AudioWrench β free audio tools that run in your browser.
AudioWrench is a toolbox of 139 currently free audio tools plus BalanceDAW, a browser music studio. Core file processing runs locally; optional models and browser speech modes explain their data path before use.
Current design commitments.
π Private by design
Core audio workflows process selected files on-device with the Web Audio API and WebAssembly and do not send them to AudioWrench. Optional models and browser speech modes explain their data path before use. See the exact data flow β
π Currently free
No account, subscription, payment or export watermark is currently required. This is the current release state, not a guarantee that availability or terms can never change. Practical duration depends on browser memory and device speed.
β‘ Instant
Nothing to install and no upload queue. Small focused jobs can finish quickly; full-file analysis, model downloads, long audio and multi-output renders depend on the browser, device memory and file length.
Voice and intelligent DSP: live or local-model transcription, system or neural speech, instant or neural vocal separation, voice matching, Auto-Tune, repair and creative voice effects. Each page states which mode it uses.
A browser-native digital audio workstation β multitrack timeline, piano roll, synths and effects, powered by a RustβWebAssembly engine. Open the studio β
π Learn hub
Step-by-step guides for the most asked-for jobs: mastering a song, removing vocals, making slowed + reverb and more. Browse the guides β
The story.
AudioWrench started as a simple idea: everyday audio jobs β trimming a clip, boosting bass or preparing a recording β should not require a heavy desktop setup. Modern browsers can handle serious audio work locally, so that is where we build.
Whatβs new.
Updates to the tools you use, newest first. Version numbers use the release date followed by that dayβs update number (UTC).
Numbered releases begin on 20 September 2026. This first entry also includes the Auto-Tune and filename fixes shipped earlier that day.
v2026.10.14.3Current release
Current sample tools for returning browsers
Fixed. Sample Pack Builder and SFZ Sampler Builder now load content-versioned entry modules, so returning browsers receive the current processing and safe status logic rather than a previously cached module. Build a current sample pack β
v2026.10.14.2
Safe sample names throughout loading and error recovery
Fixed. Sample Pack Builder and SFZ Sampler Builder now keep filenames and local failure messages as literal text in every progress and status panel, including brief decoding states. Original source names, audio processing and package data remain unchanged. This closes a transient display path found by the full filename regression suite. Build a current sample pack safely β
v2026.10.14.1
One current sample library from audition to complete delivery
Improved. Sample Pack Builder binds its real Source, Result and Result-minus-Source review to the current accepted queue and recipe. Explicit full-file builds deliver matching WAV, ZIP, JSON and spreadsheet-safe CSV bytes. New selections are atomic and memory-checked; cancellation, replacement, edits and clearing retire old players, packages, receipts and downloads not yet offered. Review and deliver a current sample crate β
Fixed. SFZ Sampler Builder keeps sample names as inert text, retires delayed audition starts and old exports, and packages the current editable key/velocity map with its PCM16 WAVs and JSON. Responsive region cards, natural keyboard focus, readable coverage and explicit stopped/error states accompany complete instructions on both existing tools. Browser audition remains a bounded monophonic check; destination-engine tuning, loops and SFZ 2 paths require verification. Map, audition and export the complete current instrument β
v2026.10.11.7
One current batch from gain review to audio or album QC delivery
Improved. Batch Loudness Matcher now provides genuine bounded per-track Source, Result and Source-minus-Result players beside its measured-versus-predicted gain plan. Full-length WAV and ZIP delivery stays separate. Pending selections, edits, cancellation and leaving the page retire old players, files and receipts; complete selections and results are committed atomically with a retry path and retained memory guards. Review and deliver a current matched batch β
Fixed. Album Loudness Analyzer blocks a previous CSV or JSON while a new tracklist loads, exposes cancellation during loading and prevents late decoded files or partial analysis from restoring a retired report. Current order and reference flags accompany offered-file receipts; CSV treats formula-leading text as literal while JSON preserves original filenames. Both existing tools have complete instructions, responsive track cards and keyboard-safe row controls; loudness and peak methods remain explicitly approximate. Inspect and export the current album QC report β
v2026.10.11.6
Keep separation instructions readable with enlarged and fallback fonts
Fixed. Long instruction labels in Vocal Remover and Stem Splitter now wrap inside narrow pages with enlarged text or fallback font metrics, without hiding overflow or shrinking the text. The stronger layout check covers both the site font and a wide fallback; separation audio, model identity and export behavior are unchanged. Use the complete current-song workflow β
v2026.10.11.5
Keep separation audition and delivery bound to the current song
Fixed. Vocal Remover and Stem Splitter retire previous audio, full results and downloads as soon as a new source or demo is selected. Late decodes, renders and MP3 encoders cannot replace a newer song or export old audio under a new name. Cancel, Escape, source replacement and leaving the page discard obsolete work with a clear retry path and conservative full-render memory checks. Separate the current song β
