Initializing... drag & drop files here
Supports: AU
.au files from your device. Batch is supported — every file in the queue inherits the same settings.pcm_mulaw (μ-law, 8-bit ~half the size of 16-bit PCM at 8 kHz, ITU-T G.711 — the historical AU default), pcm_alaw (A-law, the European G.711 variant), or stay on pcm_s16be / pcm_s16le (16-bit linear PCM, lossless but ~4x larger than μ-law).AU (also written .snd) is a Sun Microsystems / NeXT audio container introduced in 1992 with a 24-byte big-endian header followed by raw sample data. The same container can hold tiny 8-bit μ-law voicemail or 32-bit floating-point studio masters, so two AU files with identical playtime can differ in size by 8x or more. Compressing AU usually means picking a denser encoding inside the same container, not changing the file extension.
javax.sound.sampled reads AU natively, and many older interactive-voice-response platforms still expect 8 kHz mono μ-law AU. Compressing a 44.1 kHz stereo PCM AU down to 8 kHz mono μ-law cuts file size by roughly 22x and matches what the system actually plays./dev/audio and expect AU benefit from staying in-format rather than introducing a new decoder dependency.| Property | AU (.au / .snd) | WAV (.wav) | MP3 (.mp3) |
|---|---|---|---|
| Origin | Sun Microsystems, 1992 | Microsoft / IBM, 1991 | Fraunhofer / ISO MPEG-1 Layer III, 1993 |
| Byte order | Big-endian header | Little-endian header | Frame-based stream |
| Default encoding | 8-bit μ-law at 8 kHz | 16-bit linear PCM | Lossy MDCT, typically 128–320 kbps |
| Lossless options | PCM 8/16/24/32-bit, float, double | PCM, float | None — always lossy |
| Lossy options inside container | μ-law, A-law, ADPCM (G.721/G.722) | ADPCM, μ-law, A-law (rarely used) | n/a |
| Browser playback | Safari and Firefox (limited); Chrome no native | All major browsers | All major browsers |
| Typical use today | Java audio, legacy Unix, telephony | Editing masters, Windows recording | Sharing, streaming, podcasts |
| Encoding | Bits/sample | Typical bitrate (mono 8 kHz) | When to use |
|---|---|---|---|
pcm_s16be / pcm_s16le |
16 | 128 kbps | Editing or archival; lossless |
pcm_s24le |
24 | 192 kbps | High-headroom recordings |
pcm_mulaw |
8 | 64 kbps | Voice, telephony, IVR (G.711) |
pcm_alaw |
8 | 64 kbps | European telephony (G.711) |
pcm_s32le / float |
32 | 256 kbps | DSP, scientific work — overkill for voice |
For music, stay at 16-bit PCM and 44.1 kHz. For speech, μ-law or A-law at 8 kHz mono gives the smallest file with intelligible voice. For an exact target size, pick the percentage or target-size option instead of guessing a bitrate.
Most AU files are uncompressed PCM, which stores every sample at full bit depth. A 16-bit stereo PCM AU at 44.1 kHz runs ~1.4 Mbps — roughly 10 MB per minute — while a 128 kbps MP3 of the same audio is about 940 KB per minute. Switching the AU's internal encoding to μ-law cuts size by ~4x; converting to AU to MP3 cuts it further but leaves the AU world.
Use μ-law if the destination system is in North America or Japan (G.711 μ-law is the digital-telephony standard there). Use A-law for European telephony and most ITU international links. Both pack 16-bit dynamic range into 8 bits per sample at 8 kHz; μ-law has slightly wider dynamic range, A-law has less proportional distortion for quiet signals.
javax.sound.sampled?Yes for PCM (8/16-bit signed) and μ-law and A-law variants — those are the encodings the Java Sound SPI handles natively. If you encode the AU as Vorbis, FLAC, or some other non-PCM codec the JRE's default mixer won't play it without an additional SPI plugin.
Partially — you can downsample (e.g., 44.1 kHz to 22.05 kHz halves size) or downmix stereo to mono (halves size again). Both are lossy in the sense that they discard frequency or channel information, but the encoding itself stays PCM. Combining both on a voice file roughly quarters the size with little perceptible loss.
Lossless options (lowering bit depth from 24 to 16, or staying at PCM with a smaller sample rate that still covers your audio's frequency range) lose nothing audible for most material. μ-law and A-law are technically lossy — they keep about 13.5 effective bits in the encoded 8-bit byte — but voice and most non-musical content sound unchanged to the ear.
Free-tier users can compress AU files up to the size cap shown on the upload widget; signed-in tiers raise the cap. Processing is browser-session based, so very large files are limited mostly by your browser's available memory rather than a hard server quota.
Some editors decode μ-law or A-law to 16-bit PCM on import and re-save the file as PCM AU. The container looks identical but the encoding inside has been promoted to lossless PCM. Re-running it through this compressor with pcm_mulaw selected restores the smaller size.
If the consumer is a modern app or device, yes — AU to WAV keeps it lossless, AU to FLAC gives lossless plus ~50% size reduction, and AU to MP3 gives the smallest size for sharing. Compress within AU only when the destination explicitly requires the .au format (legacy Java apps, Solaris tooling, IVR systems).
Yes. Add multiple files to the queue and they all share the codec, bitrate, sample rate, and channel settings you pick. Each file is processed and downloaded independently, so a partial failure on one file does not block the rest. For the reverse direction see WAV to AU or the general Audio Compressor.