PlanetZ forums - scopeusers.com

Hey all, having fun with mid-journey,
i wantd to share those 4 images because the prompt used was "spacef-devices KrOn on bycicle".... not sure how it interpreted the sentence, but here are the images . no bycicle, and disappointingly, it did not look for spacef-devices kron on the web... Seems to make great record covers...

warning: because of the Terms and free trial, it is creative commons with attribution - no commercial use allowed
anyone else using those ? I am also following facebook group AI generated art, great images over there and sometimes people give away invites (I may do so if I get a paying license, but not on my agenda right now).
https://www.facebook.com/groups/1895623370624892

I am spending some time with AI. I have experimented with pattern recognition with a cnn network.
I am currently developing a speech recognition algorithm.
Have done some study and pre-work in natural language processing, read the "understanding natural language" (1972) by Terry Winograd
I have interest in symbolic ai, philosophy and linguistics.

https://www.youtube.com/watch?v=QAJz4YKUwqw

I am not sure about the site you mention but the upper right part looks kinda like a handle of a bicycle, the rest looks like roads... There is also something like space...
so perhaps the bicycle context + space lead the algo to that. I am not sure what it is exactly. Is it ai generated art? These work (as i suppose) by blending patterns from images associated with the keywords contexts.
Yes quite nice as covers.

i've learned a bit about AI last year and here it is not "conversational AI" like on the video, in the sense that there is no "conversation" but an instruction to start and a picture as a result. there are many different image generators but I don"t think I've heard of one that asks you "please could you specify this word that I do not understand". They are also voluntarily limited in order to not help producing "deep fakes" and realistic images (you cannot do a human face, results will be weird and sometimes funny, but not at all realistic).

I used "Midjourney" https://www.midjourney.com/ and it runs inside a discord channel. for the moment you must be invited so CPU is not overwhelmed, and you get can invites easily on various channels on facebook.


it could be a bycicle seen for space, and yes there is a bike handle with a "device". but it either skipped "kron" or used as an effect (the stripes?). the rendering style here is left to the machine. there are way too influence the style. for example ending a prompt with "moebius" was indeed in style of the drawer Moebius.

the way this one works is simple: you send a server request [ /imagine prompt (then expression of what you want to see). ) and it runs 4 versions like the above thumbnails. Then you can "upscale" (add details) to the ones you prefer. all runs through discord (which is a bit messy). generated images are also on your account page. I had an invite today and gave it a few runs. It is interesting to see how language is interpreted. makes me work my basic English too as it is not my main language. errors can lead to fun results too. I expected it to look on the net for terms that are obviously names or brand but it does not.


There are tricks and additional instructions but I just had a quick free trial this morning... used up the 25 free images in a matter of minutes (the calculations through the discord channels is super fast... results are simple compared to other software.

I like what people do with Disco Diffusion too, much slower and complex ( I like this style better because of the painter effect. you can change the style but it will always look like a painting or drawing. still, inspiring. people used that as a base to edit in photoshop (or directly) https://www.youtube.com/watch?v=YRcVoLL ... MHB4AaABAg this channel has a few tutorials) you can do videos with it but it is flickering a lot and still not really watchable (it's recalculate effect each frame so it is not smooth even though "makes sense").


all these tools require a little monthly budget like subscriptions or google drives and CPU rentals. and time. so I am not 100% into that, just tried today

a few example with midjourney. I gave a few runs with the term "hamburgers" in various contexts.... yes he spaceship is attacked by hamburgers shooting lasers. the scale is right and the burger is microscopic compared to the spaceship. there must be a way to change that. . on of the pic is in the style of "moebius", but the rest is up to the machine. Some complexity did not pass through. The spaceships are supposed to be over New york or over Paris (there is an eifel tower though, and on the first run it is "new york as a spaceship" , not sure where the giant red cloud is ... a tomatoes slice? . the batman series has my vocabulary errors like "bathroom" instead of bathtub, "hamburger floating" instead of "floating in the air".... all this must be tried though, lots of options. that's the only learning curve.


(also, the vikings series, the 8 images, " "viking eating hamburger, with moon in the sky no clouds"
it shows a pattern in creation. many viking seem to "share a hamburger with the moon", very poetic.

but i don't know how it ended up generating a Viking inside a hamburger (or "as a hamburger" ) !!!
that's hilarious !!

fra77x2 wrote: Tue Jun 28, 2022 10:10 am I am currently developing a speech recognition algorithm.

this is very technical and complex... not at all in my skills but read a lot about that as I was involved in a project that aimed at building an ai that could guide people through everyday problems, mainly consumer issues (there was a test made with humans(lawyers) to evaluate the kind of issues and it's crazy how many people are lost or panic in front of basic issues such as reading terms and conditions, finding contacts on a website, solving the issue with support teams etc. many people panic imagining problems that do not exist - or seek confirmation about what to do). i think the project failed because of "finances not found", to build something like that would require crazy amount of data, it would probably take 10 years of financing before any tangible result. you need to be google or tesla/space x and do a lot of networking with VCs and banks and earn their trust before getting any funding.

yes mostly it is as you describe. The google speech recognition is very good I think.

I have 3 invites for whoever needs one. PM please. first arrived first served, no selection.