The Refiner
Private betaSign-in is currently invite-only. Email support@therefiner.app for access.

Refine Suno tracks. Studio audio quality.

Suno and Udio outputs sound compressed even when you export as WAV — the lo-fi is in the generated audio itself, not the file format. The Refiner upgrades the actual audio fidelity, restoring high-frequency detail and clarity. Drop a file, get back a release-ready WAV at 44.1 / 48 kHz.

Start free — no card neededSee pricing

Free monthly credits on signup. $9/mo for 16 credits, $29/mo for 30, $49/mo for 60. Spend them on any preset.

Stacked spectrogram of a real Refiner job: the Suno original (top) cuts off around 11 kHz where the lossy compression dropped high-frequency detail; the Flow Matching output (bottom) restores the full band.
A real job, not a demo. Top: Suno original, flat above ~11 kHz. Bottom: Flow Matching output, high-frequency detail rebuilt.

Hear the difference

Same track, two refinements. Headphones recommended.

Enhanced

Fast

Enhanced runs in seconds and is what we recommend for most tracks.

0:00 / 0:00

Refined with Enhanced (44.1 / 48 kHz WAV)

Flow Matching

Detailed

Flow Matching takes about ten minutes and reconstructs the finest detail. Costs 3 credits on Normal quality or 4 on High.

0:00 / 0:00

Refined with Flow Matching (44.1 / 48 kHz WAV)

The "before" tracks cut off around 11 kHz — that's where the lossy compression threw away high-frequency detail. Enhanced restores most of it almost instantly. Flow Matching rebuilds it more carefully — the cost is time, the gain is finer detail in cymbals, breath, and reverb tails.

See the difference

Spectrograms of the passage above. Brighter horizontal bands at the top mean more high-frequency content recovered.

Source · original Suno outputmp3, 128 kbps · cuts off around 11 kHz
Source · original Suno output — mp3, 128 kbps · cuts off around 11 kHz
Refined with Enhanced44.1 / 48 kHz WAV · high-frequency detail restored
Refined with Enhanced — 44.1 / 48 kHz WAV · high-frequency detail restored

How it works

Step 1

Drop your track

mp3, wav, or flac, up to 8 minutes. Stereo or mono.

Step 2

Pick a refinement

Three models, each tuned for different damage: Enhanced (recommended default), Restoration (stronger on heavily-degraded mp3s), and Flow Matching (Pro) for the finest detail.

Step 3

Get a clean WAV back

44.1 / 48 kHz, ready to master or upload. We keep your file for 7 days, then delete it.

Built for AI-generated music, end to end

AI tracks have a specific shape of damage: high-frequency rolloff, compression artifacts, sometimes detectable temporal patterns from the generation process. Generic audio restorers weren't built for this. We were.

Built for AI sources
Why it matters
Models trained on the failure modes specific to Suno, Udio, and lossy mp3s
Generic restorers introduce their own artifacts trying to fix problems they don't recognize
Flow Matching preset uses a specialized generative model that rebuilds spectral detail rather than guessing it
The diffusion process strips and reconstructs the sound from scratch, so the output sounds like a real recording rather than an upscale. As a side effect of the rebuild, most generation-time AI artifacts are eliminated. The point is sound quality — not a tool to bypass AI detectors.
Optional AI-artifact cleaning on Enhanced runs
Enhanced uses a neural vocoder, so its output carries faint vocoder artifacts; the optional cleaner pass scrubs them after refinement. Not designed to defeat AI detectors — it's a finishing step.
44.1 / 48 kHz WAV output, distribution-ready
Drop into release without re-exporting from a DAW. Pro adds 32-bit float for further mixing.
Also available

Split any track into stems

Isolated vocals, drums, bass, and everything else — or speech vs. background for spoken-word audio. Drop a track and pick an engine; get back one WAV per stem plus a ZIP for one-click download.

Demucs

4 stems · fast classic

Vocals, drums, bass, and everything else. Quick and solid for everyday remixing or DAW work.

