Auphonic or Podcastle, which AI tool makes a podcast voice sound better?

Auphonic is a processing engine with a leveler, four denoisers and LUFS targets; Podcastle, now Async, is a browser studio with Magic Dust. Free tiers, prices as of October 2026, and why neither works on a live call.


A 48-minute episode lands in your Downloads folder on a Wednesday night. Your half was recorded on a USB mic at the desk. Your guest joined from her kitchen, and under every pause sits the hum of a refrigerator compressor. When she laughs, the meter jumps into the red; when she leans back to think, she drops so low that anyone listening in a car will reach for the volume knob. Two browser tabs are open: Auphonic and Podcastle. Both promise AI that fixes this. Both will happily take your file. They are not doing the same job.

Here is the short answer. Auphonic is a processing engine: you hand it a finished file, choose a loudness target and how much leveling and noise reduction you want, and a processed file comes back. Podcastle, which renamed itself Async in January 2026, is a browser studio: remote recording, a multitrack editor, transcripts, AI voices, and a one-click cleanup called Magic Dust inside that editor. If the question is "which one gets my finished episode to even, correct loudness with the most control", pick Auphonic. If you want to record, edit and clean in one tab and can live with fewer dials, pick Async. Neither works while you talk. Both process audio after the recording exists.

If what bothers you is how you sound live, on a Zoom or Discord call nobody will edit afterwards, read our piece on real-time audio enhancers instead. That article covers what can run while you speak. This one stays with files.

Podcastle is now called Async

Search for Podcastle in October 2026 and you land somewhere else. The company announced on January 19, 2026 that Podcastle is now Async, and the old pricing address, podcastle.ai/pricing, answers with a permanent redirect to async.com/pricing. The editor most guides still call Podcastle is the studio part of Async, and the podcast hosting pages still carry the Podcastle name. Same tools, same Magic Dust, new logo.

That has one practical consequence. Comparison posts from 2024 and 2025 quote plans and prices from a page that no longer exists under that name. Treat any Podcastle price you find in an older review as a rumor until you see it on async.com.

What Auphonic does to a file

Auphonic runs a chain of separate algorithms, and each link has its own switch. That is the real difference from a one-button cleaner: after processing you can say exactly what was done to the audio, and change one thing next time.

The Adaptive Leveler, the part people come for

The Adaptive Leveler is what most people mean when they search for "auphonic leveler". It sorts the file into speech, music and background, then rides each speaker's level the way an engineer rides faders on a desk. Quiet speakers come up. Noise, wind and breaths are excluded from that boost, so the fridge does not rise along with your guest's quiet sentence. And because the episode is already finished, nothing forces the leveler to decide in a few milliseconds, which is exactly the constraint a live tool lives under.

Its strength is a number, and the documentation is specific about what each value does:

  • 100% is full leveling and the default.
  • 70% keeps roughly 3 dB more dynamic range than 100%. Good for narration that should breathe.
  • 110% (Fast Leveler) reacts much faster. It exists for the quiet audience question in a lecture, or a voice that swings between a murmur and a shout.
  • 120% (Amplify Everything) lifts as much as possible, background noise included. Rarely what a podcast wants.

A separate Compressor setting (Auto, Soft, Medium, Hard, Off; Medium is the documented default, and Auto picks Medium at strengths up to 100%) handles short spikes such as plosives and laughter. Before you worry about over-processing: if your input already sits inside a narrow loudness range, the leveler leaves it alone. Running a well-recorded file through Auphonic does not flatten it for the sake of it. For broadcast work there is also a Broadcast Mode that targets a maximum loudness range instead; Auphonic's help cites about 3 LU as typical for news and 5 LU for talk programs.

Noise and reverb reduction

Auphonic offers four denoisers, and picking the wrong one is the most common way to get a bad result:

  • Static Denoiser: steady broadband noise and reverb, and it tolerates music in the mix. It will not touch breaths or noise that changes over time.
  • Dynamic Denoiser: keeps speech and music and removes everything else, breaths included.
  • Speech Isolation: keeps only speech. Your theme music goes too, so use it on voice-only tracks.
  • Classic Denoiser: the older profile-based method for static noise and hum. It needs a silent stretch to learn from and should not be used where music plays.

The three modern ones have separate Remove Noise, Remove Reverb and Remove Breaths amounts (the Static Denoiser has no breath control). The fridge in the kitchen example is steady, so the Static Denoiser is the honest first try. The Dynamic Denoiser is for the dog, the street, the espresso machine that starts mid-answer.

Filtering, loudness and cuts

Then come the finishing steps. A high-pass filter removes rumble adaptively. Voice AutoEQ adjusts the tone of each speaker over time so a guest who drifts off-mic does not turn boomy. Bandwidth Extension synthesizes missing high frequencies, the kind of thing a guest on a phone line lacks. Studio Voice, still in beta, aims to rebuild a voice from damaged recordings with codec artifacts or clipping.

