It has two releases (well 4, but those aren’t relevant for this example), a standard version released on most platforms, and a hi-res version released on Apple Music, Tidal, and Qobuz.
The maximum resolution artwork available on the standard version is 1400, since Apple Music doesn’t have the standard version. But the hi-res version has the full 3000 version, from Apple Music.
Should the 3000x artwork be uploaded to the standard version, since it’s identical but technically is a different release?
Sorry for the direct reply, and not sure if this should go in a separate topic but I was looking for some advice, in Edit #149877995 - MusicBrainz and… many of this artist’s releases I indeed did not follow the advice given on this thread, but rather of having an issue with a bigger/smaller image, what are the guidelines on DSP provided images that are simply riddled with artifacts? Edit #149726522 - MusicBrainz Is a very similar example of an edit like this.
My most recent example (and most likely mistake) is https://musicbrainz.org/edit/150914086
There are no good sources for this release, everything has artifacts, but… there are okay source pieces, so I thought maybe combining them into one better one was an acceptable solution, however it would not be 1:1 with the true sources just due to the nature of attempting to do this manually.
*That and the fact it can’t be directly linked to any DSP or provider for the release (obviously…)
I would not replace these, instead I would, as mentioned in the other examples, set the non-artifacted cover as the release group cover.
I don’t see the benefit in duplicating the perfect image across every release. I’d even say that’s data-loss (even if the data that is lost is - accurately - representing an objectively inferior source!)
I should note that I personally would not consider cover artifacts on their own a good enough reason to split releases in the first place. But if the releases are already separate, why not store the cover differences, imo.
I see, so adding the properly linked cover art for each release is what I’ll do next, then just change the release group cover for the “best” one.
If you were also looking at Edit #150914086 - MusicBrainz should I cancel it as per your advice? In this case there is just one release but it most likely constitutes as fan art, correct? Just trying to not waste effort with such exercises.
I do a lot of photography, so I care a lot about embedded ICC color profiles in my images. Do you know who doesn’t care about embedded ICC color profiles? Apple Music! They strip out embedded metadata (as far as I currently know, I haven’t done a lot of research on this (yet).) Do you know who doesn’t? Bandcamp! Bandcamp will give you the raw pristine image data exactly as the artist uploaded it. Bandcamp is the best!
For example, Bandcamp accidentally supports animated GIFs, but you will never see them on the standard web interface because all of the scaled thumbnails are converted to JPG from whatever the original format, like GIF, was. (Except for the banner image, which is probably a PNG to work better as a clickable image map (which is a feature of Bandcamp).) There are a couple of artists using an animated GIF for their profile, but you must download the “original size” version of the profile image to get the GIF. Bandcamp will set the “content-type” header correctly to “image/gif” and it doesn’t care whether you use “.png” or “.jpg” in your initial request.
So… resolution is not the only criteria for selecting which cover art provider to use.
Deezer is always scaled, recompressed, and has every bit of metadata stripped out, even the “last-modified” header is missing. There is no “original version” you can get from Deezer. (As far as I currently know.)
I’m still researching this (just started a few days ago), but Tidal appears to also give you the original bits exactly as uploaded, just like Bandcamp does, but I might be misinterpreting something and I want to take a closer look before I’m confident in making this assetion.
I follow the exact words of @aerozol . Release artwork should always be as close to the real thing as possible. Never mocked up. Never faked from merging multiple versions. Artwork is about identification of a source and NOT about tagging. It is never about tagging.
One of the most horrendous things is seeing someone having uploaded the same image to every release in a release group. It is so obviously clearly wrong. Especially bad when a digital image appears on a CD or Vinyl.
Too often you see an editor do this and totally fail to notice different scaling and placing of part of the artwork. Title positions and fonts often change between CD and Digital Versions. Shapes change.