Improved. The two existing separation workflows now distinguish short review from full-length output, show current-source export receipts and explain cancellation, bleed and file verification. Stem Solo/Mute exposes its listening state without changing the individual downloads. Neural beta stays opt-in, can retire model fetch or late inference and never silently changes to DSP after an error. Review and deliver the current full stems β
v2026.10.11.4
Complete current mastering delivery with readable interaction states
Fixed. Mastering Studio keeps secondary action text readable after hovering or focusing cancellation, pre-processing, codec and WAV buttons. The same current-source mastering journey now also gives module names adequate space in its existing guide's comparison table, without shrinking text or removing horizontal keyboard scrolling. Audio processing and measurements are unchanged. Use the complete current mastering workflow β
v2026.10.11.3
Keep mastering feedback, analysis and delivery bound to the current mix
Fixed. Mastering Studio retires the previous audio, frozen measurements, peak holds and download results as soon as a new file or demo is selected, including preflight and decode. Late Assistant, reference and pre-processing results cannot replace newer source or manual decisions. Cancel, Escape and leaving the page discard pending work and codec playback; failed replacement leaves a clear retry path rather than silently reusing old audio. Master the current mix β
Improved. Rendering and encoding now have owned pending states, caught failures and current-source download receipts for WAV 16-bit, WAV 32-bit float and MP3 320. The seven existing visualizers and audio algorithms are preserved. The expanded existing guide explains settings versus measurements, audition versus export range, cancellation, codec checks and reopening delivery files, with readable enlarged-text and keyboard workflows. Read measured mastering feedback and verify the file β
v2026.10.11.2
Complete the current metadata workflow with readable enlarged text
Fixed. MP3 Metadata, FLAC Cover Art and BWF Metadata now let long headings, action labels and expanded workflow diagrams wrap within narrow screens with enlarged text and different system-font metrics. This completes the same current-source and verified-file workflow introduced in the preceding update, without shrinking text or hiding overflowing labels. Use the complete current metadata workflow β
v2026.10.11.1
Keep metadata edits, artwork and delivery files bound to your current selection
Fixed. MP3 Metadata and FLAC Cover Art retire previous source files, queues and download receipts when a new selection begins, including file checks and demo loading. Pending cover reads cannot export the previous artwork or undo an explicit Remove. Cancel, Escape and leaving the page retire unfinished selections; failed reads offer a clear retry or another-file path. Edit current MP3 tags and artwork β
Fixed. BWF Metadata keeps imported descriptions, chunk identifiers and before/after values as literal text. A newer WAVE retires old fields and downloads immediately. Editing one field preserves the original line endings of unedited coding history, INFO comments and iXML; verified output still compares original format, audio data, unknown chunks and padding bytes. Inspect and verify the current Broadcast Wave β
Improved. Artwork replacement now checks actual JPEG/PNG signatures and browser image decoding before accepting a cover. All three existing pages include worked examples, cancellation and retry guidance, preservation boundaries and recipient checks, with matching visible structured instructions and readable mobile export states. Batch cancellation keeps controls locked while an unabortable file read settles, preventing overlapping full-file exports. Prepare current FLAC artwork and a verified bundle β
v2026.10.09.4
Control generated music and effects from current settings to audition and file
Fixed. Arpeggiator, Melody Generator, Chord Progression Generator and SFX Generator now own their audition through preparation, browser audio resume and playback. Stop stays actionable while Play is preparing; Stop, Escape, editing or leaving the page retires that audition without a late restart. Finished playback and audio errors return to an explicit, retryable state without closing shared browser audio. Generate and control an arpeggio audition β
Fixed. Editing, Escape or page leave discards obsolete pending audio downloads across the four generators. Current MIDI/JSON events and deterministic synthesis remain intact. SFX removes rendered waveforms and PCM measurements when settings change, so old audio is not presented as the new result. Inspect and save the current procedural effect β
Improved. All four existing tool pages explain preparing versus playing, Stop versus cancel, bounded chord audition versus complete render, and MIDI events versus reference-synth WAV/MP3. Visual captions distinguish calculated or illustrative plans from actual rendered output. Matching visible steps and structured instructions, accessible status messages and narrow-screen text wrapping make the complete workflow easier to follow. Read the full progression-to-file workflow β
v2026.10.09.3
Keep a binaural stereo session faithful from tone plan to saved file
Fixed. Binaural Beats retains the actual Volume in preview and files: 0% is silence and quiet exports are not boosted. One owned stereo graph gives separate left/right frequencies, native audio-play-time stop deadlines, immediate Stop and cancelled starts. Current-setting ownership discards obsolete WAV/MP3 results and page-leave jobs; memory and storage errors have a usable retry or normal download route. Build and save a current stereo-tone session β
