#compression — Public Fediverse posts

I love the efficiency with which ImageMagick  has been created. The look of the user interface is deliberately simple; the software suite needs to be as blazingly fast as possible from the shell

@rl_dane

#imagemagick #Image #Magic ✨ #OpenSource #programming #compression #Graphics

May I assume that you used this wonderful software suite in your endeavor?

https://imagemagick.org/#gsc.tab=0

@rl_dane

#Image #Magic ✨ #OpenSource #programming #compression #Graphics

Not actually back yet, just wanted to toot this for fun... ;)

Raw (TGA/Targa format) 1200ppi US Letter scan: 381MiB
PNG version: 27MiB
After contrast enhancement: 2.3MiB
After reducing to 10 colors: 1.2MiB
After re-doing reduction, first converting to grayscale: 1001KiB
Reducing to 4 shades of gray (from original 256-shade image): 727KiB
Reducing to monochrome (no perceptible loss in quality): 556KiB
Conversion to PDF + OCR with aggressive JBIG2 compression: 264KiB

Overall compression factor from original scan: 1480.6:1

Overall compression factor from uncompressed monochrome version of scan: 61.7:1

Overall lossless compression factor (uncompressed monochrome to PNG monochrome): 29.2:1

Lossless compression factor for full-color image: 14:1

Data compression is a heck of a thing. ;)

#DataCompression #Compression #PNG #JBIG #JBIG2

https://medium.com/the-story-well/tighten-prose-without-losing-meaning-79dbcd94c2a3

Compression Techniques in Writing.

#compression #writing #writingtips #medium #writingadvice #storytelling #fiction #writerscommunity

🤖😂 Oh, the "AI Situation" - where software developers ponder the meaning of life through context windows and compression! 🧠💥 It's like watching someone explain how to teach a goldfish to do calculus - both time-consuming and utterly pointless. 📈🐠
https://srikanth.ch/posts/the-ai-situation/ #AI #Situation #SoftwareDevelopment #MeaningOfLife #ContextWindows #Compression #HackerNews #ngated

Ngrok's blog attempts to convince us that "compression is prediction" while simultaneously squeezing out more #buzzwords than a #Silicon #Valley startup pitch 🏢🔮. If you manage to navigate the labyrinth of tech jargon and promises of making AI smaller, faster, and cheaper, congratulations—you've just wasted time you'll never get back 🤦‍♂️.
https://ngrok.com/blog/compression-is-prediction #Ngrok #Compression #Prediction #TechJargon #HackerNews #ngated

Compression Is Prediction

https://ngrok.com/blog/compression-is-prediction

Comments: https://news.ycombinator.com/item?id=49263497

#HackerNews #Compression #Prediction #DataScience #Technology #Insights

borg 2.0.0b23-pre benchmarks from some essential components of borg:

https://github.com/borgbackup/borg/discussions/10056

#borgbackup #blake3 #hashing #chunking #zstd #compression #algorithms #simd

📢 In case you missed it: "Not an island: bringing compression to the tabular ecosystem".

Most compression libraries ask you to move in — learn their API, convert your data, leave your tools behind. We think that's backwards: compression should be a fast, compact layer *underneath* the tools you already use.

Blosc2 4.9.1: DuckDB, Polars, PyArrow and pandas 3 read a CTable directly via Arrow's PyCapsule protocol.

📖 https://blosc.org/posts/not-an-island-tabular-ecosystem/

#Arrow #DuckDB #Pandas #Compression #OpenSource

Chance by Friendgineer Labs 🎛️
Randomized multi-FX: delay, mod, reverb, harmonics + multi-band comp/limiter. Blend effects, explore presets.

💻 Mac/Win (VST3/AU/AAX)
🎁 FREE https://tokyospeirs.com/chance

More freeware 👉 https://linktr.ee/legalvst

#freeplugin #multifx #randomization #modulation #reverb #compression #friendgineer #audioplugin

В один эмбеддинг можно впихнуть больше тысячи токенов, но что с этим не так?

