An audio extractor for pulling sound tracks from video files. Supports MP4, AVI, MOV, WebM, and other video formats — export audio as MP3, WAV, or FLAC directly in your browser.
Upload your video file — drag and drop or click to browse
Preview the video and check file information
Choose extraction mode: quick extract or convert to format
Optionally trim, adjust volume, or add fade effects
Click "Extract Audio" and download the result
Quick extract re-encodes nothing — it takes the track out of the container as it stands. So the format of the result is not your choice but whatever the sound was recorded as. The table shows which file you end up with.
| Audio codec | Where you meet it | File after “Original” |
|---|---|---|
| AAC | MP4, MOV, M4V, FLV, 3GP — phone and camera footage, editor exports, almost everything downloaded | .m4a |
| MP3 | AVI and older downloads | .mp3 |
| Opus | WebM, some MKV, voice messages from messengers | .opus |
| Vorbis | WebM and MKV of earlier years | .ogg |
| FLAC | MKV from music releases and concert recordings | .flac |
| ALAC | MOV and MP4 from the Apple ecosystem | .m4a |
| PCM | MOV and AVI straight off a camera or out of an editor — uncompressed sound | .wav |
| AC-3 | broadcast TS recordings, films in MKV | .ac3 |
| E-AC-3 | streaming captures and high-definition broadcast | .eac3 |
| DTS | films in MKV | .dts |
| MP2 | older broadcast and TS recordings | .mp2 |
| TrueHD | Blu-ray discs and their MKV rips | .thd |
| Codec not recognised | rare and damaged files | .mka |
If the codec is not recognised the track still comes out — it lands in an MKA container, which VLC, mpv and foobar2000 all open.
The frames below were captured on this very page: the same panels and the same buttons that open above. You can see what will happen to your file before you upload anything.
No narration — there is nothing to narrate, everything happens on screen. The video does not start downloading until you press the button; until then the page holds only a still image.
By drag-and-drop, by clicking, or by pasting. It takes MP4, AVI, MKV, MOV, WebM, WMV, FLV, 3GP, TS and M4V. The size ceiling is worked out from your device's memory and printed in the zone itself — higher on a computer than on a phone.
The tool reads the container and shows duration, size, resolution and video codec, and on the line below — the audio codec, its sample rate, channel count and bitrate. That second line is what decides everything that follows.
“Original” takes the sound out as it is, with no re-encoding. The presets alongside fill in format, bitrate and channels for a particular job in one click, but the format and quality selects stay in charge: a preset only puts values into them, and you can change them.
This is where trimming by time, fade-in and fade-out, volume and EBU R128 normalization, channel count, sample rate, speed and noise reduction live. The counter next to the heading shows how many settings are on, so a switch you forgot does not go unnoticed.
The result block names the format, size and duration of the finished track and lets you listen to it. Next to it sits a row of other formats: if the one you picked does not suit, the track is rebuilt from the audio already extracted, not from another pass over the video.
The presets on this page are named after jobs, not formats. Here is what sits behind each one, and when to take it.
A phone needs a short fragment in a format it will accept as a ring. On iPhone that is M4A, and it will not take more than forty seconds.
What to press: the “iPhone ringtone” preset — it sets M4A at 128 kbps and caps the length at forty seconds; the part you want is chosen with trimming.
What matters here is intelligibility and an even level. Stereo only doubles the size, because the speaker sits in the middle anyway.
What to press: the “For a podcast” preset (MP3 128 kbps, mono), then “Normalize loudness” targeting −16 LUFS — the level podcast platforms expect.
The sound is going into an editor, where it will be cut and processed. A second generation of compression is dead weight there, and compatibility matters more than size.
What to press: the “For video editing” preset — WAV. If your editor understands the original codec, “Original” is better: it costs neither time nor quality.
Keep the track alongside the video without losing anything on the way.
What to press: “Original” — an exact copy of what is written in the file. Take FLAC only if you need one format across a whole collection: it compresses already-decoded audio losslessly, but it does not bring back what the original codec threw away.
