# Cut, merge, and retime audio without avoidable quality loss

You cannot make a compressed audio file better by editing it again. What you can do is avoid extra damage: start from the best source you have, make all cuts and timing changes in one pass, choose a sensible output format, and verify the result before deleting the original.

Lutrakit's free [Audio Cutter](/audio-cutter/), [Audio Merge](/audio-merge/), and [Audio Speed](/audio-speed/) run in the browser with no account required. The selected audio bytes stay on the device. The tools use FFmpeg-style decoding, filtering, and encoding inside the browser session, so the quality rule is the same as in desktop audio editing: filtering decoded audio and exporting to MP3 or Ogg means a new lossy encode. Exporting to WAV avoids a lossy codec stage, but the file is much larger.

## The simple rule: edit once from the best source

FFmpeg's own documentation separates stream copy from transcoding: stream copy is fast and has no quality loss because it does not decode or re-encode, while transcoding decodes and encodes again and lossy encoders can degrade quality ([FFmpeg documentation](https://ffmpeg.org/ffmpeg.html), accessed 2026-08-17). Lutrakit's audio tools apply filters such as trim, fade, crossfade, concat, and tempo. Filters work on decoded audio frames, so plan the session as an export step, not as a reversible mastering system.

Use this decision table:

| Task | Best source | Safer output choice | Why |
| --- | --- | --- | --- |
| Cut a ringtone or excerpt | Original WAV/FLAC if available; otherwise the highest-quality copy | Same delivery format if final, WAV if you will edit again | Avoid repeated MP3/Ogg encodes. |
| Merge clips with fades | Clean source clips at similar loudness | WAV for a production intermediate, MP3/Ogg for final sharing | Crossfades require filtering and re-encoding. |
| Speed up speech | Original recording | Same format for final listening, WAV for further work | Tempo change creates new audio samples. |

Do not run "cut, then merge, then speed, then convert" as separate downloads unless you need to. Each lossy export can compound artifacts.

## Cutting: choose boundaries and fades deliberately

[Audio Cutter](/audio-cutter/) accepts common audio files up to 300 MB. Its default selection is 0:00 to 0:30 in "keep" mode, with optional fade-in and fade-out. It can also remove a selected middle section and keep the surrounding audio.

For speech, leave a small breath before and after the selected sentence. Cutting exactly on a plosive or a loud transient can create a click. A very short fade, even 30-80 ms, is often enough to hide the edit. For music loops, listen to the beat and the reverb tail; a cut that looks exact on the timer can sound late because the musical phrase resolves after the waveform peak.

Lutrakit exposes a ringtone preset that keeps the first 30 seconds and adds a 2-second fade-out. Treat that as a convenience, not as a universal format rule. Phones, platforms, and social apps have their own limits.

## Merging: hard joins, crossfades, and loudness

[Audio Merge](/audio-merge/) takes 2 to 32 files, with a 200 MB per-file limit and a 300 MB combined input limit before any lower device-memory cap. It can output MP3, WAV, or Ogg; the current defaults are `merged.mp3`, no crossfade, no fade-in, no fade-out. Crossfade and fade durations are limited to 0.1-10 seconds.

FFmpeg's `concat` filter joins streams one after another, while `acrossfade` overlaps the end of one stream with the start of the next for a set duration ([FFmpeg filters: concat and acrossfade](https://ffmpeg.org/ffmpeg-filters.html), accessed 2026-08-17). Use a hard join when the clips already have clean starts and ends, such as numbered lecture parts. Use a crossfade when room tone, applause, or music beds would make the join obvious.

Before merging, put files in the final order and rename them if the order might be confusing. After export, listen to each join. A waveform that looks smooth can still contain a volume jump or a word cut in half.

## Retiming: speed without pitch shift

[Audio Speed](/audio-speed/) changes playback speed from 0.25x to 4x and keeps pitch by using FFmpeg's `atempo` filter chain. FFmpeg documents `atempo` as a tempo filter and notes that large tempo factors can be chained to avoid the sample-skipping behavior of factors above 2 ([FFmpeg filters: atempo](https://ffmpeg.org/ffmpeg-filters.html), accessed 2026-08-17). Lutrakit chains factors internally for speeds outside the simple 0.5-2 range.

For speech, 1.25x or 1.5x usually preserves intelligibility better than jumping straight to 2x. For music, tempo changes are more destructive because rhythm, transients, and room reverb all change. If the output sounds metallic or smeared, use a gentler factor or a dedicated audio workstation.

## Worked example

Use a redistributable sample set:

- `voice-master.wav`: 30 seconds of spoken text owned by Lutrakit.
- `intro.wav`: 3 seconds of generated tone or original music sting.
- `outro.wav`: 3 seconds of generated tone or original music sting.

Workflow:

1. Cut `voice-master.wav` from 00:04.000 to 00:22.000 with 0.05 s fade-in and 0.08 s fade-out.
2. Merge `intro.wav`, the cut voice clip, and `outro.wav` with a 1.5 s crossfade, output WAV.
3. Speed the merged file to 1.25x for a review copy.

Expected checks, to be recorded after a real run:

| Check | Expected result |
| --- | --- |
| Duration after cut | About 18 seconds, allowing encoder/container rounding. |
| Join audit | No click at either crossfade; no clipped first word. |
| Speed copy | Duration is about original merged duration divided by 1.25; pitch remains recognizably the same speaker. |
| File privacy | Browser network log shows no upload of selected audio bytes. |

For a reproducible check, record the exact files, browser, OS, Lutrakit revision, and observed outputs.

## Browser limits are quality limits too

Compressed input size is not the same as editing memory. FFmpeg decoders turn compressed packets into raw frames, and Web Audio's `AudioBuffer` model represents memory-resident 32-bit floating-point PCM for short audio assets ([Web Audio API editor's draft](https://webaudio.github.io/web-audio-api/), accessed 2026-08-17). A tiny MP3 can expand into much more decoded audio. That is why Lutrakit keeps explicit file-size and combined-size limits, and why very long projects belong in a desktop editor.

If the browser tab runs out of memory, reduce the number of files, use shorter clips, export an intermediate WAV, or move to desktop software. Do not keep retrying a large merge with more tabs open; memory pressure can make the browser slower and less predictable.

## Verification checklist

Before sharing the result:

- Keep the original files unchanged.
- Listen to the first 3 seconds, every join, and the final 3 seconds.
- Check the expected duration against the visible timeline.
- Compare the voice pitch before and after speed changes.
- If exporting MP3/Ogg, use it as the final delivery copy rather than a new source for more edits.
- If the output is important, keep a WAV or FLAC master outside the browser session.