Хабр, привет! Меня зовут Дмитрий Тарасов, я исследователь в лаборатории FusionBrain AIRI. Мы недавно свозили нашу свежую работу — “Progressive Cramming: Reliable Token Compression and What It Reveals” —на ICML 2026 , и здесь я хочу рассказать, о чём она. Начнём с почти магического факта. Год назад в замечательной работе “Cramming 1568 Tokens into a Single Vector and Back Again” моего коллеги по AIRI Юры Куратова и его команды было показано, что в один‑единственный входной эмбеддинг замороженной LLM можно «упаковать» до 1568 токенов текста так, что модель потом слово в слово его восстанавливает. Полторы тысячи токенов — это несколько абзацев — в одном векторе размерности несколько тысяч, невероятная плотность памяти. Разбор этой статьи уже был на хабре . У нас этот результат вызвал не только восторг, но и подозрение. Одно дело — восстановить текст, и совсем другое — понимать его. Что, если «идеальная» реконструкция на самом деле не хранит смысл? В нашей статье мы: – покажем, что стандартный протокол token cramming скрывает катастрофические провалы, и предложим более честную процедуру — progressive cramming ; – исследуем траектории оптимизации сжатых последовательностей; – и, самое интересное, покажем, что идеальная реконструкция не сохраняет способность модели рассуждать — а причина этого сидит в первых слоях трансформера. Поехали.

https://habr.com/ru/companies/airi/articles/1062296/

#LLM #compression #cramming

Exploring Hidden JPEG Features

https://fed.brid.gy/r/https://hackaday.com/2026/07/24/exploring-hidden-jpeg-features/

Estrarre i file compressi da terminale senza sapere il comando giusto per ogni tipo, con il comando exex al volo e senza fatica.

#opensurce #UnoOpensource @opensource #compression

https://vitforlinux.wordpress.com/2026/07/07/estrarre-i-file-compressi-da-terminale-senza-sapere-il-comando-giusto-per-ogni-tipo-con-il-comando-ex-al-volo-e-senza-fatica/

How to pack ternary numbers in 8-bit bytes

https://compilade.net/blog/ternary-packing

Comments: https://news.ycombinator.com/item?id=48912812

#HackerNews #ternary #numbers #packing #binary #encoding #data #compression #computer #science

Crunchy Data author Christopher Winslett does a deep dive into the new column compression algorithm PostgreSQL 19 uses to compress some column types by default (LZ4), in this July 16, 2026 article.

He also goes into detail about the history of the previous algorithm (pglz, first appearing in version 7), and how LZ4 beats it to the punch in benchmarked performance.

"Postgres 19 Compression: from pglz to LZ4"

https://www.crunchydata.com/blog/postgres-19-compression-from-pglz-to-lz4

#databases #internals #compression #algorithms #postgresql

Lossless model compression experiment: GLM-5.2 in 25% less memory

https://brianbell-x.github.io/weight-compression/

Comments: https://news.ycombinator.com/item?id=48979201

#HackerNews #Lossless #model #compression #GLM-5.2 #memory #efficiency #AI #research #machine #learning

Regressive JPEGs (progressive JPEGs exploited): https://maurycyz.com/projects/bad_jpeg/
#graphics #jpeg #compression

I Thought AVIF Would Make Every Image Smaller—It Didn’t, by @malarkey:

https://stuffandnonsense.co.uk/blog/webp-to-avif-gotcha/?ref=frontenddogma.com

#avif #images #compression #webp

Regressive JPEGs

https://maurycyz.com/projects/bad_jpeg/

Comments: https://news.ycombinator.com/item?id=48954851

#HackerNews #RegressiveJPEGs #JPEG #compression #DigitalArt #TechInnovation #ImageProcessing

misa77 - a codec that decodes 2x faster than LZ4 (at better ratios)

https://github.com/welcome-to-the-sunny-side/misa77

Comments: https://news.ycombinator.com/item?id=48922838

#HackerNews #misa77 #codec #LZ4 #faster #data #compression #technology