Rawpixel images contain watermarks

[AetherUnbound]

Description

It appears that all Rawpixel images (both in thumbnails and the upstream URL) contain watermarks for Rawpixel. This was initially reported to us by a user who was uncertain whether or not they would be able to use the Rawpixel images on their website (even with attribution), since watermarked images typically convey that the image is a sample and must be acquired through other means. The results (for now) can be seen here: https://staging.openverse.org/image/source/rawpixel/ - all of these images have visible watermarks on them.

I've verified that when trying to use the Gutenberg integration to insert images, the watermarked image is the one that gets inserted. Here's an example: https://openverse.org/en-gb/image/0f65b7e2-a23b-4375-b0fe-598b7f659ff4. When attempting to download this image from the Rawpixel website, a login is required. I've verified that after creating an account and downloading the image, the downloaded image does not have the watermark.

To me, this feels very similar to the discussion we had in #3499, the only difference being that this applies to the entire provider and not a subset of images. Particularly as it relates to using integrations, this creates a very frustrating experience when trying to insert media (I had to try several times in order to get a suitable image for this Make WP post). I wonder if we might want to take a similar approach with Rawpixel:

• Deprioritize results in the index
• Add an indexed field (something like "Login required for download"?) which can be filtered
• Recommend removing Rawpixel from the results that appear in the Gutenberg integration

@WordPress/openverse-maintainers - anyone have additional thoughts here? Any action we should take now that I might be missing?

Additional context

Related to #3531 as well

[sarayourfriend]

Add an indexed field (something like "Login required for download"?) which can be filtered

It's actually very similar to freesound #1676, in that it requires logging in to get the "full" version.

It's a shame that the thumbnail has the watermark. Just to confirm, there's no thumbnail URL on their end that returns a compressed/resized version without the watermark? Seems like probably not, considering they put the watermark even on their own page.

It would be a bummer to lose RawPixel or push them down in the results, the images on RawPixel are rather high quality.

I don't really see why they would do this for CC licenced images other than to prevent hot-linking the thumbnail? What a shame.

[fcoveram]

It's a shame that the thumbnail has the watermark. Just to confirm, there's no thumbnail URL on their end that returns a compressed/resized version without the watermark? Seems like probably not, considering they put the watermark even on their own page.

I was thinking exactly this. And also agree with @sarayourfriend that pushing them down doesn't sound like the best solution.

The logging wall is definitely a pain point that can cause frustration. I was already logged in in Rawpixel and didn't notice the breakdown. I'm in favor of creating a ticket to address that UX flow as it might be happening with other existing sources and with the upcoming ones.

[krysal]

There are alternative links that don't have the watermark, namely pinterestImage, google_teaser and google_teaser_transparent.

For example, for this image, we would use this direct link (currently style_uri with image_1300 as the size):

https://images.rawpixel.com/image_1300/cHJpdmF0ZS9sci9pbWFnZXMvd2Vic2l0ZS8yMDIyLTA1L3Vwd2s2MTY3MDIxNi13aWtpbWVkaWEtaW1hZ2Utam9iNTcyLTEuanBn.jpg

But we could use one of these instead:

{
"pinterestImage": "https://images.rawpixel.com/image_social_landscape/cHJpdmF0ZS9sci9pbWFnZXMvd2Vic2l0ZS8yMDIyLTA1L3Vwd2s2MTY3MDIxNi13aWtpbWVkaWEtaW1hZ2Utam9iNTcyLTEuanBn.jpg",
"google_teaser": "https://images.rawpixel.com/image_800/cHJpdmF0ZS9sci9pbWFnZXMvd2Vic2l0ZS8yMDIyLTA1L3Vwd2s2MTY3MDIxNi13aWtpbWVkaWEtaW1hZ2Utam9iNTcyLTEuanBn.jpg",
"google_teaser_transparent": "https://images.rawpixel.com/image_800/cHJpdmF0ZS9sci9pbWFnZXMvd2Vic2l0ZS8yMDIyLTA1L3Vwd2s2MTY3MDIxNi13aWtpbWVkaWEtaW1hZ2Utam9iNTcyLTEuanBn.jpg"
}

I can't say what differences rawpixel applies for each platform, but in this case, the image for Pinterest is smaller than the one we use and a bit bigger than the links for Google. I think we can go with that one as replacement.

[sarayourfriend]

Nice find, Krystle! That sounds like the perfect solution.

[fcoveram]

Great solution Krystle 👏

[rawpixel-vincent]

Hi,

Sorry for the trouble occasioned by this update, for the sake of not adding complexity to our system we couldn't treat public domain image differently than others when applying the watermark, we still have in mind to find a solution for this and we will update you if anything changes.

as an alternative to the image_1300 and social_landscape preset you can use the editor_1024 preset which should give you something closer to the 1300 preset when stretched because it has higher quality settings and no watermark.

you can replace the {} substring in the value of style_uri property with editor_1024

https://images.rawpixel.com/editor_1024/cHJpdmF0ZS9sci9pbWFnZXMvd2Vic2l0ZS8yMDIyLTA1L3Vwd2s2MTY3MDIxNi13aWtpbWVkaWEtaW1hZ2Utam9iNTcyLTEuanBn.jpg

https://images.rawpixel.com/image_social_landscape/cHJpdmF0ZS9sci9pbWFnZXMvd2Vic2l0ZS8yMDIyLTA1L3Vwd2s2MTY3MDIxNi13aWtpbWVkaWEtaW1hZ2Utam9iNTcyLTEuanBn.jpg

you can @ me in the future if you have any issue with the rawpixel feed 🙏

Edit: note that editor_1024 is serving webp format where social_landscape is jpeg