Improved. The workspace names the left carrier, right-minus-left gap and both channel frequencies, with a readable fixed-scale calculated tone plan rather than brain-state telemetry. Frequency-only presets, independent file duration and timer semantics are explicit. Short 10/30-second exports join 1/5/10-minute sessions; local WAV, MP3 and editing handoff preserve current levels, with lossy and non-gapless limits stated. Read the channel plan and session controls β
Improved. The existing binaural-beats research guide now covers the complete stereo-to-file workflow, worked frequency and file-size tables, digital gain versus ear sound level, timers, fresh fades, import checks and troubleshooting. Six static steps, matching FAQ/schema, keyboard-scrollable tables and enlarged/fallback-font checks keep the explanations usable. Research limits remain qualified; there is no sleep, focus, brain-state or calibrated hearing guarantee. Read the expanded stereo, file and research guide β
v2026.10.09.2
Take a defined metronome bar from practice to a current click-track file
Improved. Metronome now names what BPM counts, with explicit dotted-quarter, quarter or eighth units in 6/8, relative subdivisions and a readable generated onset plan. One owned native loop supports genuinely silent Volume 0, immediate Stop, cancelled starts and no queued catch-up clicks. Save 1β16 current whole bars as WAV, MP3 or a local handoff with the chosen rhythm and volume retained. Device latency, suspension, file rounding and lossy MP3 remain explicit limits. Use the metronome practice-to-file workspace β
Added. A complete tempo-unit, meter, subdivision and click-track guide includes worked bar durations, equivalent 6/8 tempos, an original explanatory diagram and import checks. Named keyboard-scrollable tables, enlarged-text and fallback-font checks keep the examples usable on narrow screens. Sources explain the timing boundaries; there is no practice-benefit or DAW-sync guarantee. Read the tempo, meter and click-track guide β
Fixed. The colored-noise guide social titles and descriptions identify its own subject instead of copied hearing-tone text. A regression guard checks all Learn route, image and cross-article social identities without forcing legitimate abbreviated titles to be identical. Read the colored-noise guide β
v2026.10.09.1
Keep the noise guide readable with enlarged and fallback fonts
Fixed. The colored-noise guide wraps long terms safely at 320 px and 200% text, including fallback font metrics. A reproduced compound-word overflow is removed; full explanations and keyboard-scrollable tables remain readable. This closes the same noise mix-to-file journey without changing generated audio, timer or exports. Read the complete colored-noise guide β
v2026.10.08.7
Keep the noise preview beside controls, clear of export
Fixed. The desktop Noise Generator preview is contained by its control row. It stays useful beside the sliders but stops before the save panel, so scrolling cannot cover WAV, MP3 or Continue editing. This completes the same mix-to-file journey; samples, gain, timer and export behavior are unchanged. Use the complete noise mix-to-file workspace β
v2026.10.08.6
Keep the noise mix you chose from controls to file
Improved. Noise Generator uses one stable four-color pattern for playback, the actual sample preview and the current thirty-second WAV or MP3. Master volume stays meaningful in the file instead of being erased by peak normalization; zero really mutes. Reducing-only headroom, digital peak/RMS, one owned playback graph, editable native audio-clock timers, immediate stop and cancellable starts make the complete workflow explicit. Changed settings discard obsolete encodes; Continue editing uses the existing local workspace without uploads. Mix and save your current noise β
Added. A worked colored-noise guide connects the four implemented textures to independent gains, headroom, real sample readouts, timer limits and WAV versus MP3. A gain-flow illustration and accessible arithmetic table explain why quiet exports stay quiet. Primary technical sources and explicit limits distinguish digital samples from acoustic measurements or sleep/health claims; the guide remains readable without JavaScript. Read the mix-to-export guide β
v2026.10.08.5
Control every test tone and understand its limits
Fixed. Hearing Test & Tone Generator owns one audio source across generator, sweep and range checks. Replay and source changes replace old sound; Stop all audio, Escape, hiding or leaving the page stop it without automatic restart. Settings alone are silent, pending starts can be cancelled, Volume 0 mutes every mode and generator tones auto-stop after 30 seconds. Range tones honor the selected channel and digital gain cap. Results preserve submitted answers, setup and unknown untested tones rather than inferring an unmeasured cutoff or diagnosis. Open the controlled tone workspace β
Added. A worked tone-and-setup guide explains the actual controls, logarithmic sweep, digital gain versus acoustic levels, replay and stop behavior, scripted answer records and equipment limitations. Its nominal-sweep illustration is not an audiogram. Primary listening and technical references, explicit non-diagnostic limits and the full guide remain readable without JavaScript. Read the tone and setup guide β
v2026.10.08.4
Read real signals with four oscilloscope exercises
Added. Audio Oscilloscope now offers four locally generated learning signals: a clean sine, hard clipping, a known DC offset and exact stereo polarity cancellation. Selecting the next example leaves the current signal untouched; generating it loads a stopped, measurable frame without microphone permission or autoplay. Real timebase, channel, trigger, cursor, freeze and PNG controls work on the generated samples. Original file, microphone and stereo demo workflows remain available. Try the signal exercises β
