Hear the difference
Listen to the source and result. Sample peak and RMS measurements explain amplitude; FFmpeg reports integrated loudness separately.
Your story deserves
a clearer listen.
Trim a recording, compare its waveform and normalize loudness with real FFmpeg. Make a useful audio export before you plan the episode, short clip or creative handoff.
Useful audio work.
One honest workflow.
Make an actual file
Export lossless PCM WAV or a smaller MP3. The processor trims the selected segment and applies the requested loudness target.
Keep your recording local
Files stay in browser memory. The processing engine downloads from this site when needed; recordings are not uploaded.
Your local audio workbench.
This tool normalizes loudness and trims time. It does not remove room noise, restore clipping, transcribe speech or identify speakers. Listen before and after every edit.
Lavender: original signal. Lime: processed signal. Each waveform spans its own duration; after trimming, positions do not represent identical source timestamps.
The −16 LUFS starting value is editable, not a universal platform requirement. Check the destination's delivery specifications. True-peak target: −1.5 dBTP; loudness-range target: 11 LU.
Load audio or generate the demo. No recording is uploaded.
Original
No audio loaded.
Processed
No processed result.
FFmpeg loudness report
These values describe the selected segment at the normalization filter, before output encoding, not necessarily the complete original recording or final encoded-file loudness. Sample RMS above is not LUFS. The browser may resample decoded audio to the device audio-context rate.
Run processing to see measured input/output LUFS and true peak.
Processing log
Compatibility: browser audio decoding varies by codec. WAV or MP3 is a practical starting point. Video must contain an audio track. Large recordings require substantial memory; use a shorter excerpt when practical. The single-thread FFmpeg engine is about 31 MB and can process more slowly than playback.
Cleaner levels.
A more intentional episode.
Separate an audio problem from a story problem
A quiet recording needs level management; an unclear argument needs editing. Write the listener's question first. Preserve the answer, the evidence and enough context to understand who is speaking. Normalization cannot make an unfocused episode useful.
Trim a complete thought
Listen to the intended excerpt. Leave enough space around consonants and breaths so the edit feels natural. A waveform shows amplitude, not meaning, so do not infer sentence boundaries from a visual gap alone.
Compare at a consistent playback volume
Keep device volume steady when switching between original and result. Check voice peaks, pauses and background noise. Raising the average level can also make room noise more noticeable; this tool does not selectively remove that noise.
Choose the format for the next step
WAV keeps a PCM working copy for further editing. MP3 is smaller and practical for sharing, but uses lossy encoding. Keep the source; repeated lossy exports can compound degradation. This tool exports at 48 kHz in either format.
Repurpose without losing the point
Choose a standalone answer instead of the loudest or shortest moment. Write a title that accurately describes the excerpt, preserve attribution and supply any context the listener would otherwise miss. Review captions separately if you create a video version.
Pick the right next action.
Uneven levels
Use the loudness report, listen to the result and inspect the waveform. Dynamic loudness normalization may change the relationship between quieter and louder moments.
Noise or clipping
Return to the source recording and a suitable restoration editor. Normalizing volume cannot recover clipped waveform information or isolate every noise source.
A short excerpt
Make sure the segment contains a complete idea. Trim with explicit seconds, listen again and export the format needed by your publishing workflow.
What the measurements mean.
Sample peak
The largest absolute decoded sample, expressed in dBFS. It is different from true peak, which accounts for the reconstructed signal between samples.
RMS
The root mean square of decoded samples, expressed in dBFS. It measures average signal amplitude. It is not an integrated perceptual-loudness measurement and must not be labelled LUFS.
Integrated loudness
FFmpeg's loudnorm filter reports input and output integrated loudness in LUFS for the selected segment. The target can differ from the measured output, especially for short or unusual content. Read the report and listen rather than assuming an exact match.
What happens to my file?
The browser decodes a copy for visualization and puts another copy in FFmpeg's in-memory filesystem for processing. No media is sent to a server. Reloading closes the tool's in-memory session; keep exported files you want to retain.
Can I cancel?
Yes. Cancel terminates the processing worker. The source remains available, and the next processing request reloads the engine.
Sources you can inspect.
Written October 5, 2026. No AI transcription, customer testimonials or third-party performance claims are presented here.
Audio processing
FFmpeg loudnorm filter describes targets and normalization behavior. ffmpeg.wasm installation explains the browser worker architecture.
Waveform and amplitude
MDN decodeAudioData documents browser decoding. MDN AudioBuffer.getChannelData explains access to decoded PCM samples.