MB is not about the “best” art, it is about accurate artwork. If you want to make up prettified artwork then join fanart.tv or theaudiodb as they create those “best image possible” versions of artwork. (And Picard has add-ons that let you use them as tagging images)
@foxgrrl - interesting stuff. And why does it not surprise me that the biggest shops make the biggest changes.
Physical releases would have to have “data-loss” cover art since the general public doesn’t have access to the original data used to manufacture the physical release before a random person scans it.
You say we should require lossy compression data loss for digital releases when lossy compression artifacts are obviously not artist intent. How does the same image with less compression artifacts provided by the artist not fall into some kind of artist intent?
Why should error correction and artist intent apply to the text of third-party store standardization errors but not to the cover art with undesired compression errors?
Taken literally, no digital media releases could be merged because you are trying to use exact pixels of lossy image formats of various resolution and quality to say the cover art no longer matches.
You are arguing for the sake of arguing and know that is not what he is saying.
We make scans or photographs of the artwork in our hands in front of us. Or borrow images from a source like Discogs who are working from a human scanning\photographing an object. We make a “best image” possible of that source material. The source material being the physical paperwork in the hands that one purchases from a store.
What he is saying is “don’t fake it”. Don’t take an image degraded by a shop and then guess what it should look like by editing it in photoshop. Do a best effort of getting the best image from the best store, but don’t fake it.
If “artist intent” has been to put their artwork on a store that then degrades their image, then that is the image we all get to see. If spotify has degraded an image, then that is literally what all spotify users see. And therefore that is what MB should represent. If deezer has better art, then choose that instead. If Bandcamp has the best artwork then go that way. Just don’t upload edited art.
When Spotify takes a lossless quality WAV of an artist and then puts that out as a compressed MP3 audio stream, then that is what the store is representing of the artist. We can only represent what we the customer can purchase.
This is very interesting, often I do find Apple Music and Qobuz provide very similar if not the same images, but I didn’t look much further than artifacts.
Will be keeping an eye on your findings if there ends up being a topic about this, learning how to check for the correct ICC at some point would be awesome as well.
I just had a great idea for a GreaseMonkey UserScript that does a HEAD on document.getElementsByClassName("band-photo").item('img.band-photo').src.split("_")[0]+"_0.jpg" and if there’s a Content-Type: image/gif header, then actually GET the the “original size” image and replace the small img.band-photo JPG with the animated GIF.
Slight tangent, but I just remembered another thing to watch out for with regards to the image dimmcensions of cover art. Some online music services will upscale a small low-resolution (original source) image up to whatever size you request (like SoundCloud) if you’re not careful about what you’re asking for. I think Bandcamp will actually do this too if, for example, an artist uploaded a 345x432px image, Bandcamp will scale and crop that to generate the 700x700 and 1200x1200 pixel “thumbnails” on the album page and embedded cover art in the FLAC and MP3 tags when you download the audio.
Additionally… I remember a few years ago, I discovered that the 3000x3000 pixel JPG image that came from the iTunes Store (I guess it would be Apple Music now) was actually a scaled up copy of the [I forget exactly but it was small like] 1400x1400 pixel image I got from somewhere else. The 1400x1400 px image was actually sharper and “higher resolution” because the 3000x3000 one had every pixel quadrupled in area and then blurred to try to hide the jagged stair-step artifacts on any diagonal high-contrast edge between two colors.
I assume the record label or artist did the upscaling to 3000x3000 and that it wasn’t Apple… I hope… I don’t actually know this for certain.
I’ve also seen “official from the record label” cover art that someone employed by the record label itself and acting in an official capacity for the record label had obviously just downloaded from Discogs.
(Some of these record labels are literally just a single human, and sometimes just a single artist themselves.) Most online music stores have a process for getting recognized by them as a music distributor/label, and once they believe that you’re legit, you can upload anything you want for sale to their online store. If you’re surprised by the unexpectedly low quality of cover art, and metadata, and ISRC usage, and … etc. the barrier for entry is extremely low, and the online music store itself doesn’t care about anything other than making money and not getting sued.