Added. A worked oscilloscope guide connects each example to its actual controls and expected measurements. It explains period-to-frequency calculation, display zoom versus clipped samples, short-window RMS and DC limits, and why silent Mix can coexist with audible stereo playback. Each exercise link selects its matching reference without starting sound; the guide, mathematical illustration and technical sources remain readable without JavaScript. Read the worked guide β
v2026.10.08.3
Keep every sample at impulse-response control boundaries
Fixed. IR Convolver keeps the entire selected impulse response when its exact duration falls between the sliderβs 0.1-millisecond steps. Choosing the maximum by settings or keyboard End no longer silently drops final samples on 44.1 kHz outputs. Visible and accessible values describe the actual bounded setting; a changed control retires the previous export before a fresh complete-tail render. The processing algorithms, original audio, safety limits and free access are unchanged. Open IR Convolver β
v2026.10.08.2
Correct sample clocks for two-source audio effects
Fixed. Spectral Morph prepares a local Source B copy at Source Aβs sample rate before its chosen timeline alignment. Combining authored demos with decoded files now uses the correct frequency scale and physical duration in full renders and bounded 10-second auditions. WAV and MP3 export the current full result; preview-only audio remains non-exportable. Original source buffers and displayed formats stay unchanged. Open Spectral Morph β
Fixed. IR Convolver accepts valid source and impulse-response pairs with different decoded sample rates, preparing an IR copy in the source clock for trimming, pre-delay and complete-tail export. Its live graph prepares the IR at the actual audio-output rate, fixing 48 kHz demo playback on 44.1 kHz outputs. Conversion memory is checked before allocation; cancellation, changed inputs or controls and late conversion failures cannot revive obsolete audio. Existing DSP algorithms, defaults, safety budgets and free access are preserved. Open IR Convolver β
v2026.10.08.1
Keep six complex processors on your current inputs
Fixed. Vocoder, Multiband Compressor, Audio Exciter, Match EQ, Drum Replacer and Stereo Phase Aligner retire previous comparisons and exports when audio, settings or trigger plans change. Refused, failed and pending imports clear the replaced input; Match EQ and Drum Replacer preserve the other valid input. Late decoding, renders and MP3 callbacks cannot overwrite newer work. Current WAV and MP3 downloads can run together, and failed demos, previews and stereo estimates can be retried. Processing algorithms, preset sounds, original audio, safety limits and free access are unchanged. Open Vocoder β
Fixed. Match EQ prepares a reference copy at the target sample rate when the two inputs differ, so both fingerprints use the same frequency scale in bounded previews and full renders. Raw inputs and their metadata stay unchanged; conversion memory stays within existing limits and cancelled or obsolete conversions cannot replace current audio. Open Match EQ β
Fixed. Match EQ imported filenames and processing errors, and Drum Replacer custom one-shot labels, display as literal text rather than page markup. Replacing a custom sample also updates its displayed identity. Open Drum Replacer β
v2026.10.07.3
Keep twelve advanced effects on the current audio
Fixed. Phaser, Bitcrusher, De-reverb, Dynamic EQ, Formant Shifter, Guitar Amp Simulator, IR Convolver, Ring Modulator, Spectral Freeze, Vinyl Effect, Vocal Harmonizer and Spectral Morph retire obsolete comparison playback and exports when inputs, controls or presets change. Refused, failed and pending imports clear previous audio and source details; replacing one input in IR Convolver or Spectral Morph preserves the other valid input. Late decoding, renders and MP3 progress, failures or completion cannot overwrite newer work, and current WAV and MP3 downloads can run together. Demo and live-audition failures have a clean retry path. Audio algorithms, preset sounds, safety limits, native export formats and free access are unchanged. Open Spectral Morph β
v2026.10.07.2
Keep imported text safe across twelve advanced audio effects
Fixed. Phaser, Bitcrusher, De-reverb, Dynamic EQ, Formant Shifter, Guitar Amp Simulator, IR Convolver, Ring Modulator, Spectral Freeze, Vinyl Effect, Vocal Harmonizer and Spectral Morph display imported filenames and processing or export errors as literal text, not executable page markup. Both the source and impulse-response inputs in IR Convolver and both sources in Spectral Morph are covered. Accented characters, quotes and literal HTML entities stay readable. Audio algorithms, presets, source identity, native WAV/MP3 exports and free access are unchanged. Open IR Convolver β
v2026.10.07.1
Keep effect auditions and exports on the current audio
Fixed. Chorus, Flanger, Tremolo / Auto Pan and Distortion retire obsolete comparison playback when controls or presets change, so hidden players cannot keep playing an old render. Replacing audio clears old player buffers and source details even if the new file is refused or cannot decode. Late MP3 progress, failures and completion cannot alter a newer result or encoding job; current WAV and MP3 downloads can still run together. A failed live audition clears its obsolete audio while keeping an error and recovery path. Distortion also prevents an older render from clearing a newer result. Processing algorithms, presets, live-preview controls and free access are unchanged. Open Audio Distortion β
v2026.10.06.1
Keep filenames and error messages safe across four audio effects
Fixed. Chorus, Flanger, Tremolo / Auto Pan and Distortion now display selected filenames and decoder, processing and export errors as literal text, never page markup. This also protects temporary loading messages, file-size refusals and Distortion source details, while preserving accented characters, quotes and literal HTML entities. Audio processing, presets, source identity, WAV/MP3 exports and free access are unchanged. Open Audio Chorus β
