Four JPGs from one image: what makes each one different?
See the implemented filenames, dimensions, and quality calculations for the 1600px social file, 2200px portfolio file, watermarked sample, and 600px thumbnail.
Read the output specifications →
DROWPACK KNOWLEDGE BASE
Tested in practice,
Tested in practice,
verified with numbers
Instead of repeating feature names, we explain which files DrowPack creates, which settings affect the result, and where automatic conversion reaches its limits. Every article is edited directly from the tool's actual code specifications and conversion examples in the project repository.
Start with the problem you need to solve
Different outputs call for different formats and settings. The topics are separated so you can go straight to the answer you need.
An original-quality SVG is not the same as a pure vector SVG
Both carry the SVG extension, but their internal structures, scaling behavior, and editability are entirely different. Compare the criteria for photos, illustrations, and logos.
See the format decision guide →When should you change threshold, color count, or detail?
We publish DrowPack's line-art, logo, silhouette, simple-color, photo, and illustration preset values and explain the order in which to adjust them.
Read the settings guide →Why can an edge look jagged even with 60,114 paths?
We directly analyzed a 10.63MB auto-traced SVG, including its path count, half-pixel coordinates, tiny fragments, and before-and-after cleanup renders.
View the measured case study →What data goes where when you run AI editing?
Understand the boundary between browser processing and AI transfer, the 511px input limit, the current provider, review requirements, and responsible-use standards.
Read about the AI data flow →Once you understand the criteria, compare the results yourself
Your source image stays in the browser for standard conversions. Start with a fast preset, compare the preview and file size, and raise only the settings that need it.
Open DrowPack →How we write these guides
We focus on leaving the evidence users need to judge a result, rather than padding pages with generic search phrases.
Actual specificationsWe match filenames, long-edge limits, and preset values to the code the tool currently runs.
ReproducibleInputs, adjustment order, and inspection points are documented so the same conditions can be tested again.
Limits disclosedWe do not hide cases where auto-tracing cannot replace manual Bezier work or where AI results are uncertain.
Revision historyWhen a feature or provider changes, we review the article date and the related specifications together.