The one from Twitter has a higher resolution, but the one from Bluesky has a larger file size. Visually, they look the same

https://pbs.twimg.com/media/GqHm-8Jb0AARO5Z?format=jpg&name=large

https://inkcap.us-east.host.bsky.network/xrpc/com.atproto.sync.getBlob?did=did:plc:wjowrrey6d5tnnihomdw7dkt&cid=bafkreifcvlbymnarls6vyroqffiydfxzxr4bakqo7eq3sswcyfl5rzsehi

Twitter image is 2048x1799, larger than the BSky one which is 2000x1756.

generally I'd go with the larger res if they're both jpegs

dba_afish said:
Twitter image is 2048x1799, larger than the BSky one which is 2000x1756.

Yeah, but the file from Bluesky is 25% larger, so I'm uncertain

(And I used that script that finds the best version on BlueSky)

The Bluesky version is marginally higher quality in this case. The compression artifact patterning is overall more severe in the Twitter version, despite the minor resolution gain. I would replace the Twitter version with the Bluesky version for the quality gain.

Extra info:

1) Most Bluesky images at around 1999/2000x in one or both dimensions will be superior to the Twitter version at 2047/2048x dimensions or below. The reason I include 1999 and 2047 is because these services sometimes use a bound one pixel smaller for more accurate downscaling. Twitter's max size of 4095/4096x is also affected by this.

2) Some artists reupload posts from one social media account to another, recompressing the image. This is thankfully not the case here. You can tell because the Bluesky version has a different pattern of artifacts than the Twitter version when zooming in on both. If the pattern looks identical, then it's typically the artist copying the Twitter version to Bluesky.

3) When sourcing from Twitter, make sure you are using the "orig" parameter and not a sample size like "large". In this and most other cases, the images will be identical, but it's better to be safe by getting what is guaranteed to be the best version Twitter hosts. orig also guarantees that you are not artificially converting a JPEG to PNG, which only increases file size with no visual gain.
A userscript to reformat the URL to always get the Twitter orig version is available here: https://e621.net/forum_topics/39960.
A userscript for getting the better Bluesky image and post permalinks is available here: https://e621.net/forum_topics/56919

4) Some extra useful search terms for finding potentially inferior versions:

- Posts that usually have the source twitter.com or twimg

- Posts that have the source x.com. A post may be missing the twimg direct image link, and this catches those

- Posts that are within the common Twitter boundary where the Bluesky version will generally be better
- Combine with the above for posts specifically from Twitter

- Posts that are specifically at the Twitter boundary, either due to being the full orig size or being large sample sizes
- .. is range syntax for numbers, saying "return results from range x to y"

~width:679..680 ~height:679..680 width:<681 height:<681
~width:899..900 ~height:899..900 width:<901 height:<901
~width:1199..1200 ~height:1199..1200 width:<1201 height:<1201

- Other common Twitter sample size resolutions. The top one will get the most true positives. The bottom one will get the most false negatives
- As stated above, pair with source:*twi or source:*x.com for posts specifically from Twitter

Hopefully this longer breakdown was helpful for you! If you have any questions, always feel free to ask.
tl,dr: the answer to your question is on the first line.

song said:
The Bluesky version is marginally higher quality in this case. The compression artifact patterning is overall more severe in the Twitter version, despite the minor resolution gain. I would replace the Twitter version with the Bluesky version for the quality gain.

Extra info:

1) Most Bluesky images at around 1999/2000x in one or both dimensions will be superior to the Twitter version at 2047/2048x dimensions or below. The reason I include 1999 and 2047 is because these services sometimes use a bound one pixel smaller for more accurate downscaling. Twitter's max size of 4095/4096x is also affected by this.

2) Some artists reupload posts from one social media account to another, recompressing the image. This is thankfully not the case here. You can tell because the Bluesky version has a different pattern of artifacts than the Twitter version when zooming in on both. If the pattern looks identical, then it's typically the artist copying the Twitter version to Bluesky.

3) When sourcing from Twitter, make sure you are using the "orig" parameter and not a sample size like "large". In this and most other cases, the images will be identical, but it's better to be safe by getting what is guaranteed to be the best version Twitter hosts. orig also guarantees that you are not artificially converting a JPEG to PNG, which only increases file size with no visual gain.
A userscript to reformat the URL to always get the Twitter orig version is available here: https://e621.net/forum_topics/39960.
A userscript for getting the better Bluesky image and post permalinks is available here: https://e621.net/forum_topics/56919

4) Some extra useful search terms for finding potentially inferior versions:

source:*twi

- Posts that usually have the source twitter.com or twimg

source:*x.com

- Posts that have the source x.com. A post may be missing the twimg direct image link, and this catches those

width:<2049 height:<2049

- Posts that are within the common Twitter boundary where the Bluesky version will generally be better
- Combine with the above for posts specifically from Twitter

~width:2047..2048 ~height:2047..2048 width:<2049 height:<2049

- Posts that are specifically at the Twitter boundary, either due to being the full orig size or being large sample sizes
- .. is range syntax for numbers, saying "return results from range x to y"

~width:679..680 ~height:679..680 width:<681 height:<681
~width:899..900 ~height:899..900 width:<901 height:<901
~width:1199..1200 ~height:1199..1200 width:<1201 height:<1201

- Other common Twitter sample size resolutions. The top one will get the most true positives. The bottom one will get the most false negatives
- As stated above, pair with source:*twi or source:*x.com for posts specifically from Twitter

Hopefully this longer breakdown was helpful for you! If you have any questions, always feel free to ask.
tl,dr: the answer to your question is on the first line.

Thank you!

I also noticed that the picture from Twitter is a bit lighter, and from BlueSky a little darker. Can this also be used as a hint?

nikopol said:
Thank you!

I also noticed that the picture from Twitter is a bit lighter, and from BlueSky a little darker. Can this also be used as a hint?

Nope. You have to look for compression artifacts and image clarity when evaluating quality. It's worth noting that colors like magenta and bright red especially can appear darker or murkier due to compression artifacts, but overall image brightness is not a factor.

You can check howto:sites_and_sources next time if you want to know what sources are better.

If both are considered as unfavourable, highly consider finding another source the artist (or commissioners/character owners) posts to that may have a better quality. You may even ask politely from them for the original quality as well as their blessings to repost it, if you are courteous enough.
Otherwise, you would have to visibly compare the two images and choose the one which is higher in quality/less compressed like the others have said.