v2026.10.05.1
Keep transient shaping and clipping repair on the current audio
Fixed. Transient Shaper clears the previous source, processed and difference playback, measurements and pending exports as soon as you choose a replacement, including files rejected before decoding. Late uploads, renders and synthetic drum demos cannot replace newer work, and changing shaping controls retires stale output and MP3 downloads. Fresh WAV, MP3 and Continue editing exports belong to the current render. Fast/slow envelope processing, presets and free access are unchanged. Open Transient Shaper β
Fixed. Audio Declipper retires the old recording, all three players, region reports and exports immediately when you select another file, even when the new file fails safety checks. Older decoding, bounded previews and full-file repair cannot overwrite newer work; settings changes cancel obsolete previews and pending MP3 exports. Fresh audio exports and JSON/CSV reports use the current repair. Flat-top detection, interpolation, safety ceiling and free access are unchanged. Open Audio Declipper β
v2026.10.04.1
Keep loudness measurements and exports on the current file and target
Fixed. Loudness Normalizer now discards cached output, playback and pending exports when you change the LUFS target or preset, so a new target cannot simply relabel the previous audio. Choosing a replacement immediately clears the old source, measurements and output, even if the new file fails memory preflight. Delayed decoding, measurements, normalization or bundled demo loading cannot overwrite newer work. Repeated requests share the current normalization; fresh WAV, MP3 and Continue editing exports use its captured source and target. Silence and incomplete measurements cannot be exported as normalized results. Loudness estimation, peak-protection processing and free access are unchanged. Open Loudness Normalizer β
v2026.10.03.1
Keep voice cleanup previews, reports and exports on the current recording
Fixed. Hum, Mouth Click, Plosive and Breath Remover now retire the previous recording, previews, events, reports and exports immediately when you choose a replacement, even if the new file is rejected by memory preflight. Late older uploads, previews or renders cannot replace the current source. Local voice demos supersede uploads still being checked or decoded, and choosing a source or changing settings cancels pending obsolete MP3 downloads and handoffs. Fresh WAV, MP3 and Continue editing exports stay with the current result. Breath Remover's six controls also retain their proper labels and value associations. Cleanup algorithms and free access are unchanged. Open Breath Remover β
v2026.10.02.1
Keep envelope edits, split parts and ZIP exports on the current file
Fixed. Audio Fade In/Out and Audio Splitter now retire the previous source, timeline, previews and exports when you choose a replacement, even if the new file fails memory preflight. An earlier decode, render or split plan cannot overwrite the latest selection. Pending envelope MP3 exports are cancelled when the source or points change; individual split-part exports can now be cancelled too, and replacing the source cancels unfinished part or ZIP downloads. Fresh WAV, MP3, local handoffs and ZIP manifests stay tied to the current result. Processing algorithms and free access are unchanged. Open Audio Splitter β
v2026.10.01.1
Keep dynamics processing and exports tied to your current audio
Fixed. Noise Gate, Audio Limiter and Dynamic Range Compressor now clear the previous audio, previews and measured receipt as soon as you select a replacement, even if it fails the memory safety check. A delayed earlier decode or render cannot replace the latest file. Changing the source or settings while MP3 preparation is pending cancels the obsolete download and local handoff; fresh WAV, MP3 and Continue editing exports use the current result. Processing algorithms and free access are unchanged. Open Dynamic Range Compressor β
v2026.09.30.1
Keep audio quality reports tied to the current file
Fixed. Audio File Analyzer and Stereo Phase Checker now retire the previous report and downloads the moment a replacement is selected, including when the new file fails the memory safety check. A delayed earlier decode cannot replace the latest result, and the stereo previews stop and clear with the old source. Valid files can still be scanned and exported as fresh JSON or CSV. Measurements and free access are unchanged. Open Audio File Analyzer β
v2026.09.28.5
Keep musical analysis tied to the current song
Fixed. Key Finder, BPM Detector and Chord Progression Detector now clear the previous song's result when you choose a new file or demo. A delayed earlier load cannot replace the newer analysis, and an invalid replacement cannot leave the old BPM or chord receipt available for export. New results and exported receipts remain tied to the current file. The detection algorithms and free access are unchanged. Open BPM Detector β
v2026.09.28.4
Keep three editors bound to the file you chose
Fixed. Audio Trimmer, Reverb Designer and Stereo to Mono now retire the previous preview and export as soon as you choose a different file. A slow earlier decode or demo cannot replace the newer selection; an invalid replacement leaves no old result available. Each tool can still render and download a fresh WAV from the current file. Audio processing settings and the free tools are unchanged. Open Audio Trimmer β
v2026.09.28.3
Keep Master EQ readable when a band has no signal
Fixed. Mastering Studio's EQ visualization no longer errors when the audio has measurable mid frequencies but no measurable high-frequency signal. It shows the mid-band change and marks the high band as No signal. The seven-module mastering chain, audio processing and export settings are unchanged. Open Mastering Studio β