Last is loudness normalization with a true peak limiter. Auphonic's own help lists the targets: -16 LUFS for Apple and the AES recommendation, -14 LUFS for Spotify and Tidal. For a spoken podcast, -16 LUFS is the usual choice. Optional automatic cuts remove silences, filler words such as "uhm", coughs, and music segments longer than about 20 seconds, and an "Export Uncut Audio" mode lets you review the cut list before trusting it.

Auphonic vs Levelator

"Auphonic vs Levelator" is an old question that keeps coming back, because The Levelator was free and many podcasters learned on it. The rights moved to The Conversations Network in 2008. It combines compression, normalization and limiting to even out levels between voices, reads only PCM files such as WAV and AIFF, and writes a copy with ".output" in the name. Official support stopped at the end of 2012; after macOS Catalina dropped 32-bit apps, a 64-bit version appeared in the Mac App Store, still free, and its listing now asks for macOS 12 or later.

So the comparison is short. The Levelator does one thing: it evens out loudness offline, on your Mac, for nothing. No noise reduction, no LUFS target, no MP3 input. Auphonic does that plus everything above, on its servers, with two free hours a month. Auphonic also used to sell a desktop Leveler; the last announcement about it on Auphonic's blog is from June 2017 (version 1.8), and the old desktop pages on auphonic.com now return "not found". In 2026, "Auphonic leveler" means the web service.

If the file is clean and only the levels jump around, try the Levelator first. If there is noise, or a platform wants a specific LUFS figure, go to Auphonic.

What Podcastle (Async) does to a file

Async does not show you a chain. It shows you a studio. Recording happens in the browser with remote guests on separate tracks, editing is multitrack and transcript-based, and the AI tools act on the track itself.

Magic Dust is the headline feature. Right-click the audio track, choose Add Magic Dust, apply. Async's own guide describes it as the tool for removing echo and background noise and does not list what happens inside. Next to it sit AI Auto leveling (a dynamic equalizer plus loudness matching to a target), AI Noise reduction, which the pricing page describes only as reducing background noise in any file, and AI Silence removal. The Audio Assistant runs Magic Dust, Auto-Leveling and the Silence Remover in one pass.

What you give up is the dial. There is no way to tell Magic Dust "keep 3 dB more dynamics" or "this is steady hum, not chatter". For a host who wants one tab and one button, that is the appeal. For someone who wants to know why episode 14 sounds different from episode 13, it is the problem.

Remote recording is the other half of the product, and the one Auphonic does not offer at all. If your guests need a link they can click and a separate track for each voice, that is a reason to look at Async that has nothing to do with AI cleanup. Our comparison of remote podcast recording tools puts it next to Zencastr, Riverside, SquadCast and Cleanfeed.

Side by side, as of October 2026

AuphonicPodcastle (Async)
What it isProcessing engine: file in, file outBrowser studio: record, edit, clean, publish
LevelingAdaptive Leveler, strength 0 to 120%, compressor presets, Broadcast ModeAI Auto leveling, one switch
NoiseStatic, Dynamic, Speech Isolation and Classic denoisers, separate noise, reverb and breath amountsMagic Dust (echo and noise), AI Noise reduction
Loudness targetExplicit LUFS, such as -16 or -14, plus true peak limiterTarget loudness inside Auto leveling
Free tier2 hours of audio a month, jingle added, credits do not roll over, multitrack under 20 minutes10 AI credits for the life of the account, 1 audio export, MP3 up to 160 kbps, 1 hour of remote recording
Entry paid planS: 9 hours a month, €12 / $13 monthly or €10 / $11 a month billed yearly, before VATEssentials; the price loads by script on async.com/pricing, check it there
Remote recordingNoYes
Works during a live callNoNo

What an hour actually costs

Auphonic bills by the length of the audio you send, with a 3-minute minimum per production. A 48-minute episode with a 1-minute intro uses 49 minutes. Output formats do not add to it, so an MP3 for the feed and a WAV master from one run cost the same as one file. On the free tier that means two episodes of that length a month, each carrying Auphonic's jingle. Plan S at 9 hours covers eleven. Monthly credits reset at the end of the month; the one-time packs (5 hours for €11 / $12, larger packs cheaper per hour) never expire, which suits a show that publishes in bursts.

Async's free plan works as a trial: one audio export in total and 160 kbps MP3. The paid tiers (Essentials, Pro, Teams, and Business on request) are where Magic Dust gets daily use. The page advertises 40% off for yearly billing, but the dollar amounts are drawn by a script rather than printed in the page, so read them on async.com/pricing on the day you buy.

