Covermakers Chat Thread

[msfry]

A lot of our images are fuzzy on the Youtube platform, but the one I remember offhand is one of my own: Comparing archive to youtube:

That one on YouTube, the image is zoomed in. The title is cut off, and the girl is even barely all in the image. If it were regular size, it would probably look fine.

The image I should have linked to (and thought I had) for comparison is when you click on the rectangle and it brings you to this blown up square:
https://www.youtube.com/watch?v=FcZS1KkHd1s&t=13106s
Compared to the original, it is fuzzy.
https://1drv.ms/i/s!AqmYAdtBQdx8ittzQ0eBWm16rgUv7w?e=RunXw4

I should have kept us focused on this square image for this discussion. The rectangular display that cuts off half the cover is another issue entirely.

All I'm trying to do is see if we can get our 600 x 600 images to BengjW somehow -- the ones we all already have sitting on our computers.

I don't understand Anne's problem with this easy solution, which sits at our fingertips and would make our work look a tad or two more professional, i.e., make example_itemimage.jpg the 600 x 600 size rather than a duplicate of the example.jpg image. It's not like it would cost us anything in time or money but to rename one file. As to server space, 4 times our wee little footprint in one piece of cover art is almost nothing, and in terms of Internet Archive's 145 Petabytes of storage space, uploading/displaying our 600 x 600 covers there really is "nothing". Besides, IA is well funded by grants and donations from far wealthier entities than ourselves, all with philanthropy at heart. I have absolutely no fear that this enhancement would contribute to carbon emissions a la Bitcoin farms or the Airlines industry.
https://archive.org/about/

[msfry]

Many Youtube makers just turn out rectangular images into squares - and if they care so little they are unlikely to do anything different, it will not increase their earnings. I always try to make our ad hard to remove just so we at least get some credit
Anne
We are discussing BengtW's channel in this thread aren't we ?

Yes we are discussing BengtW's LV Youtube channel, which displays our artwork.

What does this mean "I always try to make our ad hard to remove". What ad?

"Other Youtube makers" who pirate our stuff don't use the Librivox name on their channel, and while BengtW's channel is not actually affiliated with Librivox, that is not easily discernable by the public. We are very fortunate that he is willing to work with us on improving its appearance.

[quartertone]

If cover-makers are making art that needs that resolution, they ought to start their own online galleries or something!

[...]

Technical limitations are one thing, but we have generously been given this space for the purpose of public domain audio, and there ought to be no doubt that's what we're using it for, whether we also donate money to Archive's broader mission or not.

Yes! This I agree with. I have my own private site where I host my cover images and audio files. Mostly to keep my own archive copies. A drawback of this is that there may be barriers to achieving this for some people, including server cost, moderate tech knowlege requierement, interest in the project, and uncertain longevity/legacy, etc.

That kind of work should be seen, but LV's catalog site isn't the place for showing it off. Neither, I would guess, are most YouTube channels.

I also (sort of) agree that LV/IA perhaps may not be the best place to showcase high-res digital art. But Youtube channels are "privately" managed (ie, outside of LV's sphere of influence/jurisdiction), so I opine that they could do whatever they want within the terms of service and laws etc.

Librivox (the audio, images, texts, etc) are all Public Domain, so really there is nothing stopping anyone from "stealing"/"pirating" our work, or using it "with permission"/"without permission" in whatever way they prefer. Anyone could, for example sell a print collection of book covers to make a profit with no reference to LibriVox. Such is Public Domain. It's not a nice thing to do, but it's allowed as far as I understand. (right?)

It seems there may not be (or there is resistance to) a way to automate high-res upload to Youtube, so maybe a workaround might be to send BengtW the high res images directly? This would involve extra work on his end though, and probably not sustainable in the long term.

[msfry]

LV's catalog site isn't the place for showing it off.

We aren't talking about LV's catalog site for displaying 600 x 600 jpgs. We are talking about supplying BengtW with those images for his Youtube channel, by changing the _itemimage file to the higher resolution and posting it to IA which will readily accept it. Anne has already revealed that the _itemimage file at 300x 300, which is a duplicate of another image size, is not used anywhere.

For the record, I have no interest in displaying my covers on my own website for 10 people to admire, when there are thousands of eyes on our audiobooks on LV, IA, and LV YouTube. Every book and magazine publisher knows the value of a great cover, or they would not invest in it. That goes without saying. Decorating LV's audiobooks to attract attention, is the only reason I bother to make these covers.

So having made my case, multiple times, for decorating LV the best we can, easily, with assets already in hand, I have run out of things to add to this topic.

[redrun]

Sorry for adding to the confusion: that phrase was referring to 1080p images that are essentially high-resolution desktop wallpapers -a completely different thing from the decorating you proposed!

