|
I, for one, am looking forward to the Star Trek computer being able to generate me novels in the styles of my favorite authors except featuring anthropomorphic hamburgers and hotdogs instead of humans If you're worried this will make your job obsolete, just remember that AI is going to make practically everyone's jobs obsolete eventually. Hopefully this will mean a Star Trek style utopia, but I'm thinking it'll be more like Judge Dredd Maybe we'll at least get mutants soon
|
# ¿ Apr 23, 2022 02:03 |
|
|
# ¿ May 14, 2024 13:00 |
|
If you haven't been able to get into dall-e 2, replicate has a nice alternative text to image generator that seems to do a better job at faces than dall-e mini, at least in the sense that they don't end up blurry looking: https://replicate.com/jack000/glid-3-xl It's limited to 30 mins of their cpu time a day but that should still be enough for a few dozen images depending on how much you tweak the generation settings Here's "a candid photograph of an evil horror clown Donald Trump"
|
# ¿ Jun 29, 2022 13:46 |
|
"A press photo of Donald Trump eating a burger" through the replicate generator:
|
# ¿ Jun 29, 2022 15:09 |
|
Last Visible Dog posted:Playing with DALL-E's edit function. I took a watercolor painting I'd done of a raccoon's head, made the paper transparent, cropped off the bottom, and resized it to give DALL-E some space to work with. This is incredible! Theoretically, you could also give a certain person a bunch of different outfits (or bodies) by just cropping their head into this and giving it a prompt, right? Also, most of the simple generators I've played around with are pretty good with horror type images, but that might just be because the things they produce are always a little disturbing. Would someone with DALL--E-2 or Midjourney mind trying the prompt "a press photo of an evil nightmare horror magician"? I'm curious as to how it'll compare to one of the replicate generators or dall-e mini's stuff. Here's replicate's attempt at it:
|
# ¿ Jul 4, 2022 21:06 |
|
huh, so it IS a completely different kind of result! very interesting! and these are great, thank you!
|
# ¿ Jul 4, 2022 21:20 |
|
Now these are definitely more disturbing. That top left one is especially great. I've noticed that some of the image generators it treats the words in prompts as weights, so if you put the same word multiple times it'd weigh it more heavily for the final image. Like, if I put "a grinning grinning grinning man" it would generate images of men with grins extending off the sides of their face and/or rising up past their eyeballs, usually with multiple additional rows of teeth. Do DALL-E-2 and Midjourney work the same way?
|
# ¿ Jul 5, 2022 03:47 |
|
TIP posted:which ones do the thing you mention? I've noticed it having a mild effect in mini dall-e/craiyon, but more of just weighting in a certain direction and not really the multiplicative effect you're talking about I've been using https://replicate.com/jack000/glid-3-xl for these, with the guidance scale at 17.77 and the seed at -1, as you said the effect isn't really as strong in mini dall-e Here's "a grinning man" Okay, screwed up the arms, but the grin isn't too awful. "A grinning grinning grinning man" Now we've got extra rows of teeth and the mouth is extra long. Also the eyes are getting a little crazy. "A grinning grinning grinning grinning grinning grinning man" Still getting extra rows of teeth and the huge cartoonish mouth, but now it seems to also always try to make the eyes into mouth or teeth like things.