Melband

4 stems · cleaner separation

Mel-Band RoFormer — a newer transformer that pulls stems apart with less bleed. Takes longer in exchange for cleaner isolation.

Voice Cleanup

2 stems · speech / background

Speech on one stem, everything else on the other. Built for podcasts, interviews, and voice notes.

Pro feature · 3 credits per split, any engine · Free users get one lifetime trial split to try it.

Pricing

Free to start. $9/mo when you need more.

Plans

Credits are credits — costs vary per preset, not per plan.

Free
$0
2
credits / month
  • Enhanced model✓
  • Restoration model✓
  • Flow Matching model✓
  • AI detection✓
  • AI artifact removal✗
  • Stem separation1 lifetime trial
  • Bit depth24-bit
  • Rate44.1 kHz
  • SupportEmail
Hobby
$9/mo
16
credits / month
  • Enhanced model✓
  • Restoration model✓
  • Flow Matching model✓
  • AI detection✓
  • AI artifact removal✗
  • Stem separation✗
  • Bit depth24-bit
  • Rate44.1 kHz
  • SupportEmail
Pro
$29/mo
30
credits / month
  • Enhanced model✓
  • Restoration model✓
  • Flow Matching model✓
  • AI detection✓
  • AI artifact removal✓
  • Stem separation✓
  • Bit depth24 or 32-bit float
  • Rate44.1 or 48 kHz
  • SupportEmail
Studio
$49/mo
60
credits / month
  • Enhanced model✓
  • Restoration model✓
  • Flow Matching model✓
  • AI detection✓
  • AI artifact removal✓
  • Stem separation✓
  • Bit depth24 or 32-bit float
  • Rate44.1 or 48 kHz
  • SupportPriority email
See full pricing →
Field notes

Guides

Practical writing on getting AI-generated music release-ready.

Why Suno tracks sound compressed

The two layers of audio quality loss in a Suno export, and how refinement fixes both.

Enhanced vs Flow Matching

Which refinement preset to pick — the speed, cost, and fidelity trade-offs explained.

Releasing AI music

Format, loudness, and AI-disclosure checklist for Spotify, Apple Music, and the rest.

All guides →

Frequently asked questions

Quick answers. Full guide on the help page.

What is The Refiner?
The Refiner is a web app that cleans up and upgrades audio tracks. It works on anything — AI-generated music from Suno or Udio, demos, podcast recordings, old mp3s, live captures — drop in a lossy or low-fidelity file and get back a release-ready WAV with restored high-frequency detail, better clarity, and a more polished sound.
What's the difference between Enhanced and Flow Matching?
Enhanced is the fast preset — it runs in seconds and restores most of the high-frequency detail that lossy compression discards. It's what we recommend for most tracks. Flow Matching takes about ten minutes and reconstructs the finest detail (cymbals, breath, reverb tails) using diffusion. Costs 3 credits on Normal quality or 4 on High.
What audio formats can I upload?
MP3 and WAV. Tracks can be up to 8 minutes long. Outputs are 44.1 or 48 kHz WAV at 24-bit (32-bit float is available on Pro and Studio for Flow Matching, useful if you're feeding the result back into a DAW for further mixing).
How long does a refinement take?
Enhanced typically completes in 10–30 seconds. Flow Matching takes around 6–12 minutes depending on quality setting and track length. You'll see live progress in the dashboard and can keep working while jobs run.
Do I need a subscription?
No. The free tier gives you a small number of credits each month to try the service. Paid plans (Hobby, Pro, Studio) and one-time credit packs are available if you need more.
Can I split a track into stems?
Yes — stem separation is a Pro feature. Drop a track on the Stems page and pick an engine: Demucs (fast 4-stem classic), Melband (newer transformer, cleaner separation), or Voice Cleanup (speech vs. background, great for podcasts and interviews). Output is one WAV per stem plus a ZIP of all of them. 3 credits per split. Free users get one lifetime trial split to try it.

Used by indie musicians and podcasters

Preparing AI-assisted tracks for release.

Start free