So which one makes a podcast voice sound better?

For a recorded episode where the raw sound is decent and the problems are uneven levels, steady noise and platform loudness, Auphonic. Not because its AI is smarter in some general sense, but because you can match the tool to the fault: Static Denoiser for the fridge, leveler at 100% for the guest who drifts, -16 LUFS for the feed. When the result is off, you change one setting and run it again. Async wins when you also need the recording and editing, or when your guests need a link-and-record studio and you would rather not juggle three services.

Neither one recovers what the mic never captured. Every denoiser works by deciding which parts of the signal are voice. Where noise and voice share the same frequencies, the cleaner either leaves some noise or takes some voice with it, and heavier settings push that trade toward thin, watery speech. The fix that costs nothing happens before the upload: the mic closer to the mouth, a softer room, a guest wearing headphones so your voice does not bleed into her track. Our guide to podcast recording software for a Mac covers that side, and removing noise from a recording compares the free cleanup routes, Audacity included, if neither paid service appeals.

A starting chain for the kitchen-guest episode

Auphonic, single-track production
Input:            WAV, or MP3 at 192 kbps or higher
Adaptive Leveler: Default Leveler, Leveler Strength 100%, Compressor Medium
Denoiser:         Static Denoiser for steady fridge hum
                  Dynamic Denoiser instead if the noise changes (traffic, dog)
Remove Noise:     default is full; lower it if the voice turns thin
Filtering:        High-pass on, Voice AutoEQ on
Loudness:         -16 LUFS, true peak limiter on
Cuts:             off for the first run, or Export Uncut Audio to review
Output:           MP3 for the feed, WAV as the master

When each host recorded a separate track, a multitrack production levels every speaker on their own track and then mixes them; on the free tier those productions must be under 20 minutes.

How to use Auphonic in five minutes

  1. Register an account. The free tier gives two hours of processing a month.
  2. Open the new production page (auphonic.com/engine/upload) and choose the file. For lossy formats Auphonic asks for 192 kbps or higher, so export WAV if you can.
  3. Add intro and outro if you use them. They are only loudness-matched, not cleaned, so prepare them beforehand.
  4. Set the algorithms as in the chain above and save the lot as a preset. Next week's episode then takes one click.
  5. Start the production, download the result and keep your own copy: Auphonic stores files only for a limited number of days.

Why neither helps on a live call

Both services wait for the recording to end. Loudness normalization, for one, measures the finished file and then applies a single constant gain to reach -16 LUFS. A call never becomes a finished file. On Zoom, Google Meet or Discord your voice leaves the Mac within a fraction of a second, and any cleaner on that path has only the last few milliseconds to decide what is voice and what is noise. Upload-and-wait processing cannot sit in that path at all.

So if the people who hear your podcast also hear you on weekly calls, you need two different tools: a post pass for the episode and a live chain for everything else. Free live options exist. Zoom has its own audio settings, covered in our Zoom guide, and on a Mac from 2018 or later, Mic Mode in Control Center offers Voice Isolation in apps that support it, FaceTime among them. Our overview of noise cancelling software compares the rest.

That live slot is where TunedMic fits, and only there. TunedMic is a Mac app that adds one more microphone to the system, also called TunedMic, which any app with a mic menu can select: Zoom, Meet, Teams, Discord, Telegram, OBS, a browser. It takes your real mic, processes the voice locally in real time with less than a tenth of a second of delay, and every app that picks it hears the processed sound. Setup is a one-minute Sound Check: you read a short passage aloud and it takes more than 30 measurements of your voice, your mic and your room, and its nine profiles are built from that measurement rather than a stock preset. A recording app can pick the TunedMic input too, so the file you later send to Auphonic starts from the processed voice instead of the raw mic. Whether that helps depends on the episode: Auphonic's help, for one, asks you not to level or gate audio before its Classic Denoiser, so compare a short test both ways.

Its limits are plain. TunedMic needs a Mac with Apple silicon (M1 or newer) and macOS 14 or later; there is no Windows or Intel version. It does not remove every noise, and it does not replace Auphonic's loudness targeting on a finished episode. It costs $19 once, with a free 7-day trial that starts at first launch and needs no card or account. Payments are final, so use the trial to decide.

Your next step

Take last week's episode, the one with the quietest guest. Run it through Auphonic's free tier with the chain above, then listen to two places: the first two minutes and the loudest laugh. If the levels were the whole problem, you are done, and the free two hours may be all you ever need.

If the real problem is how you sound when nothing is being recorded, on the calls in between, watch the 38-second video on the TunedMic home page: you read aloud for a minute, TunedMic tunes the mic to your voice, and after the call you get a Speech Report. Then decide whether the 7-day trial is worth an afternoon, and check pricing only after that.

Keep reading