DEV Community

Lank_M profile picture

Lank_M

404 bio not found

Joined Joined on 
Your compressor dropped color_range, so now the player decides

Your compressor dropped color_range, so now the player decides

Comments
4 min read

Want to connect with Lank_M?

Create an account to connect with Lank_M. You can also sign in below to proceed if you already have an account.

Already have an account? Sign in
Blurry canvas export: give preview and export separate render paths

Blurry canvas export: give preview and export separate render paths

Comments
4 min read
PNG beat JPEG on one PDF page and lost 7.7x on another

PNG beat JPEG on one PDF page and lost 7.7x on another

Comments
3 min read
Hair-edge error in background removal: about 54 from the model, about 1 from mask upsampling

Hair-edge error in background removal: about 54 from the model, about 1 from mask upsampling

Comments
4 min read
One text upscale won on SSIM and lost on PSNR, so let the downstream pick the metric

One text upscale won on SSIM and lost on PSNR, so let the downstream pick the metric

Comments
4 min read
Four DOCX converters agreed on screen and disagreed in the markup

Four DOCX converters agreed on screen and disagreed in the markup

Comments
4 min read
One runtime, one GPU, two matting models: one matched WASM to 1.3e-6, the other returned an empty mask

One runtime, one GPU, two matting models: one matched WASM to 1.3e-6, the other returned an empty mask

Comments
4 min read
Screen recordings get fat at capture time: 50,002 bytes per still frame at constant bitrate, 77 after re-encoding

Screen recordings get fat at capture time: 50,002 bytes per still frame at constant bitrate, 77 after re-encoding

Comments
4 min read
Each OCR engine had its own minimum text height, and none had a maximum up to 60 px

Each OCR engine had its own minimum text height, and none had a maximum up to 60 px

Comments
4 min read
Size PDF downsampling tiers by the smallest text you must read, not by dpi

Size PDF downsampling tiers by the smallest text you must read, not by dpi

Comments
4 min read
Three color checkpoints for HEIC to JPEG, and why sips -g profile isn't one

Three color checkpoints for HEIC to JPEG, and why sips -g profile isn't one

Comments
4 min read
Inside a pngquant PNG: 23 palette colors, 6 tRNS bytes, 7 alpha levels

Inside a pngquant PNG: 23 palette colors, 6 tRNS bytes, 7 alpha levels

Comments
5 min read
What an OCR accuracy report must state before you compare its numbers

What an OCR accuracy report must state before you compare its numbers

Comments
4 min read
Same matte, MSE 1.41 or 9.03: what a matting report has to state first

Same matte, MSE 1.41 or 9.03: what a matting report has to state first

Comments
4 min read
Where a not-worth-re-encoding check belongs in a video transcode pipeline

Where a not-worth-re-encoding check belongs in a video transcode pipeline

Comments
4 min read
Where to put a GIF size budget check in an asset pipeline

Where to put a GIF size budget check in an asset pipeline

Comments
4 min read
Where to put a pixel-budget check in a PDF rendering pipeline: after the viewport, before the canvas

Where to put a pixel-budget check in a PDF rendering pipeline: after the viewport, before the canvas

Comments
4 min read
Where to put an alpha-convention check in an image pipeline: at the resize call, not in the images

Where to put an alpha-convention check in an image pipeline: at the resize call, not in the images

Comments
4 min read
Where to put the lossless PNG pass in an asset pipeline: after the palette step, and last

Where to put the lossless PNG pass in an asset pipeline: after the palette step, and last

Comments
3 min read
A font-license checkpoint for PDF-to-HTML: what fsType can and can't tell you

A font-license checkpoint for PDF-to-HTML: what fsType can and can't tell you

Comments
4 min read
A transparent video went opaque with no error: put an alpha check after every hop

A transparent video went opaque with no error: put an alpha check after every hop

Comments
4 min read
A 25 MB video target came back at 14 MB, and the fix is checking how far under you landed

A 25 MB video target came back at 14 MB, and the fix is checking how far under you landed

Comments
4 min read
Table structure after OCR: why I'd rather get word boxes than tab-separated text

Table structure after OCR: why I'd rather get word boxes than tab-separated text

Comments
4 min read
Why PDF image extractors return the stored picture, not the one on the page

Why PDF image extractors return the stored picture, not the one on the page

Comments
4 min read
Why a JPEG white background isn't 255: blocks, quantization tables and chroma subsampling

Why a JPEG white background isn't 255: blocks, quantization tables and chroma subsampling

Comments
4 min read
Budgeting browser memory for a 300-photo batch: the ZIP step added 830 MiB at the very end

Budgeting browser memory for a 300-photo batch: the ZIP step added 830 MiB at the very end

Comments
4 min read
In a frame-diffed GIF, ordered dither cost 15% more bytes and Floyd–Steinberg cost 91%

In a frame-diffed GIF, ordered dither cost 15% more bytes and Floyd–Steinberg cost 91%

