Does Editing Audio Lose Quality? What Re-Encodes and What Doesn't
Only edits that decode and re-encode the audio cost you anything. Cutting, looping, reversing and speed changes all re-encode. Muting a video and extracting a soundtrack to WAV do not. One extra generation is usually inaudible; the mistake that actually hurts is stacking several, or ending a chain on a higher bitrate than the audio deserves.
Key takeaways
- A cut segment comes out of the Audio Splitter at LAME quality level 2, a variable bitrate averaging about 190 kbps — whatever went in.
- A loop comes out of the MP3 Looper at a fixed 128 kbps constant bitrate. Two tools on the same site, one full quality step apart.
- Doing those two steps in the wrong order gives you a file 48% larger that sounds identical, because the 128 kbps stage caps the sound either way.
- Remove Audio from Video copies the video stream instead of re-encoding it, so the picture is bit-for-bit identical to the original.
- The MP3 Reverser stops at 10,240,000 bytes: 10 min 40 s at 128 kbps, 4 min 16 s at 320 kbps, 58 seconds as CD-quality WAV.
- CD-quality WAV runs at 1,411 kbps, about 10.6 MB per minute — 11 times an MP3 at 128 kbps, and the reason browser tools cap what they accept.
On this page
- What does a lossy generation actually cost?
- Which browser audio edits re-encode?
- What bitrate does each tool write out?
- How much audio fits inside a browser tool?
- Worked example: the same edits in two orders
- Getting audio out of a video without a second loss
- The edits that touch no audio file at all
- Four rules that cover most workflows
- Terms worth knowing
- All 12 tools in this guide
What does a lossy generation actually cost?
A lossy format such as MP3 or AAC does not store your recording. It stores a description of it, built by throwing away the parts a model of human hearing says you are least likely to notice. Decode that description and you get audio back, but not the audio you started with — you get the encoder's reconstruction.
Encode that reconstruction again and the second encoder works on already-damaged material. It cannot tell which artefacts came from the first pass, so it spends bits preserving them and discards a fresh layer of real detail. That is a generation, and the damage compounds. It is the audio version of photographing a photograph.
The practical news is better than that sounds. One extra generation at a sensible bitrate is inaudible to most people on most equipment. Three or four generations at 128 kbps is not: cymbals turn to hiss, reverb tails smear, and voices pick up a watery edge. So the goal is not to avoid re-encoding, which is often impossible in a browser. The goal is to know which steps cost a generation and spend as few as the job needs.
Which browser audio edits re-encode?
This is the table to keep. It is not a general statement about audio editing — it is what these specific tools do, read off their own processing settings, so you can check any row yourself by inspecting the file that comes out.
| Job | Tool | What happens to the data | Adds a generation? |
|---|---|---|---|
| Silence a clip | Remove Audio from Video | Video stream copied, audio track dropped | No — picture identical |
| Pull a soundtrack out, for editing | Audio Extractor, WAV output | Decoded to plain samples | No new lossy stage |
| Pull a soundtrack out, to keep | Audio Extractor, MP3 or AAC | Re-encoded at 192 kbps or 128 kbps | Yes |
| Cut a segment | Audio Splitter | Re-encoded, LAME quality level 2 | Yes, at about 190 kbps |
| Build a repeating loop | MP3 Looper | Decoded, then re-encoded at a fixed rate | Yes, at 128 kbps |
| Reverse a track | MP3 Reverser | Decoded, samples reversed, re-encoded | Yes |
| Speed a clip up or down | Adjust Video Speed | Timeline rewritten, output re-encoded | Yes |
| Swap the soundtrack | Add or Replace Audio in Video | New audio encoded into the container | Yes, for the new audio |
| Transpose a chord chart | Music Key Transposer | Text in, text out | No audio involved |
| Measure a tempo | Music BPM Tapper | Times your taps, reads no file | No audio involved |
| Produce a reference pitch | Tone Generator | Synthesised live in the browser | No source to lose |
Two rows deserve a second look. Remove Audio from Video is the only tool here that is genuinely free: it hands the untouched video stream straight to the output container, so a two-hour recording is processed in seconds and the picture is byte-identical. And the Audio Extractor appears twice because the answer depends entirely on the format you pick in its dropdown — the same button is lossless or lossy depending on one choice.