|
# ¿ Jul 5, 2022 04:56 |
|
I did "a mad max *famous person*, manga scan, artwork by kentaro miura, high resolution" and got some decent results with stable diffusion: Richard Ayoade Benedict Cumberbatch Andy Samberg Selena Gomez Jeff Bridges Idris Elba Tom Hanks Natalie Portman Christopher Walken It also gave me some uh, more interesting results with certain celebrities: Ron Perlman Rosie O'Donnell Danny Devito
|
# ¿ Sep 25, 2022 12:48 |
|
https://cdn1.suno.ai/8e8542cd-4d80-421a-960a-e1298a5d70d2_0.mp4
|
# ¿ Nov 18, 2023 12:14 |
|
giving suno bizarre prompts and/or complete gibberish and seeing what it spits out is so much fun "aggressive glitch trap shrieking squealing SQUUUEEEALLL SQUUEAAALLL SQEUUAAALLLLL SEUQQALLLLLLLLLLE" https://cdn1.suno.ai/59fab262-3d7f-4bb8-aa89-cc8338b9f637_1.mp4 "hbfogh aeorohgoiarhg vrnirnvshi rhisd psgoehgs psergjosep grehig sbgiergsih axhiso pxshyyspa dfhdhea" https://cdn1.suno.ai/7df3be96-fe9f-42f6-87d7-9a648551a177_0.mp4 SCheeseman posted:death metal ba da pi da bop la de dah de doppity doo zipitty bippity boppity blooby bob doodah woop this is amazing i need more credits
|
# ¿ Nov 20, 2023 09:06 |
|
KakerMix posted:http://puu.sh/JUT0f/196252a496.mp4 Amazing how good it can make some of these, including this one. Seems like a great leap forward. Since it's using GPT3 or whatever for the lyric generation, it seems to have a generalized knowledge of things, too. For instance, I was inspired by the goonswarm rap posted earlier in the thread so I did "experimental aggressive grand opera rock dramatic intense track about the Goonfleet conquest of EVE" and it spit this out, which mentions space and ships in the lyrics: https://cdn1.suno.ai/19322906-2a00-4399-9569-18824e14d492_0.mp4 It also comes up with some interesting stuff when you give it complicated genre requests: "experimental aggressive complex sludge funk metal about hordes of people feasting at the dinner bell" https://cdn1.suno.ai/a1cae3b8-9592-4b49-9fb4-18bba364366a_1.mp4 Gashroom posted:“A song composed entirely of palindromes”… sort of. I like both versions These are both great and I'm impressed at the amount of palindromes it gets right (or pretty much close to it) in there.
|
# ¿ Nov 20, 2023 10:58 |
|
Bored posted:Dalle does not know that the word "gremlins" predates the 1980's movie. :nerdsnort: pretty sure most of those are mogwai
|
# ¿ Nov 25, 2023 03:40 |
|
There's too many good Suno generations to quote, keep posting em My contribution for this page is pizza time https://cdn1.suno.ai/970a227f-bfab-4df8-a792-57b0496a6e3a_1.mp4 Now that you can put your own lyrics, you have a little more control over song structure. So far I know you can use [Verse], [Pre-Chorus], [Chorus], and [Outro]. I haven't tested it but [Intro] probably works too. [Bridge] can also work but [Pre-Chorus] seems to work more consistently in what testing I've been able to do with my free credits.
|
# ¿ Nov 26, 2023 12:47 |
|
Lots of good tips for Suno generations ITT, it's great discovering all the nuances you can use to alter the outputs. https://cdn1.suno.ai/d6d622ab-d31b-474d-9bef-3d5c6db006e3_0.mp4 That parenthesis to voice over trick posted earlier also works sometimes if you use the [Pre-Chorus] input: https://cdn1.suno.ai/5ef99659-9f32-4b79-ba8d-95ce4ad10f90_0.mp4 One thing I am running into is it censoring all sorts of innocuous lyrics. Can't use a lot of copyrighted terms, which, sure, okay, but then I also get lyrics rejected when I use words like "turds" or "feces". How are some of you guys getting it to drop uncensored f-bombs?
|
# ¿ Dec 2, 2023 06:44 |
|
|
# ¿ May 14, 2024 13:00 |
|
mcbexx posted:Suno question: The whole "running jobs at once" is the number of different tracks you can have generating in the background. As far as I know Pro plan users still get two generations for each prompt, they can just have more prompts being generated at the same time.
|
# ¿ Mar 25, 2024 04:16 |