v2026.09.28.2
Trust the file shown in Local Workspace
Fixed. Local Workspace now opens or forgets only the exact output shown on its card. If another tab replaces that file first, the card refreshes and asks you to review the new output instead of opening or deleting it unexpectedly. Expired or incomplete local records are cleaned without removing a newer saved file; handoffs saved before this update still work. Audio stays on your device and the 24-hour expiry is unchanged. Local storage details β
v2026.09.28.1
Keep Audio Converter bound to your latest file
Fixed. Audio Converter now clears the previous source and preview as soon as you choose a new file or demo, so a pending load cannot export the old audio. A delayed earlier demo cannot replace a newer upload, and an in-flight conversion for an older source is discarded without writing a stale status over the new workflow. Failed demo loads leave no old file available; after retry, the new file's full WAV or MP3 export remains explicit. Open Audio Converter β
v2026.09.27.3
Make Audio Compressor size targets dependable
Fixed. Audio Compressor now warns before encoding when a full file cannot meet a selected sharing cap at the available bitrates. For feasible targets it checks the actual encoded size, retries lower bitrates if needed, and downloads only a fitting result; an over-target file requires an explicit choice. A newer file or changed setting also cancels a stale export, and the measured receipt distinguishes encoded from downloaded files. Manual MP3 and WAV export remain available. Open Audio Compressor β
v2026.09.27.2
Make Volume Booster's sample ceiling measurable
Fixed. Volume Booster's compressor threshold alone could let an export peak far above the selected ceiling. An added sample-peak safety stage now applies the ceiling in live preview and before WAV or MP3 encoding; the output meter sits after it. The limiter-off bypass remains available. This is not a certified true-peak or MP3 peak guarantee, and strong boosting can still distort. Open Volume Booster β
v2026.09.27.1
Keep Volume Booster exports in sync with your settings
Fixed. If you change Volume Booster controls while an audio file is rendering, the pending export now refreshes with the current settings instead of caching the older result. Repeated downloads reuse only a render that matches the loaded file and controls; a failed render no longer downloads the previous audio. Live preview and microphone handling are unchanged. Open Volume Booster β
v2026.09.26.4
Resume Volume Booster preview where you paused
Fixed. Volume Booster's file preview now resumes from the paused position. Changing the limiter while a file plays keeps the current position instead of jumping to the start; selecting a new file still starts from zero. Microphone handling and audio processing settings are unchanged. Open Volume Booster β
v2026.09.26.3
Keep Visualizer microphone sessions under your control
Fixed. Music Visualizer now shows Cancel while microphone permission is pending. Repeated clicks open only one request; late grants close after Cancel, file playback or a new file selection, and leaving the page ends the active track. Existing video export and audio processing are unchanged. Open Music Visualizer β
v2026.09.26.2
Keep Volume Booster's microphone under your control
Fixed. Volume Booster now lets you cancel a pending microphone request. A late permission grant cannot take over after Stop or switching to file preview, and limiter changes reuse the active microphone instead of requesting it again. Leaving the page closes the track. Use headphones when monitoring live audio. Open Volume Booster β
v2026.09.26.1
Keep the tuner microphone under your control
Fixed. The Instrument Tuner now prevents overlapping microphone starts. Stop can cancel a pending permission request, late grants are closed, and leaving the page ends an active microphone track. Pitch detection is unchanged. Open Instrument Tuner β
v2026.09.25.13
Keep the microphone meter demo on real time
Fixed. The local Microphone Level Meter demo now advances its level history and two-second peak hold at real-time speed, instead of running faster on high-refresh screens. Live microphone measurements are unchanged. Open Microphone Level Meter β
v2026.09.25.12
Keep Mastering A/B controls inside small screens
Fixed. The Source, Mastered and Difference switch now fits a 320px phone viewport across browser font metrics. Its tap targets remain at least 44px high. This completes the mobile visualizer update without changing audio processing. Open Mastering Studio β
v2026.09.25.11
Make Mastering Studio visualizers readable on phones
Improved. Mastering Studio charts now draw at their actual phone width instead of shrinking a wider canvas. Stereo Width, Glue Compressor and Maximizer stack their measurements on narrow screens; analyzer controls, readouts and explanations are easier to read. Audio processing is unchanged. Open Mastering Studio β
v2026.09.25.10
Refresh cached Whisper workers after the security fix
Fixed. Speech to Text now requests a fresh entry script and Whisper worker after the WebAssembly security-policy fix. This avoids reusing an earlier browser-cached worker response that still blocks transcription, while keeping the model and audio processing unchanged. Open Speech to Text β
v2026.09.25.9
Restore local Whisper transcription in production
Fixed. Speech to Text's Whisper worker can now compile the WebAssembly needed by the on-device model. The production security policy had blocked real transcription even though the earlier synthetic worker test passed. Open Speech to Text β
Improved. The WebAssembly permission applies only to the dedicated Whisper worker. The page and other tools keep their existing security policy; JavaScript eval remains blocked. A browser regression now tests this boundary.
