Website Image Downloader Benchmark

Allbirds homepage: larger source image test

For our first test, we scanned the Allbirds homepage, a popular Shopify shoe store, with all ten tools and compared the same two fixed product images. Three tools returned 4000×4000 versions of both images: Source Image Finder, extractpics.com, and Grabit. Six returned 1024×1024 versions. ToolNova did not return either fixed homepage target in the retained core test. Larger source images preserve more product detail, which can be useful when shoppers want a closer look at materials, texture, and construction.

The split panel below shows the difference between a 1024×1024 image and the 4000×4000 version returned in this test.

The same image at two sizes

15.26× as many pixels

1024×1024 downloaded file1,048,576 pixels

4000×4000 downloaded file16,000,000 pixels

↔

Detail view: drag the divider to compare the same matched area of both images. Use the links below to open the complete files.

Detail crop: showing 50% 1024×1024 file and 50% 4000×4000 file.

Open 1024×1024 test fileOpen 4000×4000 test file

Dimensions measured from the files · No AI upscaling or editing. The larger result is a public file that was already available from the website. The viewer uses an optimized WebP copy of the 4000×4000 file for page performance; the original PNG is linked above.

The 1024×1024 file shown in the split panel was returned identically by Website Image Xtractor, extract.pics, IMGDownloader, and Image Downloader by Pact Interactive. extractpics.com and Grabit are not shown on the smaller side because both also returned 4000×4000.

Allbirds homepage results

Test settings: Source Image Finder used Scan Single Page with Include Site UI off and Include all verified versions off. extractpics.com used Quick Extract. Grabit used its unchanged common/default extension settings: Prefer full size on, Original format, Medium speed, and a 60 px minimum detection size. The other tools used their normal single-page or default workflow.

Wix: Mighty Spice and Eor Glas

We used the same four public Wix URLs throughout this part of the benchmark: each store homepage and one product page from each site. The matched-family test compares Source Image Finder with extract.pics across 46 directly matched Wix image families. We also completed four-page downloads with eight other tools, including ToolNova and Grabit, so the fixed product-image comparison covers all ten tools.

46-family comparison with extract.pics

Across the 46 matched families, 45 had higher pixel dimensions in the Source Image Finder download and one tied. extract.pics was larger on none of the directly matched families in this test.

We left one matching image hosted by Unsplash out of the Wix store-asset count because it did not come from Wix. One Eor Glas image also came back from extract.pics with different proportions. We report its dimensions, but do not use a pixel-count ratio to compare the quality of that image.

Test URLs: Eor Glas homepage, Léviter product page, Mighty Spice homepage, and Butter Chicken product page.

Four-page Wix download evidence across all ten tools

This table includes all ten tools from the benchmark period. Source Image Finder and extract.pics use the same-URL matched-family record, which did not capture selected → downloaded totals in the same format, so those cells show directly matched families instead of inventing counts. For seven of the other tools, counts are shown as the interface download/selected count followed by the files present in the downloaded ZIP or folder. Grabit’s per-page interface counts were not retained, so its row reports downloaded image files only. Tools define found, shown, selected, and downloadable items differently, so these are observations rather than normalized failure rates.

* extractpics.com showed 80 selected on the Léviter page. The resulting ZIP contained 10 downloaded payload files, nine images and one TTF font, plus images.csv. The CSV is not included in the 10-file payload count.

† The retained Mighty Spice ToolNova archive also contained a nested prior-run folder unrelated to ToolNova. That folder was excluded; the ToolNova top-level payload contained 38 image files and 36 Forbidden text responses. ‡ On Eor Glas, the retained ToolNova payload contained 25 images + 39 Forbidden text responses on the homepage, and 11 images + 6 Forbidden text responses on Léviter.

For Source Image Finder and extract.pics, the retained evidence records directly matched image families rather than a normalized selected/downloaded total, so the table reports only what that evidence supports. GetAnyImage's manifest.json and README are excluded from the image counts above. Pact's folder counts include its SVG image file. Exact-duplicate counts are image-only and exclude CSV, JSON, README, manifest, and other support files.

Among the eight four-page runs with exact-duplicate checks, only Imageye contained byte-identical duplicate image files: four files forming two duplicate pairs, both on the Mighty Spice homepage. The retained Source Image Finder and extract.pics records did not measure exact duplicates in this same four-page count format.

File naming, organization, and download completeness

Finding the right image is only part of the job. We also compared what the downloaded files were like to work with afterward: whether filenames stayed recognizable, whether pages or products stayed organized, whether source information was preserved, and whether exact duplicate image files remained in the download.

For this downloaded-file analysis, we used the Allbirds homepage and Cruiser Canvas product page with the same ten tools shown above. Where multi-URL downloading was supported, we also used the same three-page set from Allbirds, Mighty Spice, and Eor Glas.

File naming and organization

Observed download counts and exact duplicate images

How to read the counts: tools use different definitions for found, shown, selected, filtered, and downloadable items. We report the counts exactly as observed and do not assume every difference means a failed download. An exact duplicate here means two downloaded image files with byte-for-byte identical contents. CSV, JSON, README, manifest, and other metadata/support files are excluded from duplicate-image counts. Different responsive sizes are not exact duplicates unless the downloaded image-file contents are identical.

Key findings

Scope: the retained record covers 14 services across four public websites and seven named webpage URLs. Not every service completed every test, so detailed tables compare only like-for-like results. These findings apply to the named pages and test dates, not every website or downloader.

Limitations and disclosure

This is first-party testing by Source Image Finder, not independent third-party testing.

Open the core benchmark record (JSON) → · Open the file-output record (JSON) → · Open the Wix 10-tool record (JSON) → · Open the ToolNova standardized record (JSON) → · Open the Grabit standardized record (JSON) →

The saved record includes services and observations that are not all shown in the main tables. We keep the public tables focused on results that can be compared directly.

The saved 1024×1024 and 4000×4000 Allbirds files on this page came from these tests. We used full competitor ZIPs for analysis but do not republish competitor archives.

Results apply to the named URLs, settings, files, and dates. Websites and downloaders can change, and the page given to a tool can matter. In the Allbirds test, several tools returned only 1024×1024 from the homepage but reached 4000×4000 for the primary image when given the product page directly.

Pixel dimensions are useful for comparing downloaded file size in pixels, but they do not by themselves prove better visual quality. We separately note whether a tool found the image, preserved the source format, returned many responsive variants, or completed the download.

This benchmark does not measure every image downloader, every website platform, authenticated/private pages, or every feature offered by the tested services.

Run the same single-page workflow

If you want to try the same single-page workflow yourself, paste a public webpage URL into Scan Single Page to review the image files Source Image Finder can verify. The free single-page workflow uses the same core extraction and source-recovery system tested in this report.