I don't have any first-hand knowledge of how _itemimage is or isn't used. I agree you've thoroughly laid out your case if that's so, but I'll have to leave it at that as well.

[carthinius]

I have tired these tools already and it does not work. Many extracts includes parts outside the cover or splits into multiple images. Sometime its rotated. As the PDF files are not following the strict format it cannot be use reliably. I use them to find the cover artist but this is again a best effort using OCR methods. Mistakes in cover artist can be manually updated in the description if needed. For the cover I need something more reliable as a mistake will encode the video incorrectly which cannot be corrected manually without a complete re encode.

Just to be sure: Your tools would check if the provided file (_itemimage) fits the necessary pixels and/or ratio and if not, use the image to scale it up to 1080 x 1080 px (or 720 x 720 px, as many thumbs aren't HD)?
So if there would be an uploaded image in 16:9 ratio with at least 720 px in height, the tools would use it? And if not, still proceed as before?

[annise]

I think I didn't make myself clear when I said Archive didn't use _itemimage files, I just meant they no longer showed the pdfs as their first choice image. Nowadays certainly for us, they displays all .jpg images.
So bigger images do affect the display on the IA page and increase our use of their resources, which makes page-opening slower at their busy moments, probably always but it is more obvious then.

Anne

[BengtW]

Just to be sure: Your tools would check if the provided file (_itemimage) fits the necessary pixels and/or ratio and if not, use the image to scale it up to 1080 x 1080 px (or 720 x 720 px, as many thumbs aren't HD)?
So if there would be an uploaded image in 16:9 ratio with at least 720 px in height, the tools would use it? And if not, still proceed as before?

No it will not check the ratio as all covers are squares at the moment. I will just scale it up and assume its a 1:1 ratio. Encoding in 1:1 ratio is not an issue. Youtube adds the empty borders itself making it a non issue. I think the current ratio is fine and I think the main improvement of a higher resolution is to generate nicer thumbnails.

Current I do not add thumbnails resulting in Youtube picking a random frame as thumbnail, since they all look the same makes little difference. I am working on a thumbnail template in 1280x720 that will use 600x600 for the cover, have the librivox logo and add some information like language, run time, part information and if solo or group project. An example (not real data):

[annise]

That is a very good plan - looks good and is "good for purpose". Anne

[msfry]

I am working on a thumbnail template in 1280x720 that will use 600x600 for the cover, have the librivox logo and add some information like language, run time, part information and if solo or group project. An example (not real data):

By thumbnails I suppose you mean the rectangle Youtube generates to display its videos, not the 150 x 150 thumbnails we produce for use in our catalog list. Looks good. My only question is how will you get the 600 x 600 image if we don't upload ours somewhere?

[BengtW]

By thumbnails I suppose you mean the rectangle Youtube generates to display its videos, not the 150 x 150 thumbnails we produce for use in our catalog list. Looks good. My only question is how will you get the 600 x 600 image if we don't upload ours somewhere?

Yes, Youtube allows a 1280x720 image to be used as video thumbnail. I will use the the same source as for the video. So either the "itemimage" if available or I upscale the 300x300 to 600x600. I might look deeper into the PDF extraction also. Thumbnails can be updated without recoding anything so I can sort out some mistakes afterwards.

[annise]

That sounds good - If you could tell us what projects the pdf caused you trouble when extracting we may be able to stop suggesting that particular template or something.
We will not be changing the size of the images but I think if you use the idea you showed us above the 300X300 would look good. Much better than the one YouTube provides

On the occasions I have needed to harvest an image from an existing PDF it usually turns out about 571X571.

Anne

[msfry]

So, I'm finishing up a cover now. Do we still need to make the 300 x 300_itemimage file?

[TriciaG]

For now, I'd say yes, since BengtW is still using that one.

[BengtW]

Ok so I spent some time with the PDF image extraction and after experimenting with about 1000 different PDF's I have made several important observations to help me use them in the cover process. Images from PDF have a kind of sizes and contains multiple images in many cases. Some PDF's has the text as overlay over the image making the image extract lose that information. To make a rule that I think consistently creates a good cover candidate I have done the following:

1. Download PDF and extract images.
2. Use the first extracted image from the PDF if "Y resolution is at least 600 pixels AND Y and X resolution is the same"
3. If the PDF images is disqualified due to above requirement try download "_itemimage" cover and use if it exists.
4. If no _itemimage file is found use the regular cover.

Now a cover is selected that can have theoretically any size which could lead to all kind of issues so I convert to one of two possible sizes that I know works well with YouTube.

• If the cover Y resolution is less than 600 pixels, the image is resized to a 600x600 image.
• Else (the cover Y resolution is equal or greater than 600 pixels) the image is resized to a 1200x1200 image.

I will try this starting today and see how it works out.