v2026.09.25.8
Stop unfinished Whisper work when you switch modes
Fixed. Speech to Text now runs Whisper model setup and transcription in a dedicated browser worker. Switching to Live or leaving the page stops unfinished work instead of leaving a hidden task running; a fresh attempt remains available. Open Speech to Text β
Improved. Whisper still requires explicit model-download consent and processes the audio locally. Existing transcripts and TXT/SRT export are unchanged. A cancelled request may already have downloaded some model bytes.
v2026.09.25.7
Keep Match EQ with the right source
Fixed. Selecting a new main track now cancels an older in-progress Match EQ reference immediately, including while a bundled demo is still loading. Its result cannot change the EQ for the wrong source. Open Mastering Studio β
Fixed. An interrupted reference returns to a clear cancelled state instead of appearing accepted or failed. Older main-track loads are also invalidated as soon as a newer track is selected; audio processing algorithms are unchanged.
v2026.09.25.6
Last selected mastering track wins
Fixed. Rapidly choosing two main tracks or Match EQ references no longer lets the slower, older file replace the last selection or its displayed filename. A bundled demo loading in the background also yields to a newer upload. Open Mastering Studio β
Fixed. Mastering Studio commits a new track only after decoding and loudness measurement finish. An obsolete load cannot leave the timeline and source measurements out of sync; mastering audio algorithms are unchanged.
v2026.09.25.5
Master the track you actually loaded
Fixed. Mastering Studio now discards an older Stem Rebalance result if you load another track, reset preprocessing or change its controls while it runs. The new source cannot be replaced by the previous track's processed audio. Open Mastering Studio β
Fixed. A render or MP3 export from older settings is no longer accepted after you change the source, range or mastering chain. Render again to export the current track; codec preview also stops when those settings change. Audio processing itself is unchanged.
v2026.09.25.4
One mastering audition at a time
Fixed. Mastering Studio's MP3 codec preview now has a Stop action. Starting the main audition, loading another track or leaving the page ends that preview instead of letting two audio sources play together. Open Mastering Studio β
Fixed. If MP3 decoding finishes after you start normal playback, the cancelled codec preview stays silent. Mastering processing and exports are unchanged.
v2026.09.25.3
Mastering measurements that agree
Fixed. Mastering Studio's live measurement cards, descriptions and accessible canvas summaries now use the same captured audio frame. Values no longer disagree while the signal changes. Open Mastering Studio β
Improved. Freeze stops repainting the held module display, and reduced-motion mode refreshes the module and audition spectra at a gentler pace. Playback and audio processing are unchanged.
v2026.09.25.2
Beat Maker steps that work on phones
Improved. The mobile Beat Maker pattern now has 44-pixel pads, persistent voice names, and four direct step groups, so all 16 steps can be reached without guessing how to scroll. Open Beat Maker β
Improved. Pattern pads now announce on/off state to assistive technology, including after a preset or Clear. The desktop grid and synthesized sound are unchanged.
v2026.09.25.1
A quieter Audio Oscilloscope display
Fixed. Freezing the Audio Oscilloscope display now stops its animation loop while audio continues. Unfreeze resumes the live trace; a hidden tab pauses display updates and refreshes when visible again. Open Audio Oscilloscope β
Improved. Changing your system's reduced-motion setting during a session now takes effect immediately and refreshes the live display at a gentler pace. Signal measurements and PNG snapshots are unchanged.
v2026.09.24.5
Equalizer spectrum follows your playback
Fixed. The Equalizer spectrum now stops redrawing when playback is paused, finished, or in a hidden tab. The last frame remains visible, and playback wakes the live display again. Open Equalizer β
Improved. With reduced motion enabled, the live spectrum refreshes at a gentler pace. Equalizer sound processing and export are unchanged.
