Identify and review hesitation sounds ('um', 'uh', 'er', 'ah', 'like', 'you know') and stutter repetitions with timestamped word tokens. Selectively audition each cut and preserve natural 0.08s–0.32s speech cushions without robotic dialogue cuts.
Ready for a local audio or video file
Reading media and computing RMS energy…
Balanced (~-38dB): Calibrated natural balance recommended for podcasts, YouTube videos, and dialogue.
Automatically removes silence before first speech and after last speech.
00:14.2 → 00:16.8 (2.6s gap)
1.6s quiet gap; keeps 0.32s breathing room.
0 silences found (0.0s dead air) · 0 will be cut, 0 kept
Removes 80Hz rumble, 50/60Hz hum, room hiss & mouth clicks
Fills shortened silence gaps with authentic room tone synthesized from your recording, eliminating unnatural "vacuum dropouts".
Protect Intentional Sentences & Pauses
Importing your subtitle transcript locks dialogue boundaries so dramatic reflections and deliberate speech pauses inside sentences are never auto-cut.
Removes verbal hesitation sounds (um, uh, er, like) and repeated stutters.
Total Proposals
0
Active Cuts
0
Restored Pauses
0
Protected Lines
0
Structural Export Validation:
Verified Export Options:
Multi-band spectral Wiener filtering & room rumble removal
Lossless 16-Bit Master Audio & Edit Ledger Export
Rendered locally in browser memory with 8ms anti-click micro-fades. Zero server upload.
Timeline files reference your media by file name only. Final Cut Pro, DaVinci Resolve or Premiere Pro will ask you to relink the clip to the original file.
100% browser-local processing · No files uploaded to any server
How this tool decodes, analyzes, and masters your dialogue locally with zero cloud latency.
Import subtitle cues (.vtt / .srt) or simulate local speech alignment. The engine maps spoken cue timestamps into individual word tokens with sub-second precision to isolate hesitation crutches and stutter loops.
Review flagged word tokens in the interactive queue. Use 'Hear Before' and 'Preview Cut' to audition each removal in context, retaining calibrated 0.08s–0.32s speech cushions so natural conversational phrasing remains intact.
Render spliced audio locally via OfflineAudioContext with 8ms anti-click micro-crossfades. Export updated millisecond-precise WebVTT/SRT subtitles, Audacity labels, or uncompressed 16-bit PCM WAV masters.
Explore companion utilities for dialogue cleanup, speech pacing, and mastering.
Generate automated episode summaries, key takeaways, and timestamped chapters from your transcript.
Smoothly fix saliva clicks, lip smacks, and transient pops without muffling speech clarity.
Shorten awkward quiet gaps and gasps while retaining calibrated 0.32s breathing cushions.
Explore the acoustic principles, ITU-R loudness standards, and pacing techniques behind this tool.
User-controlled review: differentiating linguistic hesitation markers from contextual connectors, with cushion preservation at cut boundaries.
An empathetic guide for editors: respecting authentic speech traits, avoiding synthetic smoothing of neurodivergent speech, and editing with consent.
How text-audio alignment works: importing SRT/VTT files, protecting enclosed pauses within sentences, and exporting synchronized subtitles.
Common Questions & Technical Clarifications
AutoSilenceCut tokenizes imported WebVTT or SRT subtitle cues (or local speech interval estimates) into timestamped word tokens. It matches token text against comprehensive multilingual hesitation patterns ('um', 'uh', 'umm', 'er', 'ah', 'like', 'you know', 'basically', 'literally') and stutter repetitions (e.g. 'the the' or 'w-what'), pinpointing exact timecodes on your waveform.
Yes. Rather than performing blind, destructive bulk deletion, AutoSilenceCut itemizes every flagged filler candidate in an interactive inspection queue. You can audition the interval ('Hear Before' / 'Preview Cut') to hear how the phrase sounds before and after the cut, toggling individual items with a single click or keyboard shortcut.
Removing a filler word abruptly right at the vocal boundary can cut off subtle consonant decay or create jarring pacing. AutoSilenceCut applies calibrated 0.08s to 0.32s speech cushions around splice points and blends segments using 8ms equal-power linear micro-crossfades to eliminate clicks and preserve speech flow.
When you import a subtitle file, AutoSilenceCut maps sentence boundaries across your audio. Quiet pauses that fall within intentional spoken lines are automatically tagged as 'Protected dialogue sentence' and restored, ensuring dramatic pauses and expressive reflections are never stripped by mistake.
You can import standard .vtt (WebVTT) and .srt (SubRip) subtitle files alongside your audio or video file. When exporting, you can download uncompressed 16-bit PCM WAV masters, updated timestamp-synchronized WebVTT/SRT subtitle files, Audacity label tracks (.txt), CSV edit ledgers, or EDL/FCPXML cut lists.
Never. All subtitle parsing, word tokenization, acoustic RMS scanning, and audio rendering execute 100% locally inside your browser's Web Audio engine. Zero media bytes, transcripts, or filenames are ever sent to any remote server or cloud API.