Audio Splitter → The anchor tool for this guide. Set start and end points to the millisecond in HH:MM:SS,mmm format, preview the segment before you commit, and download it as MP3. Everything runs through FFmpeg.wasm inside your browser, so the file is never uploaded anywhere. One segment per run; re-mark the points to take another from the same file without re-uploading.What bitrate does each tool write out?
Bitrate is how many bits per second the encoder is allowed to spend. It sets the file size exactly, and sets the quality ceiling loosely. Each of these tools has its own fixed answer, and none of them ask you.
| Tool and output | Encoder setting | Effective bitrate | Size per minute |
|---|---|---|---|
| Audio Extractor → WAV | Uncompressed samples | 1,411 kbps at CD quality | 10.6 MB |
| Audio Splitter → MP3 | LAME quality level 2, variable | about 190 kbps average | 1.43 MB |
| Audio Extractor → MP3 | Fixed 192 kbps | 192 kbps constant | 1.44 MB |
| Audio Extractor → AAC | Fixed 128 kbps | 128 kbps constant | 0.96 MB |
| MP3 Looper → MP3 | Fixed 128 kbps | 128 kbps constant | 0.96 MB |
The arithmetic is worth doing once, because it is the whole of audio file sizing: bitrate in kbps, times seconds, divided by 8, gives kilobytes. A 30-second cut at 190 kbps is 190 × 30 ÷ 8 = 712 kB. The same 30 seconds as CD-quality WAV is 1,411 × 30 ÷ 8 = 5.3 MB, about 7.4 times bigger. If you also work with video, the Video Bitrate Calculator runs the identical formula against a target file size.
How much audio fits inside a browser tool?
These tools run in the page, not on a server, so the ceiling is your browser's memory rather than an upload quota. The limits are stated in bytes, which means the duration you get depends on the bitrate you bring.
| Source encoding | Bitrate | Duration that fits |
|---|---|---|
| Low-bitrate voice MP3 | 96 kbps | 14 min 13 s |
| Standard MP3 | 128 kbps | 10 min 40 s |
| Good MP3 | 192 kbps | 7 min 07 s |
| High MP3 | 256 kbps | 5 min 20 s |
| Maximum MP3 | 320 kbps | 4 min 16 s |
| CD-quality WAV | 1,411 kbps | 58 s |
Worked example: the same edits in two orders
You have a 4-minute track at 320 kbps — 9.6 MB. You want a 30-second phrase from the middle, repeated 8 times, as a practice loop. Two tools, two ways round.
| Path A: split first, loop second | Path B: loop first, split second | |
|---|---|---|
| Step 1 | Splitter cuts 0:45–1:15 → 30 s at about 190 kbps = 0.71 MB | Looper repeats the phrase 8 times → 240 s at 128 kbps = 3.84 MB |
| Step 2 | Looper repeats it 8 times → 240 s at 128 kbps = 3.84 MB | Splitter trims the ends → 240 s at about 190 kbps = 5.70 MB |
| Lossy generations | 3 | 3 |
| Lowest bitrate in the chain | 128 kbps | 128 kbps |
| How it sounds | Capped by the 128 kbps stage | Capped by the same 128 kbps stage |
| Final file | 3.84 MB | 5.70 MB, 48% larger, no better |
Both paths sound the same, because both pass the audio through the same 128 kbps encoder and nothing later can rebuild what that stage discarded. Path B simply ends on the more generous encoder, which spends 190 kbps faithfully preserving audio that only ever contained 128 kbps of information. You pay 1.86 MB for nothing.
That gives the single most useful rule in this guide: whichever step has the lowest fixed bitrate, run it last. It sets the ceiling wherever it sits, so let it also set the size.
