WAV vs MP3 vs OGG in one page
WAV is raw PCM — perfect quality, huge files (≈10 MB/min stereo). MP3 shrinks ~10× by discarding what ears mask; Opus (WebM/OGG) beats MP3 at low bitrates and rules voice calls.
Archive in WAV/FLAC, distribute in Opus or MP3 192k+, edit from the lossless original — every lossy re-encode stacks damage (generation loss).
Mistakes: converting MP3→MP3 'to make it smaller' (damage compounds), sending 32-bit float WAVs to players expecting 16-bit, and assuming .ogg always means Vorbis (often Opus now).
Example: 3-min song: WAV 32 MB → MP3 192k ≈ 4 MB → Opus 96k ≈ 2 MB, transparent on most earbuds.
Try: WAV Converter · Opus/WebM Export
Normalize without clipping
Normalize to a target peak (e.g. −1 dB) so the loudest moment nearly touches full scale — matched loudness across files without touching dynamics. It turns the volume knob once, for the whole file.
Peak-normalize for consistency, then check the noise report: high DC offset or clipping % means fix the recording (gain staging, mic placement), not the file.
Mistakes: normalizing each track of an album separately (destroys intentional balance), normalizing a clipped file louder (distortion scales too), and confusing peak with loudness (LUFS).
Example: podcast takes peaking at −12 dB → normalize to −1 dB = +11 dB gain everywhere. Same dynamics, audible everywhere.
Try: Volume Normalizer · Noise Report
Fades that don't click
Fades that don't click ramp amplitude to exactly zero at the boundary — clicks are discontinuities, a waveform jumping mid-value to silence in one sample.
Use 5–50 ms fades on every cut boundary (edits), 1–3 s musical fades for intros/outros. Exponential curves sound natural for endings; linear is fine for edit cleanup.
Mistakes: fading only one side of a splice (the other edge still clicks), 0 ms 'fades', and long crossfades that smear transients (drums turn to mush).
Example: trim 12.40–48.00 s, then 10 ms fade-in + 300 ms exponential fade-out → broadcast-clean segment, zero clicks, waveform proves it.
Try: Fade In / Fade Out · Audio Trimmer / Cutter
MP3 without uploading anything
MP3 export runs lamejs (pinned 1.2.1, about 152KB, loaded once) entirely on-device: decode with Web Audio, encode CBR 128/192/320k, download. Your audio never leaves the tab.
It matters because every free MP3 converter is an upload queue with limits: here 2 seconds become 47.8 KB of real MP3 (frame sync verified) in seconds.
Mistakes: re-encoding MP3 to MP3 to shrink it (damage stacks — keep the WAV master), expecting 320k to rescue a bad recording, and exotic sample rates (auto-resampled to 44.1k with a note).
Example: tone-2s.wav → MP3 192k → tone-2s-192k.mp3 at 47.8 KB. Check the header bytes FF FB — genuine MPEG audio.
Try: MP3 Export · WAV Converter
Join podcast takes
Joining concatenates decoded PCM: one context guarantees a single sample rate, channels are matched (mono upmixed), capped at 20 minutes and 20 files, exported as WAV with a waveform.
It matters for podcasts, audiobooks, and voice notes: assemble takes locally instead of uploading raw recordings to an editor.
Mistakes: joining takes at different loudness without normalizing first (normalize, then join), clipping at seams (leave 0.2s handles), and 20 scattered files when 3 ordered ones edit faster.
Example: two 2-second tones become one 4.0s stereo WAV. Duration math is exact: it sums, never guesses.
Try: Join Audios · Audio Trimmer / Cutter