Narration peaks: choose compression or limiting deliberately
Use compression when the level difference between ordinary narration and louder passages needs controlled reduction. Use limiting when occasional peaks need a ceiling. Neither operation reconstructs speech already clipped in a recording, and changing a peak ceiling alone does not establish a delivery loudness target.
Open the relevant downloader · Platform tools and access limits
Inspect the recording before choosing an effect
Keep an untouched recording and select three listening regions: a normal sentence, the loudest phrase and a pause containing breaths or room noise. Listen at a fixed playback volume. If the loud region is already harsh in the original, inspect it for overload before applying another effect. Lowering damaged samples can reduce their level without undoing their distortion.
Separate large differences between takes from peaks within one take. An editorial gain adjustment may make two takes comparable without flattening every syllable. A compressor reacts to an input level and changes gain over time; threshold, ratio, attack and release interact. Start conservatively in an audio editor and compare the processed region with bypass at a similar perceived level.
Review the ceiling and the voice together
A limiter can restrain short peaks, but heavy limiting may make consonants rough, breaths conspicuous or the voice tiring. Do not keep lowering the ceiling to solve a background-noise problem. Check whether automatic makeup or output gain raises the result after gain reduction. An acceptable meter reading is insufficient if the narration sounds worse.
Peak protection and loudness delivery are separate checks. If a recipient specifies integrated loudness and true peak, measure those quantities on the final encoded file using a suitable meter. A sample-peak limiter is not automatically a true-peak guarantee. Re-encoding can change the measured result, so do not certify a compressed derivative from measurements of the uncompressed intermediate alone.
Verify timing after lookahead processing
Some limiters use lookahead and therefore introduce processing latency. An editor may compensate for that delay inside its timeline; a standalone render chain needs its own verification. FFmpeg documents a latency option for its alimiter filter. Check the options in the installed version instead of assuming that every limiter or host compensates identically.
Export a short test with a visible clap, lip closure or other clear synchronization cue near both ends. Compare the processed audio against the unchanged picture and inspect the beginning and tail for lost material or added silence. If timing changed, establish whether the cause is uncompensated processing delay or an existing timestamp offset before applying a correction. The website's MP3 extraction utility does not provide a compressor, limiter or loudness certification service.
Worked example
Example: a narrator's ordinary sentences are balanced, but one laugh reaches an undesirable peak. Compare a restrained limiter on that moment with whole-track compression, then recheck the first clap after export. If the laugh is distorted in the original recording, retain that evidence and request another take rather than labelling a quieter damaged result repaired.
Related guidance
Sources and verification
Reviewed 2026-10-12. Product behavior is limited to available source formats and provider access; it is not a promise that every public link works.