Benchmark Summary
Cher Ami scores with best compression ratio on all probed lyrics of European anthems; also best on bilingual text and chat with emoji; second-best on Asian anthems; and third on content with severe duplication. All in all, 18 best compression ratios out of 24 comparisons against the most capable compressors in this category.
The Cher Ami Cipher
Special offers may vary by location, store, or even declared use case.
Lossless Compression Results
All sizes are in bytes, the smaller the better. The best-compressed size for each text is in bold.
Legend
Text — the uncompressed UTF-8 BOM text files, downloadable.
Text Size — the uncompressed size of the text files in UTF-8 bytes.
BZip2, PPMd, Zstd, Brotli, Deflate, LZW, Cher Ami — compressed sizes in bytes, named after the method/algorithm/protocol.
Methods, Implementations, Settings
Deflate, web-based
Implementation: MultiUtil.com.
Output encoding: Base64.
Brotli, web-based
Implementation: MultiUtil.com.
Output encoding: Base64.
BZip2, web-based
Implementation: TxtWizard.net.
Output encoding: Base64.
LZW, web-based
Implementation: PlanetCalc.com.
Initial dictionary: Encoding, UTF-8. Otherwise, a custom dictionary per text should be considered part of the payload and added to the compressed size.
Output encoding: Binary/Radix-2.
Brotli, file-based
Implementation: PeaZip ver. 7.9.0.
Compression level: 9.
Zstd, file-based
Implementation: PeaZip ver. 7.9.0.
Compression level: 19.
PPMd, file-based
Implementation: 7-Zip ver. 18.05 (x64).
Archive format: 7z.
Compression level: Ultra.
Dictionary size: 4 MB.
Word size 32.
Non-solid block size.
Gave similar ratios to compression.
BZip2, file-based
Implementation: 7-Zip ver. 18.05 (x64).
Archive format: bzip2.
Compression level: Ultra.
Dictionary size: 400 KB.
Honorable Mentions
UHARC GUI ver. 4.0.0.2, archive format uha, dictionary size 4 MB, no media – results similar to the presented PPMd file-based compression of 7-Zip.
7-Zip ver. 19.00 (x64), archive format xz, method LZMA2:12, dictionary size 64 MB, word size 64, non-solid block size – worst compression ratios, but still popular and preferred by many for other compression inputs.
FAQ
It’s due to design. Cher Ami doesn’t reuse any third-party compression software or dictionaries. While all other tools and implementations are bound by their compatibility with legacy algorithms, structures, and protocols, this app isn’t. It practically introduces a unique approach in everything it does in the field of data compression complying only with some faultless mathematical ideas among all relevant research.
The purposes of other algorithms and dictionaries also matter. Some may claim better results on longer or linguistically improper texts which are out of scope for the purpose of Cher Ami. Claims…
Cher Ami is aimed at delivering good compression on short messages in any language. Therefore, it was essential to pick wording that is natively common in some of the world’s most popular languages. The used sentences from national anthems added a bit of cultural significance on top of that. Emoji is also important for messaging nowadays.
None validated by me. Benchmarking the compression of Cher Ami is an activity well covered with its EULA. I’m not yet making free editions for this purpose, nor other investments in similar PR. It may take a lot more time for this compressor to gain the benchmarking popularity of freeware competitors that have been around for decades.
Every reliable third-party benchmark on up to 1K code points of UTF-16 (any Unicode sequence) would be interesting. My best advice is to try your own text with the listed web-based compressors and then with Cher Ami. I wouldn’t compare compression ratios from significantly different text lengths or languages.
The most popular third-party benchmarks focus primarily on much longer texts in English which are out of scope for the purpose of this app. Even the Squash Compression Benchmark and their smallest test file (grammar.lsp) are inapplicable for such comparison. Another disappointment may arise from the fact that most other publications seldom consider the inner workings of the compressors or don’t disclose all setup details in a consumer-reproducible manner.
The usual way:
(Compression ratio) = (uncompressed size) / (compressed size)
(Compression %) = [(uncompressed size) – (compressed size)] / (uncompressed size)
However, some tools may introduce different labeling of the same numbers.
No. Cher Ami offers the option to copy the compressed contents and doesn’t save them to the file system. For the file-based implementations above, only the logical sizes of the compressed text (a.k.a. “packed size”) are taken into account. Some headers, metadata blocks, information about compression methods, CRC, file names, and timestamps were ignored in favor of the third-party compressors. This way odds and handicaps originating from the file system, cluster size, fragmentation, and storage type don’t affect the presented results. For example, the PPMd compressed archive of multilingual02_FriendlyEmoji.txt has a downloadable archive size of 563 bytes while only the packed size of 393 bytes is listed for the comparison.
Cher Ami doesn’t offer a public compression API. While API benchmarking may be easier for me, retesting it is considered harder for the target audience and even impossible for some users of the app. Manual reproducibility is of higher importance than automation in this article.
No, but in my case it is. My first proof of concept for the compression was back in 2004. So, I already had my own prototype of an ASCII compressor since my childhood and didn’t have to tackle third-party copyrights in order to develop derivative works from it. In the end, this happened to be not only more efficient but a more effective solution as well.
Disclaimer.
This is neither advertising nor a commercial offering. Cher Ami is provided as a signed installation package, strictly on an ‘as is’ and ‘as available’ basis. This page does not create any obligations for the author.
