Attached Images - Poor Quality

What’s the site doing to the uploaded images?

i (sometimes) try to practice good etiquette when embedding images at a forum (resize them smaller, compress jpgs to around 100kb or so, etc…) while still keeping them looking ok…

but here, it seems as if the forum is doing another round of compression on them but not really doing too good of a job at it…

for instance, i attached a 1000x545 jpg at 78kb earlier today… the size listed once it’s embedded is 75kb… so 3kb of space has been saved but the image quality has deteriorated very noticeably… (even upon expanding it to full size)

i guess my question is- what is the best way to upload images so they still appear crisp in the forum? because as is now, it seems like it’s better to upload full size/full quality images in order to combat the site’s image algorithm(?)… but i’d rather not go that route if at all avoidable.


@zogstrip I think this calls for a new site setting

min_saving_percent_to_recompress_images … or something along those lines.

Only run the optimiser if you see you are saving more than say 5% out-of-the-box


@sam that does not make sense, the images simply do not fit in around 700px of max width. So there is a thumbnail when clicked shows the actual image.

If you do not want the image to be thumb nailed make sure the max size is less than the post width.

But does it matter when you can click or tap on the thumbnail to see full size any time?

1 Like

yeah, the quality degrades even when viewing it at full size…

here’s the image i was talking about earlier:

then the same image when uploaded here:

it gets extra muddy here. (notice around the pink lines for obvious example but to me at least, it looks bad all around… especially for 3kb savings)

[EDIT] hmm… i can’t expand that but the version showing on the site right now is 20kb which is more inline with the image quality degredation instead of the 3kb drop before… i guess you guys are working on it right now… i’ll wait til later when you say “ok… try it out now” :smile:

When you upload an image, Discourse will check its dimensions. If it’s larger than 690x500, it will generate a thumbnail that will fit in these dimensions.

No compression whatsoever is done to either of these images.

@jeff_hammond I see no difference between the original and the one that you uploaded. In fact, these files are the same.

But, I do agree that the generated thumbnail does not look that great. We’re using image magick to generate the thumbnail, maybe there’s a setting that I could use to make the thumbnail look better, any idea?

hmm… really?
i guess it’s subjective but if you don’t see a problem with it then, ok-- you’re the boss :wink:

another example of one i posted last night… i made it 600px wide @33KB… you can’t even see the grid lines-- each of the muddy grid lines which are visible have 5 thinner lines in between… i can see them fine via my drive but once embedded here, they’re gone.

but if i’m the only one seeing a problem, i’ll just put a discourse folder in my dropbox and link from there… no biggie.
thanks for looking into it though.

this is the same pic via dropbox… i’ve already compressed to what i personally consider the threshold of decent looking… the grid lines are still visible at least…

and anyway, if i’m using my own hoster, i’ll feel less guilty about using less compression… the pictures will look better (to me) and it’s less bandwidth for you.
win-win :smile:

Well… your images won’t be visible for people behind corporate filters that deny access to anything remotely associated with social internet. But you can’t win them all… :wink:

I’m not trying to be bossy here. I just meant that I computed the SHA of both images and they are identical. This means that the files are exactly the same binary-wise. If there’s no difference in the content of the files, I don’t see why there would be any difference visually.

heh, i don’t know what SHA even means :wink:
(but in the context of your post, i think i got it figured out)

do you see a difference with the last two examples? there’s a visible grid on the dropbox and it’s not there on the discourse file…

[edit] or is this something that could be happening because of something on my end? (browser etc.) i wouldn’t think so but i don’t truly know. [/edit]

you got that right.

@jeff_hammond I must be boss too, or should I say my computer, because there are no difference between the images:

I copied each to photoshop and sat the top layer to difference and the result is completely black: (no difference)

Can you make a screencapture and show us what you are seeing?


1 Like

I’m sorry. I used the term SHA to refer to the hash value (kind of a signature) of both images. That value is computed by the SHA algorithm.

cf. @Holo’s reply for the designer-friendly explanation :slight_smile:

I actually do not. I see the grid on both images just fine.


the screenshot itself will add some blurriness but you should still be able to make out the difference i’m seeing…

ah… ok. i’m defintiely seeing a difference… hopefully the screenshot make it more apparent and i’ll look like less of a fool :slight_smile:

if i do the same thing, it does show the differences:

(it’s possible that looks black if your display gamma is lower than mine… if so, brighten the image and you should see it)

fwiw, i’ve tried it on two computers-- a macPro with osx10.7.5 and a macbookPro running mavericks… tried safari and chrome on both… the same problem happens in all 4 of those scenarios…

if i do the comparison on an iPhone, the difference between the dropbox version and discourse version becomes even worse… so, if you have an iPhone, look at this thread and compare those two pictures on it… the dropbox pic remains clean and the discourse one becomes horrible. (as in- it really is horrible)

I see the grid on both images

i guess this is something to do with macs then?.. (or, that’s the only difference i can gather so far as i’m assuming you all are on windows)…

but either A) someone is going to chime in eventually and say “yeah, i see what you’re talking about” or B) i’ll be branded the weirdo with wacky vision :wink:

[EDIT] oh… i should add- it’s not just my images that look bad… it’s everybody’s

I see what you are talking about!
But not on my computer, only on the last screenshot that compared them and on the “difference” image. I am running windows 7 on a macbook pro, so the screen is the same.

And this doesn’t make sense at all to me. Why would your system choose to do so? Is NSA running all your datastreams through an optimizer? :wink:

Ps, here is what I saw in the top of this thread.

dropbox vs discourse (273.5 KB)

1 Like

ha… ok. good (phew)

SHA comparison (i don’t know what this means but whatever… as a test, i did locally duplicate an image so i had test1.jpg and test2.jpg… they checked out the same)

(the lines starting with ‘ea4d…’ and ‘45ab…’ are the returns i get when using the shasum command in terminal)

ahh… ok. yes, those do look identical… so i guess now the question is why is this working right for everybody else? i feel so left out. :smile:

Good luck on finding it out, there is nothing worse than having a high fidelity system that refuses to be hi-fi.
(That’s why my FirePro is still collecting dust, and I bought a quadro instead, I got so distracted by the bad AA of the FirePro)

it’s not in my system(s) though… the files are actually different after i download them (i.e.- the SHA comparison which zogstrip was talking about-- he downloads both versions and they’re the same… i do it and they’re different)…
so it seems as if i’m being fed ‘bad data’

the only thing i can guess right now is that the internet i’m on is phone tethered and sites generally give me low quality stuff when applicable (but i wouldn’t guess this is one of those scenarios but maybe? discourse does it but dropbox doesn’t?)… for example, when i watch youtube with this type of connection, it feeds me 240p unless i manually go in and change the settings… i’ll be at a cable connection a little later today so i’ll try viewing the images then.

1 Like