Recordings only, as ever — these are tools for files that already exist, graded from the twelve tutorials on this site.
Four categories, four winners: Audacity is the best free desktop tool, Adobe Audition the best paid one, Adobe Podcast Enhance the best online cleaner, and CapCut the most useful mobile option. Every grade below comes from running the tool's own tutorial on real files — timings and limitations included.
| Tool | Category | Price | Batch | Formats | Learning curve | Grade |
|---|---|---|---|---|---|---|
| Audacity | Free desktop | $0, open source | No | WAV/MP3/FLAC/OGG | An afternoon | A− |
| DaVinci Resolve (free) | Free desktop | $0 | No | Anything it edits | Steep (it's an NLE) | B |
| Adobe Audition | Paid desktop | Creative Cloud sub | Yes (scripts) | Every pro format | Moderate | A |
| DaVinci Resolve Studio | Paid desktop | One-time license | No | Anything it edits | Steep | A− |
| Premiere Pro | Paid desktop | Creative Cloud sub | No | Anything it edits | Moderate | B+ |
| Adobe Podcast Enhance | Online | Free 1 hr/day · $9.99/mo | No (queue by hand) | Audio in, WAV/MP3 out | None | A |
| Cleanvoice | Online | ~€1/processed hour | Yes | Common audio formats | None | B+ |
| Auphonic | Online | 2 free hrs/mo · from €9/mo | Yes | Broad, incl. video | Gentle | B+ |
| CapCut | Mobile | Free (basic toggle) | No | Video projects | None | B |
| iPhone Voice Memos | Mobile | Free, built-in | No | Memos (M4A) | None | B− |
| Samsung Audio eraser | Mobile | Free on Galaxy AI | No | Recorder/video/notes | None | B |
The two-pass profile method is still the best noise removal $0 buys: precise, offline, endlessly redoable, and honest about its one demand — your time. It struggles only where all profile tools do: noise that changes. Budget five to ten minutes per file and it repays you with control no toggle offers.
Dialogue Leveler with Background reduction is real help hiding inside a free professional NLE — worth using when you're editing video there anyway. It levels first and denoises second, so treat it as polish; the rescue-grade tool (Voice Isolation) lives behind the Studio license.
The noise print method with better ears than anything else in this library: surgical on steady noise, scriptable for batches, with Adaptive Noise Reduction as the fallback when noise drifts. The grade is for capability; the honest caveat is that most weeks you won't need this much scalpel.
Voice Isolation is the strongest single dialogue-cleanup control inside any video editor — AI separation that survives traffic, wind and chatter, on a slider. If your work is video and your noise is serious, this is the category's deep answer; the learning curve is the whole of Resolve.
Essential Sound's Reduce Noise checkbox is the shortest path in this category from noisy to fine — one tag, one tick, one slider. It's adaptive rather than profile-based, so it trades precision for speed; for dialogue that needs actual rescue, export to Enhance and come back.
The strongest free button on the web for damaged speech — it rebuilds the voice instead of filtering around it, which saves recordings everything else abandons. Its taste is assertive (keep the strength near 60–80%) and its meter is real, but as a first stop for rough audio it has no free rival.
Noise, ums, mouth sounds and dead-air gaps handled in one automated pass, priced by the hour with credits that keep for two years. No perpetual free tier and no interest in live anything — it's a pipeline, and for weekly shows the time-for-money trade is the best on this page.
The quiet professional: loudness normalisation to podcast standards with noise and hum reduction in the same pass. Two free hours a month is a real allowance (with an audible jingle on free productions), and one-time credits never expire. If your audio sounds amateur, the fix is as often levels as noise — this does both.
One tap, no slider, sensible results on mild steady noise — and a hard ceiling everywhere else. The grade reflects reach as much as quality: it's already on the phone where the noisy video lives, and the keep-or-undo test takes thirty seconds.
Enhance Recording is the rare cleanup with an undo that costs nothing — a fixed recipe, applied non-destructively, right where memos live. It's speech-only and modest, which is exactly what a built-in should be; the export route covers what it can't.
Genuine source separation on a phone: the recording un-mixes into voices, wind and noise, each on its own fader. It's the most capable mobile tool here — gated to Samsung's Galaxy AI devices, which is the only thing keeping it from the category win.
Audacity's two-pass method costs five to ten minutes per file once you know it; Audition's noise print about five. That's fine for one file and brutal for forty: a season of weekly episodes is two full working days of profile-and-apply. The pipelines invert the cost — Cleanvoice cleans an hour of audio for about a euro while you do something else, and Auphonic's two free monthly hours cover a small show outright. My split, and the one I recommend: learn the manual method (you'll rescue things pipelines can't, and you'll understand what the machines are doing), then pay the pipeline the moment cleanup becomes a schedule instead of an event.
Every tool here has a full tutorial on this site — the grades fall out of writing them: result quality on steady and messy noise, effort per file, price honesty (meters, jingles, gates counted), and how gracefully each tool fails. Menu paths and prices were verified Jul 28–29, 2026 and are dated in each tutorial's prerequisites strip. Affiliate buttons never move a grade; two of the four category winners pay us nothing.
For one rough file with no budget: Adobe Podcast Enhance's free hour. For a regular workflow: Audition if you want control, Cleanvoice if you want your time back. There's no single winner — that's why this page ranks by category instead of pretending there is.
For steady noise under decent speech — most real cases — yes: Audacity handles it, and Enhance rescues the rough ones inside its daily hour. You start needing money when cleanup becomes weekly (metered pipelines) or when dialogue must survive heavy, changing noise (Resolve Studio's isolation).
On damaged audio, clearly — reconstruction beats subtraction when there's little signal left. On decent audio it's closer: a careful manual pass is more transparent, while AI tools trade a little character for a lot of convenience. The tutorials teach both so the choice is yours per file, not per ideology.