MJPEG to AU Converter

MJPEG to AU — Extracting Audio Into the Sun/NeXT Format

.au is one of the oldest audio file formats still in circulation — the one that originated on Sun and NeXT workstations, that Java's AudioClip API grew up with, and that Unix telephony and voice-response systems have used for decades. It has almost no structure: a short fixed header, then raw samples.

This page extracts the audio track from a Motion JPEG source and writes it as .au. Two things make the page unusual, and both are covered below: the source often has no audio at all, and the output has no compression settings whatsoever.

First, Does Your File Have a Soundtrack?

Motion JPEG is a way of coding video, and files called .mjpeg come in two shapes that behave completely differently here.

A raw Motion JPEG stream — JPEG frames written back to back, which is what a bare .mjpeg normally is — has no audio track, because the format provides nowhere to store one. We probed one: a single video stream, no duration, nothing else. Asking it for audio stops with "Output file does not contain any stream" and no file is written. That is the correct outcome; a silent .au would just look like success.

An AVI, MOV or MKV carrying an MJPEG video track that someone renamed to .mjpeg is a different animal. Files are demuxed by their actual contents, not their extension, so these work normally — we confirmed a renamed AVI probing as avi with two streams and extracting cleanly. If that is your situation, giving the file its real name and using AVI to AU is the tidier path.

Old camcorders and digital cameras produced the second kind. IP cameras, machine-vision rigs and microscopes produce the first kind, and those are genuinely silent.

How to Convert MJPEG to AU

1. Upload Your MJPEG File: Drag the file in or click "+ Add Files". If the job reports no stream to write, the source is a raw Motion JPEG stream — see above.
2. Choose Audio Channel: Original keeps the source's channel count; Mono collapses to one channel and exactly halves the output of a stereo source; Stereo forces two.
3. Choose Audio Sample Rate, or Trim: Original preserves the source rate, and 8000, 12000, 16000, 24000, 44100 and 48000 Hz are offered. Because AU output is uncompressed, this dropdown is the main size control you have. Trim takes a start time and a duration in seconds or HH:MM:SS.sss.
4. Convert and Download: Click "Convert" and save the .au. Files are uploaded over an encrypted connection, processed on our servers, and deleted automatically after a few hours — no sign-up, no watermark.

There Is No Bitrate Setting, and That Is Correct

You will notice no File Compression group on this page and no codec choices under Audio Codec. That is not an omission. AU output here is 8-bit G.711 μ-law — a companded format with a fixed one byte per sample per channel — so there is no bitrate to set and no quality slider to move. The size of the file is pure arithmetic:

bytes ≈ 32 + (duration in seconds × sample rate × channels)

We measured a four-second mono 44.1 kHz extraction and got 176,432 bytes — exactly 4 × 44,100 × 1 plus a 32-byte header. Dumping that header shows the format's whole design: the magic .snd, a data offset of 32, a data length of 176,400, an encoding code of 1 (8-bit μ-law), a sample rate of 44,100 and a channel count of 1.

If the output is bigger than you want, there are exactly two levers — drop the sample rate, or drop to mono — and they are both in step 2 and 3 above.

What μ-law Actually Costs You

μ-law is not raw linear PCM. It uses a logarithmic curve so that eight bits carry roughly fourteen bits' worth of dynamic range, spending precision where the ear is sensitive and saving it where the ear is not. That makes it far better than 8-bit linear audio and noticeably worse than 16-bit.

In short: μ-law is a telephony codec doing a telephony job well. If you are extracting dialogue, narration or a system alert, it is entirely appropriate. If you are extracting music you intend to master, it is the wrong container and you should be reaching for a lossless format instead.

Frequently Asked Questions

The conversion says there is no stream to write. Is the tool broken?

No — your source has no audio. A raw Motion JPEG file is a sequence of JPEG images with no audio track in the format at all, so an audio extraction has nothing to work with and stops rather than producing an empty file. Check what your file really is: if a media player shows a volume control and a waveform for it, it is a container in disguise and belongs on the page for its real extension.

Why is the .au file so much bigger than my MP3s?

Because nothing is compressed. MP3 at 128 kbps stores a minute of stereo in about 960 kB; the same minute as 44.1 kHz stereo μ-law is about 5.3 MB, because every sample is written out as a byte. That is the trade AU makes — trivial to decode, trivial to seek, expensive to store.

Can I choose 16-bit or A-law instead of μ-law?

Not from this page — no codec choices render here, so the output is 8-bit μ-law. For reference, A-law is the same size (it is the other 8-bit companding law, used on European telephony networks) while 16-bit big-endian PCM is exactly twice the size; we measured 176,432 bytes against 352,832 on the same four-second clip. If you need 16-bit audio, a WAV or FLAC target is the better destination.

What is the .snd at the start of the file?

The format's magic number, 0x2E736E64. It reflects the format's history — the same data structure was used for the .snd files on NeXT machines and the .au files on Sun ones. Anything that reads AU looks for those four bytes and then reads the header fields that follow: data offset, data size, encoding code, sample rate and channel count, all big-endian.

Will this play on a modern computer?

Yes, in most media players — VLC, ffmpeg-based tools, Audacity, and Java applications all read AU. What will not play it are consumer devices: car stereos, most phones' default music apps and smart speakers have no reason to support a Unix workstation format. AU is a format for programs, not for playlists.

Should I set the sample rate to Original or pick one?

Original if you want faithfulness and do not care about size. Pick one if you do — and pick it based on content, not habit. Speech carries fine at 8000 Hz, which is what telephony uses and where μ-law was designed to live; 16000 Hz makes speech noticeably clearer; anything above 24000 Hz on a voice recording is spending bytes on frequencies the source probably never captured.

What happens to the video?

It is discarded. The JPEG frames are read only far enough to work out the file's structure and never appear in the output. To keep the picture as well, MJPEG to MP4 re-encodes both streams into a single modern file — and shrinks it substantially, since Motion JPEG stores every frame in full with no compression between frames.

Can I extract just part of a long recording?

Yes, and with AU you should. Trim takes a start time and a duration, and because the output is uncompressed, trimming is the difference between a 3 MB file and a 300 MB one on a long capture. Set the start and duration before converting rather than trimming the .au afterwards.

Is stereo worth keeping?

Only if the source really is stereo. Many camcorder and webcam recordings carry a single microphone duplicated across two channels, in which case Stereo doubles the file for no information at all. Set Audio Channel to Mono and check the result — if it sounds identical, it was never stereo to begin with.