Cut AU

How to Cut AU Files Online

1. Upload Your AU File: Drag and drop or click "+ Add Files" to add .au or .snd files from your device. Batch is supported, so you can trim several clips in one session.
2. Set Start and Duration: Enter the start point and end point (or start + duration) in hh:mm:ss.ms format — for example 00:00:02.500 to 00:00:09.000 keeps a 6.5-second window. The trim happens at the byte level on the original PCM, µ-law, or A-law payload, so there is no re-encoding when you keep the source codec.
3. Adjust Output (Optional): Keep the original AU encoding for lossless cuts, or change sample rate (8000, 11025, 22050, 44100, 48000 Hz), channel layout (mono or stereo), and codec (µ-law, A-law, PCM 8/16/24-bit) if your target system needs a specific profile.
4. Cut and Download: Click "Cut". Files process in your browser session — no sign-up, no watermark, no upload to a third party once processing completes.

Why Cut AU Files?

AU (also written .snd) is Sun Microsystems' Unix audio format, introduced with SunOS and adopted by NeXT, Java applets, and a generation of telephony systems. The 28-byte header carries six big-endian 32-bit words — magic number 0x2e736e64 (".snd"), data offset, data size, encoding, sample rate, channel count — followed by raw audio. Because the payload is uncompressed (PCM) or lightly companded (µ-law/A-law), cutting a region is a clean byte-range operation: no MDCT, no Huffman tables, no frame-boundary math.

• Telephony prompts and IVR menus — Most legacy PBX, voicemail, and IVR systems (Asterisk, Avaya, Cisco) expect 8 kHz, 8-bit µ-law (G.711) mono AU clips. Trimming a 30-second recording down to a 4-second greeting keeps it under the per-prompt slot without resampling.
• Java applet and Java Sound source files — The original java.applet.AudioClip API only accepted 8 kHz µ-law mono AU. The interface was deprecated in Java 9 and is marked for removal, but legacy courseware, lab simulations, and academic Java archives still ship .au assets that need shorter, focused clips.
• Unix/Linux /dev/audio pipelines — SunOS introduced /dev/audio as a raw µ-law sink, and cat clip.au > /dev/audio still works on a few embedded BSDs and historical workstations. Cutting AU lets you build short signal/alert tones without WAV header rewrites.
• Audio dataset curation — Speech corpora distributed in the 1990s and early 2000s (TIMIT, NIST Switchboard subsets) used AU with µ-law at 8 kHz. Researchers cut utterance boundaries from these archives for re-use in modern ASR or phonetics work.
• Sound effects in NeXTSTEP / OPENSTEP archives — System sounds and game audio from NeXT and early macOS Server installations are distributed as .snd. Trimming silence or unwanted leaders preserves the original encoding for emulator and museum use.
• Cross-platform handoff to MP3/WAV/FLAC — Once trimmed, the same clip can flow into AU to MP3, AU to WAV, or AU to FLAC without going through the cut step again.

AU vs WAV vs AIFF — Format Comparison

Property	AU (.au / .snd)	WAV (.wav)	AIFF (.aif / .aiff)
Origin	Sun Microsystems / NeXT (late 1980s)	Microsoft / IBM (1991)	Apple (1988)
Byte order	Big-endian (header + PCM by default)	Little-endian	Big-endian
Header size	28 bytes minimum (fixed fields)	44 bytes (RIFF chunks)	Variable (IFF chunks)
Magic bytes	0x2e736e64 (".snd")	RIFF ... WAVE	FORM ... AIFF
Native encodings	µ-law, A-law, PCM 8/16/24/32-bit, IEEE float, G.72x ADPCM	PCM, IEEE float, µ-law, A-law, ADPCM	PCM (AIFF), plus compressed (AIFF-C)
Typical use	Unix, Java, telephony	Windows, general PC audio	macOS, pro audio
Max practical file size	~2 GB (32-bit data-size field)	4 GB (RIFF), 16 EB (RF64/W64)	~2 GB
Streamable	Yes (sequential header)	Yes	Yes

AU Encoding Quick Guide

Encoding	Bits / sample	Typical rate	Bitrate	Best for
µ-law (G.711)	8-bit log	8000 Hz	64 kbps	Telephony, IVR, voicemail
A-law (G.711)	8-bit log	8000 Hz	64 kbps	European telephony (E1 lines)
PCM 8-bit	8 linear	8000–22050 Hz	64–176 kbps	Legacy NeXT system sounds
PCM 16-bit	16 linear	22050 / 44100 Hz	352–705 kbps	CD-quality clips, music
PCM 24-bit	24 linear	44100 / 48000 Hz	1.06–1.15 Mbps	Mastering / archival capture
IEEE float 32	32 float	48000 Hz	1.54 Mbps	DSP intermediate, scientific audio

Frequently Asked Questions

Does cutting reduce audio quality?