v2026.09.24.4
Clearer active controls in Mastering Studio
Fixed. Active Loop and Hold for Source controls in Mastering Studio now use a deeper red so their white labels remain readable. Audio processing and export behavior are unchanged. Open Mastering Studio β
v2026.09.24.3
A calmer, lighter Music Visualizer
Improved. Music Visualizer no longer redraws the stage while idle. Playback and microphone input wake it again; a hidden tab pauses the preview unless you are recording a video. Open Music Visualizer β
Improved. With the system's reduced-motion setting, the live preview updates at a gentler pace while keeping audio feedback. Starting a video recording restores normal animation for the exported clip.
v2026.09.24.2
Speech to Text stops a file before a hidden model download
Fixed. Switching from Whisper to Live while an audio file is decoding now discards that import before the Whisper runtime or model can be requested. A finished result from an already-started transcription is not added after the switch. Open Speech to Text β
Improved. Returning to Whisper shows what happened to the file. If a model download or transcription was already running when you switched, it may continue in the tab; existing transcript and choices stay intact.
v2026.09.24.1
Speech to Text stops listening when you switch engines
Fixed. Switching Speech to Text engines now ends the microphone take or system speech session you are leaving. A microphone permission granted after the switch is released before a hidden recording can begin. Open Speech to Text β
Improved. An unfinished take shows a cancellation message when you return to Whisper mode. The recorder is ready for another try, while your existing transcript and privacy choices remain.
v2026.09.21.2
One microphone start at a time
Fixed. The ten tools using the shared microphone recorder now lock their recording button while microphone access is starting. Repeated clicks cannot start overlapping recordings while the browser waits for permission, and a denied request restores the button for another try. Open Voice Recorder β
Improved. A Starting microphone⦠label makes the waiting state visible. Voice Recorder's Record again button shares the same guard; Speech to Text keeps it in place when a parallel file import finishes.
v2026.09.21.1
Safer microphone cleanup and recording retries
Fixed. Tools using the shared microphone recorder release their microphone stream and level-meter resources if recording fails, finishes or the page is left. Repeated Stop calls now share one result, and a source that ends on its own can still finish decoding. Open Voice Recorder β
Fixed. Auto-Tune, Voice Changer, Voice Cloning, Speech to Song and Voice to Instrument restore their recording controls after a failed take, so you can start another recording.
v2026.09.20.5
More reliable Visualizer recordings
Fixed. Music Visualizer releases each recording's capture tracks after saving, leaving the page or a recording error, while keeping shared audio available for playback and the next take. Open Music Visualizer β
Improved. A finishing state prevents repeated stop clicks. Failed or empty recordings show a retry message instead of downloading an empty file, and the filename extension follows the browser's recording format.
v2026.09.20.4
Better wrapping across system fonts
Fixed. Long technical terms on the About page now wrap when text is enlarged, including on systems whose default fonts use more horizontal space.
v2026.09.20.3
Easier reading with enlarged text
Fixed. Contact email addresses wrap on narrow screens instead of running outside the page. The complete address and email link are preserved.
Improved. Business pages make room for enlarged text in headings, summaries and the inquiry form without widening the page.
Fixed. In-page links to release notes, legal information and business sections leave room for the taller navigation header when text is enlarged.
v2026.09.20.2
More reliable mobile navigation
Fixed. The shared header keeps its search and menu buttons at full touch size on narrow screens, and makes room for enlarged text.
Fixed. An open mobile menu now closes and releases page scrolling when the window switches to the desktop layout. Keyboard focus moves to a visible control when needed.
v2026.09.20.1
A clearer start, safer filenames, and release notes
Added. A website version in the footer and this release history, so you can see what changed with each update.
Improved. Auto-Tune now starts with a clear choice: open your vocal or try the demo. The guidance is specific to vocals, and a failed demo download can be retried. Try Auto-Tune β
Fixed. Filenames containing special characters are displayed as text across 32 audio tools, including Mastering Studio reference imports. Your original filenames are preserved for processing and export.
Operator & business details.
AudioWrench is published and operated by the Polish sole proprietorship below. These details identify the legal entity responsible for the service and its privacy notice.
Full registered name
Mateusz Konieczny - BALANCE IT
Registered address
ul. ΕwiΔty Marcin 29 lok. 8 61-806 PoznaΕ, Poland
For custom audio workflows, private deployments and clearly labeled partnerships, see AudioWrench for Business.
For the rules governing use of the service, read the Terms of service. For information about visitor data, local browser storage and optional third-party resources, read the Privacy & cookies notice.
How AudioWrench works.
Most AudioWrench tools process audio directly in the browser. Tools that need optional model files explain the download before it starts and let you decide whether to continue.
Each tool page explains what the workflow does, what it exports and where its practical limits are. Reproducible technical tests are collected separately in the benchmark and evidence hub.