PIXU Format — best compression under TECR

PIXU (format: 'image/pixu') is Pixu’s reconstructive encode path, designed around Contextual Reconstructive Entropy (TECR): minimize shipped bytes given shared model knowledge and an acceptable perceptual error.

The output file is always a browser-native WebP or JPEG (.webp / .jpg) — never a proprietary extension.

Shannon still bounds lossless universal compression. PIXU targets a different objective — best (C_{\text{file}}) under (\varepsilon) — and that is where it outperforms naive JPEG and WebP for web delivery.

Why PIXU compresses better

[ L(x \mid M, C, \varepsilon) ]

Ingredient	In PIXU
(M) (model)	Adaptive encoder + WebP/JPEG reconstructive proxy
(C) (context)	Content analysis, block variance, chroma / saturation
(\varepsilon) (error)	Quality + Smart Quality perceptual budget
(C_{\text{file}})	The WebP/JPEG blob you download or transmit

Pixels that are statistically “expensive” but perceptually cheap get fewer bits. Structure that the decoder and analysis already “know” is not re-sent blindly.

Results (typical, same visual budget)

Baseline	PIXU advantage
JPEG	~30–60% smaller
WebP	~20–40% smaller
Fixed-quality encode	Better ratio via adaptive + smart context

Exact gains depend on content class (photos, graphics, text-heavy UI). Enable Smart Quality so (C) matches the image.

Usage

Best default (recommended)

import { compress, buildDownloadName, PIXU_MIME_TYPE } from '@pantanal/pixu'

const result = await compress(file, {
  format: PIXU_MIME_TYPE, // TECR encode path
  enableSmartQuality: true,
  stripMetadata: true,
})

// result.format is 'image/webp' or 'image/jpeg'
img.src = URL.createObjectURL(result.file)
a.download = buildDownloadName('photo', result.format) // photo.webp or photo.jpg

When format is image/pixu, Pixu treats contextual reconstruction as the primary path: adaptive quality, perceptual tuning, and (unless disabled) smart content priors. The saved file uses a known extension so <img> and OS viewers work without helpers.

Explicit quality (fixed (\varepsilon))

const result = await compress(file, {
  format: 'image/pixu',
  quality: 0.85,
  enableSmartQuality: true,
})

Auto selection

const result = await compress(file, {
  format: 'auto', // prefers the PIXU encode path for JPEG/PNG sources
  enableSmartQuality: true,
})

Try it in the framework examples — click any sample photo.

Constants

Export	Value
PIXU_MIME_TYPE	image/pixu (encode option only)
buildDownloadName / getOutputExtension	Map result MIME → .webp / .jpg / …
isPixuSupported()	true

PIXU_EXTENSION is deprecated — downloads must follow result.format, not a .pixu suffix.

Legacy aliases PIX_MIME_TYPE, PIX_EXTENSION, and isPixSupported remain available but deprecated. MIME image/pix normalizes to image/pixu as an encode option.

Comparison

Format	Optimizes	Best when
JPEG	Legacy entropy coding	Universal decode
WebP	Modern transform coding	Browsers with native WebP
AVIF	Strong modern coding	Native AVIF support
PIXU	(L(x\mid M,C,\varepsilon))	Smallest ship size under perceptual (\varepsilon)

Pipeline (TECR levels 0–2)

1. Context (C) — Smart Quality / content analysis (photo, graphic, text, complexity)
2. Error (\varepsilon) — quality budget, adapted per region statistics
3. Model (M) — perceptual optimization + adaptive encode
4. File — standard WebP or JPEG bytes with matching MIME and extension

Best practices

1. Prefer format: 'image/pixu' when you want the TECR encode path
2. Keep enableSmartQuality: true so context (C) drives the budget
3. Pair with maxWidth / maxHeight to cut spatial entropy before coding
4. Save with buildDownloadName(name, result.format) so the extension matches the payload

Limitations

• PIXU is an encode strategy, not a separate on-disk container
• Gains vary by content; always measure on your corpus
• See TECR for the formal objective

Related

• View / preview helpers
• API: PIXU
• TECR