• New: record right in Denoise. Tap + and choose Recorder. Noise is removed while you record, so most recordings are clean the moment you stop. • Recording keeps going with the screen locked, and you can pause or stop it from the Lock Screen or the Dynamic Island.

• New filter, AI · Shorten: makes long recordings shorter by trimming long pauses and speeding up speech without changing the voice. Choose how much with the slider. • Easier to use with VoiceOver, Voice Control and larger text sizes, with higher contrast text. • Audio keeps playing when you lock the screen, and you can play, pause or skip from the Lock Screen.

• Adapted for iPad: your recordings stay in a list next to the editor, so you can switch between them without going back, and the editor makes use of the wider screen. • Bug fixes for older iOS versions: on iOS 17, cleaning with AI · Auto or AI · DPDFNet could close the app. That is fixed. AI · DPDFNet now needs iOS 18 or later, and on iOS 17 AI · Auto uses our earlier AI engine.

• Fixed a crash on long recordings. With recordings over 5 minutes, AI · Auto and AI · DPDFNet could close the app partway through, mostly on iPhones with less memory. Long jobs now use far less memory. • If the AI model runs into a problem in the middle of a recording, AI · Auto now finishes the job with its backup model instead of stopping. • AI · Sidon also uses less memory on long recordings.

Your recordings, heard properly. • A new noise cancellation model. Voices sound clearer and more natural, and quiet passages keep the words they used to lose. • Rebuilt speech restoration. Muffled, damaged and phone quality recordings come back far more convincingly, and short clips are restored properly now, not only long ones. • Sharper music and vocal separation. A newer model pulls singing away from the backing track more accurately. • The reduction slider responds as you drag it, so you hear a setting before you commit to it. • Large model downloads wait for Wi-Fi and ask before using cellular data. Everything runs on your iPhone. Nothing is uploaded. Tell us how it sounds. Your recordings are what we tune for, and every message reaches the person who builds the app.

• AI · Auto keeps your recording's natural sound. Recordings that are already full quality now get a single cleaning pass, so they finish sooner and keep their real high frequencies instead of having them rebuilt. Speech restoration is still used where it helps, such as phone calls, messenger voice notes and heavily compressed files. • Speech restoration is more accurate. The models behind AI · Denoise + Restore, which AI · Auto also uses for phone-quality audio, were rebuilt: restored voices stay closer to the original and get less artificial hiss added on top. Also fixed a rare glitch that could repeat or garble a short stretch of the result. • Updated for iOS 27.

• VoiceFixer is fixed. On recent iOS versions it could finish with a silent result. It now restores speech reliably, and if processing ever fails the app says so instead of saving silence. • Fewer rejected imports. Recordings the app can actually read are no longer refused, including files without an extension or with a renamed ending. When a file truly cannot be opened, the message now explains the reason. • Faster and more reliable AI model downloads. On iOS 26 and later, models are delivered by Apple's servers. Downloads that previously stalled in some regions now complete, and downloaded models are shared with the Photos extension. • Shortcuts automations can fetch a missing AI model on their own instead of stopping with an "open the app first" error. • More reliability fixes: cleaning videos with WAV output selected, honest error reporting when a save fails, and a purchase screen that recovers if products fail to load.

• WAV export now works with every engine. If you choose WAV under Options, your cleaned recording is saved as a lossless WAV file instead of M4A. Previously the WAV setting was only applied by the Classical engine, so AI cleanups always came out as M4A. • Fixed crashes while cleaning a recording. Some files could close the app instead of cleaning, and stopping a job partway through could close it while it was still saving. Files we cannot read now show a clear message instead. • More reliable saving. Fixed a rare failure that could stop a finished recording from being saved when Normalize loudness was turned on, and cancelling a job no longer discards the previous cleaned version.

• Import several files at once. Pick multiple recordings in the Files or Photos picker and they all land in your library, with progress shown while they load. • Clear feedback when a file cannot be used. If a recording is in an unsupported format, has no audio track, or is too short to clean, Denoise now tells you right away instead of adding a broken item. • Smoother imports. Adding files no longer blocks the screen, so you can keep listening or editing while they load in the background.

