fixed and conventional. Stages always run noise reduction → compression → normalisation, with metadata written last onto the finished encode. This is the standard chain order for spoken word: denoise first so the compressor doesn't bring noise up with the quiet passages, compress before loudness targeting so normalisation measures (and hits) the final dynamics, and never let anything after the limiter touch gain.
Everything is measurement-first. No stage applies textbook numbers blind. The noise profile comes from statistical analysis of your file's quietest windows; the compressor threshold is derived from your measured programme level (never above -6 dBFS), with a 20 ms attack, 250 ms release, max-linked stereo detection and no make-up gain; loudness normalisation is a full measurement pass followed by a linear (constant-gain) correction, per the ITU-R BS.1770 loudness model. Dynamics are never pumped by the normalisation stage — loudness range is preserved, and the true-peak limiter engages only on overshoots.
Your format survives. Sample rate is pinned to the source, channel layout is preserved, and there is exactly one decode/encode cycle regardless of how many stages run. Metadata-only jobs with no format change don't re-encode at all.
Edge cases are detected, not ignored. Voice-activated or hardware-gated recordings (artificial silence between phrases) are detected and the room-tone-dependent steps are skipped rather than misapplied. Noise reduction is hard-capped at 12 dB because beyond that, spectral processing on speech produces audible musical-noise artefacts — the tool declines to make your audio worse.
And what it isn't. There's no EQ, de-essing, multiband processing or mixing here, and no plans to bluff at them — this is a finishing pass for spoken-word programmes, built on a handful of well-understood processes with conservative, opinionated settings. If your show needs surgical work, do it in your DAW; Audio Audit will happily verify the result.
How Billing Works
Enhancements use the same credit system as reports, with the same simple rules:
- Priced per audio-minute. Longer files cost more; the per-minute rate depends on which enhancements you tick. Noise reduction and normalisation are the most processing-intensive; metadata is the lightest. As a rough guide, a full clean-up (all four) costs about the same per minute as a full standard analysis.
- One job, one charge. Any combination of enhancements on a file runs — and is billed — as a single job. You're never billed separately per enhancement.
- A 10-minute minimum applies per job, just as with reports, covering the fixed cost of spinning up the pipeline and moving your file in and out of storage.
- Options never change the price. Strength settings, loudness presets, custom targets, output formats — all free to change. Only the audio length and which enhancements you tick affect cost.
- You see the price before you commit. The ~credits estimate shown next to the Enhance button is calculated with the same rules as the actual charge. The exact figure appears on the results page afterwards, and in your credit history under Settings → Billing and plans.
- Failed jobs are never charged. If processing fails for any reason, no credits are taken.
- Out of credits? The job is saved with a "Not enough credits" notice and nothing is processed or charged. Top up or upgrade from the billing page, then start the enhancement again.
The "How many credits do I need?" calculator on the Pricing page lets you tick the exact enhancements you plan to use and see the cost for your episode length before you commit.
💡 Note for Organisations: running enhancements — especially on directly uploaded files — is controlled by roles and permissions. Owners and Admins can always enhance; other members need the "Enhance audio" permission.
Next Steps
- Creating Reports — Analyse an episode first so Audio Audit can suggest fixes
- Understanding Reports — Interpret the checks that drive enhancement suggestions
- Billing & Plans — Credits, top-ups and the cost calculator