We took the file the tool produced and compared the checksum of the audio packets inside it with the checksum of the packets FFmpeg demuxes from the source video. They matched to the last byte. That means the track that came out of the container is exactly the one the camera or the editor put in.
The file compared was the one saved with the Download button, not an intermediate buffer inside the page.
Same clips, same computer: on the left the time in “Original” mode, on the right conversion to MP3. The gap narrows on longer files, because converting has a fixed part (loading the engine and parsing the container) while copying spends almost all its time reading bytes.
| Video | “Original” | MP3 |
|---|---|---|
| 3 min · 20 MB | 0.8 s | 6.4 s |
| 15 min · 102 MB | 6.2 s | 22.5 s |
| 45 min · 286 MB | 15.4 s | 48.2 s |
Output size is nearly the same either way: this source already carries 128 kbps audio, so MP3 does not make the file smaller — only worse.
Copying time grows with file size: 0.8 seconds at 20 MB and 15 seconds at 286 MB. That exact phrase — “in a couple of seconds regardless of the film's length” — is what this page promised in all ten languages, and the measurement disproved it. The wording was corrected in the same change that added this block.
A caveat about the numbers themselves. The first run of this benchmark happened while the machine was busy with other work and reported 44 s instead of 15 and 91 s instead of 48 — three times higher. The figures above come from an idle machine; on a busy computer yours will be higher.
Boundaries: one desktop machine and one kind of source (H.264 + AAC). On a phone the numbers will differ, and so will they for another codec — AC-3 or PCM inside MKV copy the same way but convert differently. The measurement answers “which is more expensive”, not “how long this will take on your machine”.
Reproduce it yourself: node tests/playwright/extract-audio-copy-vs-convert-bench.mjs. The script builds the clips, drives the tool and compares the checksums.
Tembrica processes everything in your browser via FFmpeg WebAssembly — fast conversions without waiting in a server queue. You get both lossless quick extraction and format conversion with quality control — all in one tool.
Not if you take the quick route. Quick extract copies the audio stream out of the container untouched — the AAC inside an MP4 arrives as AAC, bit for bit what the camera or the editor put there, and it runs many times faster than converting, because the work is reading the file rather than recomputing the sound. Converting instead gives you MP3 or WAV and costs one generation of quality, which is worth it only when the thing you are handing the file to cannot read the original codec.
Because what sat inside your video was AAC, and “Original” does not change the codec — it only moves the track into the container that suits it, and for AAC that is M4A. You can rename the same file to .aac, though players read it either way. If you specifically need MP3, choose MP3 in the format select: that is re-encoding, and it costs one generation of quality.
Keep it when the file is going into further work — an editor, another tool here, an archive — because copying costs nothing and loses nothing. Convert when the file is going to a person or a device: MP3 opens everywhere without a question, while the AAC or Opus that came out of a phone video occasionally does not. Converting twice is the thing to avoid, so decide once and take the copy if you are unsure.
This tool works with video files on your device. You need to first download the video file, then use this tool to extract the audio.
Some formats like MKV, AVI, WMV, and FLV are not natively supported by browsers for playback. The preview will show a placeholder, but audio extraction will still work correctly.
Yes, and trimming is the one setting that gets along with quick extract. Start and end times are passed before the file is read, so the track is still copied rather than rebuilt: you can take one minute out of a three-hour recording at original quality. Volume, fades, normalization, channel or sample-rate changes, by contrast, force a re-encode — by definition they cannot be applied to a copy.
Yes. When the tool finds more than one track, a list appears above the advanced settings with their count, and each can be extracted separately. That is how people take apart films with several dubs, recordings with a separate microphone channel, and broadcasts where commentary runs as a second track. Subtitle tracks, if the file carries any, are shown in a block of their own.
We can! Just send us a quick message with your idea. If you'd like to discuss it in detail, leave your email and we'll get back to you. You can stay anonymous.