• Opens WhatsApp and Telegram voice notes. Audio saved as .opus or .ogg now imports and cleans up like any other recording, fully on device. • New: a 3 Exports pack. Save or share three cleaned recordings for a small one-time price, no subscription needed. • Small fixes and translation updates in all 20 languages.

• Opens AMR audio files. Voice notes and voicemails saved as .amr (common from Android phones and carrier voicemail) now import and clean up like any other recording. • Long recordings process more reliably: your screen stays awake while a file is processing, and you get a notification when it is done. • Fewer "couldn't open" and "couldn't save" errors across more audio formats and sample rates. • Interrupted model downloads now resume instead of failing, and AI restoration is more reliable across devices.

• Fixed a crash that could happen when opening some recordings to clean them. • Saving is more reliable: recordings that sometimes failed to export now save dependably. • Various stability improvements.

Polishing release - for the reliability sake: • Cleaning long recordings is more reliable. "Denoise + Restore Speech" no longer fails partway through a long file. If the speech-restore step can't finish, you still get the fully cleaned audio instead of an error. • Cleaning audio from the Photos editor now works dependably on longer clips. • Recovers automatically when on-device processing runs low on memory, instead of giving up. • Model add-ons (Speech Restore HD, Music) now retry on their own and resume where they left off if your connection drops. • Various stability fixes.

• Switch to other apps while a recording is being cleaned up and nothing gets lost. It picks up right where it left off when you come back. • Added a handy search to quickly find your recordings. • Polished the look across several screens. • Did you know you can clean up audio faster by opening a file straight from the Files app ("Open With" > Denoise)? It now works reliably. We fixed a crash that could happen when opening files that way.

• Switching between the original and cleaned audio on a video no longer crashes. • AI · Music separation no longer fails or crashes on some tracks. • Unsupported files (like OGG) now prompt you to convert to WAV or M4A right away, instead of failing after a long wait. More low-sample-rate recordings export reliably too. • Fixed text that was hard to read on the upgrade screen in some display modes. • Smaller app download.

• Louder by default - cleaned recordings now target -14 LUFS, so your audio is ready to post without a manual volume boost. This was the most-requested fix; you can still pick a different level in settings. • Share video into Denoise - clean the audio on a video clip and get the video back, right from the share sheet. • Save to Files, plus a new "Cleaned" folder - exported recordings show up in Files under On My iPhone > Denoise > Cleaned, ready to drop straight into LumaFusion and other editors. • Shortcuts - automate cleanup with the new "Clean Audio or Video" action. • iPad multitasking - Split View support, plus drag and drop: drag a clip in, drag the cleaned result out. • Wider format support - more voice notes and recordings (including low-sample-rate and additional M4A/AAC files) now import and export reliably. • Stability - fixed crashes in AI · Music separation and the recordings list.

Denoise 4.0: a full redesign, rebuilt around on-device AI. • One-tap AI denoise. A voice-trained neural network cleans your recording in seconds. 100% on-device, nothing is uploaded. • Pick your engine. Fast everyday cleanup, cleaner highs, or premium restoration for badly damaged audio. • Clean videos too. Denoise the soundtrack and keep the original picture. • Split music and vocals. Isolate the voice or make an instrumental. • Before and after. Toggle original vs. cleaned to hear exactly what changed. • Fine-tune it. 5-band EQ, adjustable strength, and true loudness (LUFS) leveling. • A faster, simpler editor and library. Built for podcasters, journalists, students, and anyone with a recording they thought was unusable.

- Equalizer improvements - Stability improvements

- Stability improvements on lengthy videos

- Stability improvements - Memory consumption with huge video libraries

- Fixed issues with high-bitrate memos

- Stability improvements

- AI-guided mode for automatical voice recovering - Fixed Equalizer in IOS 14+ - Stability and memory usage improvements

- Memory consumption optimizations

- Memory consumption optimizations