Getting audio out of a video without a second loss
A video's soundtrack is already lossy, usually AAC. What you do next decides whether you add another generation before your real edit even starts.
Choose WAV in the Audio Extractor when more editing is coming. WAV is not magic and will not undo the AAC compression already there, but it stores the decoded samples exactly, so nothing new is lost between extraction and the edit that matters. It is bulky — 10.6 MB a minute — and that is fine for a working file you will delete afterwards.
Choose MP3 at 192 kbps or AAC at 128 kbps when the extraction is the finished job: a podcast episode to listen to, a soundtrack to keep. One careful generation on top of the video's own is rarely audible, and the file is a tenth the size.
Three related jobs, three different tools. Remove Audio from Video keeps the picture and drops the sound, copying the video stream so the picture is untouched. Add or Replace Audio in Video puts a different track in — the usual fix when background music has triggered a copyright claim. Video Splitter and Adjust Video Speed both re-encode, so make your audio decisions before you reach for them, not after.
Audio Extractor → Pull the soundtrack out of a video and choose the output format: WAV to keep the decoded samples for further editing, or MP3 at 192 kbps, AAC at 128 kbps and OGG when the file is finished. The original video is never modified. MP3 Looper → Mark a loop on a visual waveform, set A and B points while it plays, preview it repeating, then export between 1 and 100 repetitions as a single MP3 at 128 kbps. Built for practising a passage, drilling pronunciation or working through a transcription.The edits that touch no audio file at all
Some jobs that feel like audio editing never open an audio file, which makes them the only genuinely free operations in this list. There is nothing to degrade because nothing is decoded.
The Music Key Transposer shifts written chords and notes from one key to another. It reads text and returns text, so transposing a chart up a minor third costs nothing and can be redone as many times as you like.
The Music BPM Tapper works out a tempo from the rhythm of your own taps. It never sees the recording, so it works on anything you can hear, including audio you are not allowed to upload anywhere.
The Tone Generator synthesises a pitch on demand rather than playing a stored file, so there is no source and no quality to lose. It is the reference tone for tuning, for hearing checks and for testing which speaker is which.
Four rules that cover most workflows
- Start from the best copy you can find. The version that has been forwarded through two chat apps has already spent generations you will never get back. Go to the original export or download.
- Do all your cutting in one pass. Each visit to the Audio Splitter is another generation, so mark the segment you actually want rather than trimming three times.
- Put the lowest fixed bitrate last. It caps the quality wherever it sits in the chain, so let it also cap the size — that is the 48% in the worked example above.
- Never re-encode upward to "improve" a file. Going from 128 kbps to 320 kbps produces a file two and a half times larger with exactly the same sound. Size records what the encoder was allowed to spend, not what survived.
Terms worth knowing
- Lossy
- A format that discards information to save space, keeping a reconstruction rather than the original. MP3, AAC and OGG Vorbis are all lossy.
- Lossless
- A format that can return exactly the samples that went in. WAV stores them plainly; FLAC compresses them without discarding anything.
- Generation
- One trip through a lossy encoder. The count matters more than any single bitrate, because damage accumulates and is never reversed.
- Bitrate
- Bits per second spent on the audio. Multiply by the duration and divide by 8 for the size in bytes.
- CBR and VBR
- Constant bitrate spends the same amount on every second. Variable bitrate spends more on complex passages and less on simple ones, giving better quality for the same average.
- LAME quality level
- A VBR setting on the LAME MP3 encoder where lower numbers mean higher quality. Level 2, used by the Audio Splitter, averages roughly 190 kbps.
- Sample rate
- How many times per second the waveform was measured. CD audio uses 44,100, which is why CD-quality stereo runs at 1,411 kbps.
- Stream copy
- Moving an encoded track into a new container without decoding it. This is what makes muting a video free: the picture is copied, not re-encoded.
- Transparency
- The point at which a listener cannot distinguish the encoded file from the source. It depends on the material, the listener and the equipment, not on the number alone.