HALAC 0.6.1

Most lossless audio encoders compress on a single CPU thread, so fifteen cores of a sixteen-core machine sit idle while one does the work.

HALAC spreads the job across every core, and version 0.6.1 closed most of the compression gap that used to be the price of that speed.

Download HALAC 0.6.1 - free, Windows 10 and 11, command line only.

What changed in 0.6.1

Released on 29 July 2026, and the compression work is the headline.

• Much better compression, particularly on multichannel audio. Against its own previous build at the same setting, output is around 17 percent smaller.
• Much faster decoding. Decode times drop by roughly a fifth on the developer's test sets.
• A new post filter for blocks that share a channel, which is where most of the multichannel gain comes from.
• Lower memory use, and encoding speed that adapts to the material rather than running one fixed path.

The numbers that matter

This is the developer's own benchmark on a 5.1 channel set, 16-bit 48 kHz, twenty tracks, from a 2,772,761,422 byte source.

HALAC at its highest level finishes within a quarter of a percent of Monkey's Audio Insane, beats WavPack by six percent and beats FLAC at its maximum setting by seventeen percent.

The 7.1 channel set repeats it: 1,684,430,726 bytes against FLAC -8 at 1,966,743,441.

Stereo is closer, as you would expect, but even there the picture has changed. On a 24-bit stereo set the previous 0.5.4 build already edged past FLAC 1.5.0 at -8 while encoding in 11.6 seconds against FLAC's 50.8.

Why it is fast

Two reasons, and only one of them is threading.

Instead of the Rice coding used by FLAC, Monkey's Audio and most other lossless codecs, HALAC builds its pipeline on Asymmetric Numeral Systems, usually shortened to ANS. It is a newer entropy coding method that is cheaper to run at similar efficiency.

On top of that, the encoder splits work across as many cores as you give it with the -mt flag, which is where the headline throughput comes from.

Compression levels

Compression is set by six numbered levels, -m 0 through -m 5. The older Normal, Fast and Ultra Fast names were retired in 0.5.7.

• -m 0 - faster than the old Ultra Fast, and the quickest setting available.
• -m 1 - what Ultra Fast used to be.
• -m 3 - what Normal used to be, and the sensible default. Most of the ratio, noticeably faster than -m 5.
• -m 4 - slightly faster than the old Plus mode.
• -m 5 - the old Plus mode, and the setting the benchmarks above use.

If you have batch scripts written against the old labels, they broke at 0.5.7. Update the flags before running a large job.

Using it

Parameters are space-separated, and the input filename alone is enough to drive compression or decompression.

A typical encode reads encoder input.wav -o output.halac -m 3 -b 2048 -mt 8 -ow -s, where -o sets the output path, -mt sets thread count, -b sets block size and -s switches on silent mode.

Silent mode is worth using inside loops. By default both encoder and decoder print detailed information, and that reporting costs measurable time across thousands of files.

The -lw flag tells the encoder its input is lossyWAV output, which produces markedly better results on pre-processed material.

Where it fits in a lossless workflow

1. Rip the source. Exact Audio Copy or CDex read CDs to bit-perfect WAV with offset correction and AccurateRip verification.
2. Keep the album structure. CUETools holds CUE sheets aligned through conversions, splits single-file albums and verifies against AccurateRip and CTDB.
3. Preprocess, optionally. lossyWAV removes inaudible detail first, for smaller archives at perceptual transparency. Pair it with -lw.
4. Encode. One line in PowerShell or a batch file. The CLI takes wildcards, so a folder is one command.
5. Make portable copies when needed. Decode back to WAV and feed it to LameXP, winLAME or the foobar2000 Free Encoder Pack.

The honest limitation: playback

No mainstream player opens .halac files. The source ships a player API for developers, but there is no plugin you can install today for foobar2000 or anything else.

In practice that makes HALAC an archive format rather than a listening format. Compress the masters, keep a FLAC copy for daily use, and decode from HALAC when you need the original back.

That is a real cost and it is the main reason to think twice, particularly given the format is still pre-1.0 and evolving quickly.

How it compares

FLAC is the universal default and will stay that way. Every hardware player, streaming protocol and operating system reads it, and no benchmark result changes that.

Monkey's Audio is the ratio benchmark HALAC now sits alongside, but its encoder is far slower and decoding costs real CPU during playback.

WavPack offers hybrid mode, where a smaller lossy file plus a correction file restores bit-perfect audio. Nothing else here does that.

OptimFROG still holds the outright compression crown and is still famously slow.

The summary: choose HALAC when you are archiving large amounts of audio, especially multichannel, and encoding time is a real cost. Choose FLAC when anything needs to play the file back.

Who should look elsewhere

If you are ripping a CD to listen to, this is not your tool. FLAC plays on everything you own and HALAC plays on almost nothing.

If the command line puts you off, there is no GUI. X FLAC Encoder gives you a window and buttons over standard FLAC, and our GUI versus CLI guide covers the trade-off.

The tell that HALAC is right for you: a WAV archive measured in thousands of files, cores to spare, and a memory of watching an encoder run overnight.

The same developer also makes HALIC for lossless images, and the HALAC and HALIC guide covers both.

Quick questions

Download HALAC 0.6.1 and see how it handles your archive, or tell us how you got on if you have used it before.