Comments
5 min read
The halo around a blurred video background is the dark-fringe bug from texture filtering

The halo around a blurred video background is the dark-fringe bug from texture filtering

Comments
4 min read
Falling back to WASM x265 cost 25 MB once and about 6x the encode time on every run

Falling back to WASM x265 cost 25 MB once and about 6x the encode time on every run

Comments
4 min read
I exported a long PDF page at 300 DPI and got 131: the WebP 16383 cap

I exported a long PDF page at 300 DPI and got 131: the WebP 16383 cap

Comments
3 min read
I compressed a signed PDF with five tools. Every signature broke or vanished

I compressed a signed PDF with five tools. Every signature broke or vanished

Comments
4 min read
Counting colors stops telling screenshots from photos after one JPEG save

Counting colors stops telling screenshots from photos after one JPEG save

Comments
4 min read
A Web Worker fixed the freeze, not the wait: two checks for in-browser AVIF encoding

A Web Worker fixed the freeze, not the wait: two checks for in-browser AVIF encoding

Comments
4 min read
The first fillText requests your web font and draws without it

The first fillText requests your web font and draws without it

Comments
4 min read
The frame queue emptied and the graphics buffers stayed: a WebCodecs reading

The frame queue emptied and the graphics buffers stayed: a WebCodecs reading

Comments
4 min read
WebCodecs gave me portrait video lying on its side in all four browsers I tested

WebCodecs gave me portrait video lying on its side in all four browsers I tested

Comments
4 min read
Word crops are non-destructive, and your DOCX pipeline inherits that

Word crops are non-destructive, and your DOCX pipeline inherits that

Comments
4 min read
How I carry CJK fonts into an HTML-to-PDF job I don't control

How I carry CJK fonts into an HTML-to-PDF job I don't control

Comments
6 min read
Split your OCR error rate in two before you touch preprocessing

Split your OCR error rate in two before you touch preprocessing

Comments
4 min read
What I check before a transparent PNG gets cut to 256 colours

What I check before a transparent PNG gets cut to 256 colours

Comments
5 min read
A PDF permission password is one integer, and the reader decides

A PDF permission password is one integer, and the reader decides

Comments
4 min read
My video capture test needed a pixel reference beyond currentTime

My video capture test needed a pixel reference beyond currentTime

Comments
3 min read
A moov box at byte 28 did not make this MP4 a faststart file

A moov box at byte 28 did not make this MP4 a faststart file

Comments
3 min read
A readable PDF page is not a text extraction test

A readable PDF page is not a text extraction test

Comments
4 min read
Moving canvas work into a Worker won't raise its size limit

Moving canvas work into a Worker won't raise its size limit

Comments 1
4 min read
Premultiplied alpha didn't remove my cutout's white fringe

Premultiplied alpha didn't remove my cutout's white fringe

Comments
3 min read
Progressive JPEG saved 4–6% of bytes and cost 2.5x decode time

Progressive JPEG saved 4–6% of bytes and cost 2.5x decode time

Comments
3 min read
JPEG quality 75 at 4:4:4 beat 0.99 at 4:2:0 on red text

JPEG quality 75 at 4:4:4 beat 0.99 at 4:2:0 on red text

Comments
5 min read
Same 200 KB, three formats: why WebP and AVIF look cleaner than JPEG

Same 200 KB, three formats: why WebP and AVIF look cleaner than JPEG

Comments
4 min read
In-browser AI cold start: the 6 MB ONNX Runtime WASM took longer than the 44 MB model

In-browser AI cold start: the 6 MB ONNX Runtime WASM took longer than the 44 MB model

Comments
3 min read
Base64-inlined WASM still costs 37% more after gzip

Base64-inlined WASM still costs 37% more after gzip

Comments
3 min read
Stop UA-sniffing canvas encoders: probe toBlob once at page load

Stop UA-sniffing canvas encoders: probe toBlob once at page load

Comments
4 min read
Your HTML-to-PDF export has zero selectable characters

Your HTML-to-PDF export has zero selectable characters

Comments
5 min read
A .docx does not store where its pages end

A .docx does not store where its pages end

Comments
4 min read
The knob I reached for first was the only one that hurt

The knob I reached for first was the only one that hurt

Comments
6 min read
Your WebGPU benchmark may be running on a CPU rasterizer

Your WebGPU benchmark may be running on a CPU rasterizer

Comments
4 min read
Your GIF exporter is fine — the container is the problem

Your GIF exporter is fine — the container is the problem

1
Comments
5 min read
A 5-page PDF, 8 image placements, 3 files: what an extractor should count

A 5-page PDF, 8 image placements, 3 files: what an extractor should count

Comments
5 min read
Effective DPI has two axes: what a PDF compressor actually removes

Effective DPI has two axes: what a PDF compressor actually removes

Comments
5 min read
The compression ratio looked great and the deliverable still failed

The compression ratio looked great and the deliverable still failed

Comments
6 min read
loading...