The paper is attributed to "Anonymous Authors" and the GitHub has an image saying "comming soon" [sic] added five days ago. Of course the image is generated using this technique, but it's anyone's guess what it represents, a person sitting for meditation?
This is pretty cool, but also spooky. For example it seems like this has good commercial promise, as well as a serious amount of work invested, very strange that there's no person or company taking credit yet.
(edited to add, I see now that "comming soon" uses the "o" from their "Ashtanga Yoga" example)
Are the chinese examples as legible as the english ones? The english transforms are comprehensible because only one letter was changed, if more or all were it would get difficult to understand.
To me it looks like proportionally more of the content-carrying form has been modified in the chinese. But I can't read chinese and so can't evaluate.
Native Chinese speaker here. All transformations seem legit to me, the 兔/rabbit one was especially good since it is beautifully transformed, and I would say it still readable for a native speaker. For the 猫/cat and 冲浪/surfing cases, I feel like the transformation is too intentional, or less natural compared to the other cases (including the English ones). They are more readable though than 兔/rabbit.
(Not a native Chinese speaker). 兔/rabbit is marginal to me - I don't think I would even register it as a character unless it was in the right context. The others are perfectly legible. For the bold-face 猫/cat (the first one), you do technically lose some "content carrying form" as the 田 becomes a shape that does not appear in Chinese (sideways 曰), but I don't think that would actually confuse anyone in the context of the rest of the character.
In Japanese 田 is extremely common, meaning rice paddy, pronounced ta in names like Tanaka (田中, middle of rice paddy). It is simplified from the Chinese.
the Chinese examples are legit. Though the example on the far right (冲浪) is a word comprises of two characters, which means surfing only when the word is read as a whole.
Also noted that that is simplified Chinese, used by mainland China, that is, well, simpler than traditional Chinese used in Taiwan and maybe Hong Kong. For instance, 冲浪 is written as 衝浪 in traditional Chinese, which is more complicated and I'm wondering how they handle it.
Perfectly happy to be labeled as a grumpy snob, but I'm sad that things like this are going to make our visual world less and less beautiful as we replace humans--with training and experience and taste--with computer-generated triteness.
Some of it will replace humans with bad taste. Humans with experience and training are expensive, and we've often substituted our own mediocre work.
It will, unfortunately, also take the place of some good designers, usually the ones working for the least money (because they're the ones people don't want to be paying at all).
As for the rest... I don't know. I suspect it will mostly level things out. The worst will get better. The median will remain median. The best, hope, will continue to exist, for the people who do indeed have money to spend and want what still can't be automated.
As for the people who lose their jobs... that's a problem we've been grappling with for a long time. Thus far things often seem to even out, eventually. Unfortunately. I keep expecting people to start throwing off the whole notion of make work jobs, but that's an even bigger revolution.
Seems like the excitement comes from the fact that it's automated, more so than from the actual end-result. None of the logos shown here look visually balanced or well drawn. It will replace mediocre designers but not good ones.
This is a modern version of WordArt in MS Office in 1998. Those who would never employ an artist get something passable, and those who want to differentiate from the masses will adopt new styles from talented people.
I sort of think this might be the first step someone uses to make something cool.
For example the croissants and cupcakes example, the C in croissants sucks - so I wouldn't do that. The C in cupcakes is ok, but so maybe keep that but make the top cherry on the C cherry red. And relatively quickly you have a better result than just Croissants and Cupcakes although sure - sort of trite still.
My parents own a small construction business and their logo is... well, not very good IMO and relatively old and outdated. It's just "Family Surname + Construction Co." with the surname having a vague suggestion of a house. It wasn't that cheap back then, so they never want to change it because "we spent some money on it, why do it again?".
Can't wait to try this and see what comes up, bet I can improve it a bit and I have zero typography/design skills.
I can assure you they're not going to need another wordmark with another vague suggestion of a house in it. Just write the business name in a solid and readable typeface.
I have zero cello skills but I can discern between Yo-Yo Ma and a beginning cellist, or even between a brand new beginning cellist and a slightly more advanced beginning cellist.
"You can't judge X if you don't have skills in X" is a tired trope.
This is so cool, especially the Chinese characters! I really appreciate all of the creative projects like this the open nature of StableDiffusion has enabled.
Reminds me of the kids' show WordWorld which has animals and objects be built out of the word they represent. For example a sheep will be formed out of the 3D letters for "sheep".
If the author is here. Is there a way to try it myself?
The Demo[0] link is broken and the GH repo is empty[1].
[0] https://huggingface.co/spaces/SemanticTypography/Word-As-Ima...
[1] https://github.com/WordAsImage/Word-As-Image