No, not when you keep the original AU encoding. AU's PCM, µ-law, and A-law payloads are uncompressed at the codec level (µ-law and A-law are companding, not frame-based compression), so trimming is a byte-range copy with a rewritten 28-byte header. Quality only changes if you switch sample rate, bit depth, or codec on the way out — for example exporting a 16-bit PCM AU as 8-bit µ-law will collapse dynamic range from ~96 dB to ~38 dB.

What is the difference between .au and .snd?

They are the same file format with different conventional extensions. NeXTSTEP soundfiles typically used .snd; Sun and most Unix workstations used .au. Both share the 0x2e736e64 magic number (".snd" in ASCII) at byte 0. You can rename a .snd to .au (or vice versa) without re-encoding — the cutter accepts both.

Will the cut clip still play in Java applets and modern Java?

If you keep 8 kHz, 8-bit µ-law, mono, it will play on the original java.applet.AudioClip API. That interface was deprecated in Java 9 (and is marked for removal), so for modern code use javax.sound.sampled — which accepts AU at 8–48 kHz, mono or stereo, in 8 or 16-bit PCM as well as µ-law and A-law. Keep the codec unchanged in step 3 to preserve the broadest compatibility.

Can I cut an AU file recorded by Asterisk or another PBX?

Yes. Asterisk, FreeSWITCH, and most commercial PBX systems write 8 kHz, 8-bit µ-law mono AU as their Format::ulaw output. Cut to the segment you need and leave the encoding unchanged — the trimmed file will drop back into /var/lib/asterisk/sounds/ or your dialplan without conversion. If your switch expects A-law (common on European E1/PRI lines), set the codec to PCM A-law in step 3.

Why does my AU file sound noisy or "telephone-like"?

µ-law and A-law are logarithmic 8-bit encodings designed for 200–3400 Hz voiceband telephony. They give roughly 38 dB of perceived dynamic range — better than linear 8-bit PCM (48 dB SNR for voice) but well below CD-quality 16-bit (96 dB). If your source is music or wideband speech, switch the output encoding to 16-bit PCM at 44100 Hz before cutting to avoid baking the telephony band-limit into the trimmed clip.

What is the maximum AU file size or duration I can cut?

The AU header stores data size in a single unsigned 32-bit big-endian word, which caps a single file at roughly 4 GB of audio payload (some readers treat the field as signed and cap at 2 GB). At 8 kHz µ-law mono that is around 596 hours; at 48 kHz 16-bit stereo it drops to about 6.2 hours. xconvert applies its own per-session upload cap, but the AU container itself is the longer-term ceiling for very long recordings.

Will cutting preserve the annotation / info field in the header?

The AU header reserves space between byte 24 and the data_offset for an optional ASCII annotation block (description, copyright, recording notes). xconvert keeps the annotation when you cut without changing the encoding. If you re-encode (different codec, sample rate, or channel count), the annotation is rewritten with the new offsets but the text content is preserved.

Can I cut multiple regions and join them?

The cutter trims one contiguous range per pass. To assemble a highlight reel — for example two non-adjacent quotes from a 30-minute lecture — cut each region separately, then chain through a converter or merge them after converting to a more widely supported container. Alternatively, cut to WAV first, which gives you more downstream editing tools.

How is this different from converting AU to MP3 and trimming there?

Cutting AU keeps the original codec and avoids the lossy MDCT pass that MP3 encoding adds. If your destination is genuinely MP3 (podcast distribution, web embed), it is more efficient to cut first in AU and then run AU to MP3 on the trimmed clip — that way the lossy encode only processes the seconds you need, not the full source.