Koke_Cacao-TextSynthesis

Koke_Cacao-TextSynthesis

(05-TextSynthesis)

inferkit
narrative-device

Using: Narrative Device
Themes: coding, homework
Result: The teacher looked up from his computer screen and saw that his student had not yet arrived. “I’m going to
have to start grading your homework,” he said, “but I want to make sure you are doing it right.”
Comment: This has nothing to do with coding homework. The model interpret coding as teacher’s computer
screen. The result might by highly “plagiarized” from a story book.

Using: inferkit
Result:

The teacher looked up from his computer screen and saw that his student had not yet arrived. "I'm going to have to start grading your homework," he said, "but I want to make sure you are doing it right."

The student sighed, rolled his eyes, and said, "There's no point.

This is the ninth grade.

You'll do whatever I say and there's nothing I can do about it. "He sat down, and opened his book.

His teacher put his name in the "Now On and shook his head.

"I get to do this each and every night. Why can't you do it during the day?"

"Well," the teacher said, "when I set my alarm clock for noon, my friend buys me coffee.

I set it back to 5: 30 and I'm still awake.

It's got to be midnight by the time you go to bed.

There's no point in going to sleep."

The student sighed and closed his book.

Comment: The model only seems to pay attention to the previous 2 sentences and major characters. Although attention (to
characters) is all you need, but it is hard for the model to generate a global level theme of the text that is
consistent and relevant.

Koke_Cacao-ArtBreeder

Koke_Cacao-ArtBreeder

(05-ArtBreeder)
ArtBreeder
ProsePainter

Genre Picture
Genre Picture
Genre Parameter
Genre Parameter
Breeder Picture
Breeder Picture
Breeding Parameter
Breeding Parameter

The art breeder model is very cool. It magically preserves the lower-level pattern of oil paintings with relatively
high resolution. This quality is often, at least to me, hard to achieve and requires a lot of hyperparameter tuning
and GPU power. The fact that it can run on the browser this fast is amazing (is it running on a browser though? Or
does it use backend GPU with querying?). Also, it is very rare to have a user interface that allows both breeding
and gene-editing. I wonder if “breeding” means parameter breeding rather than pixel-wise breeding.

Koke_Cacao-Pix2Pix

Koke_Cacao-Pix2Pix

(05-Pix2Pix)

Phone Cat
Phone Cat
Phone Cat with Black Screen
Phone Cat with Black Screen

I have played this model years ago. The model tries very hard to fit both low-level pattern and border structure to
the input data, which creates funny-looking images. But the model is only good at interpreting thin lines, with
equal stroke weight and without color.

Wormtilda-TextSynthesis

 

The InferKit Tool has definitely sampled from a lot of writing because I kept trying to make the text generator give new Fight Club rules, experimenting with changing the syntax of the sentence and removing words, but without fail it kept saying that the first rule is “you don’t talk about fight club.” That answer is accurate of course, but I kind of wish the generator was worse at it’s job so it could output a funnier answer. But otherwise, I thought the InferKit created surprisingly realistic and natural writing most of the time. The Narrative Device website created similarly natural-sounding writing. However, in my little experimentation I did with it, I felt that one theme usually overpowered the other. Like here I kept switching the second theme, keeping the first as “dogs with guns,” and each new prompt always had nearly nothing to do with the second theme, and was focused squarely on the exploits of armed canines. Maybe the generator is just really really interested in dogs with guns.

Wormtilda-ArtBreeder

These are the images I ended up generating with ArtBreeder. I gravitated towards creating somewhat bleak landscapes, and I think a lot of the landscapes ArtBreeder generates are really interesting. I could see myself using the tool to help generate the base aesthetics of some alien/sci-fi/alternate dimension landscape. Also, I found it super surprising that this was made by a BCSA almunus. Although I hadn’t used ArtBreeder before,  I have definitely seen it talked about all  over the internet. I’m surprised that CMU doesn’t advertise this fact front and center on the BSCA info page.

Wormtilda-Pix2Pix

These are two images  I generate, using the shoe and cat modes respectively. I thought this was pretty cool, but at the same time pretty frustrating, because with the cat one I gravitated to making the cat have really complex, unnatural forms, but Pix2Pix kept having trouble interpreting the image and it would keep coming out as an incomprehensible, blurry mess. I do realize why this is the case, however.

kong-VQGAN+CLIP

Textual Prompt: joyful beans

Textual Prompt: alice in the wonderland

It was interesting to see how both of my trials created a part of animal-like features (cat feet, fox?) when my prompts didn’t include direct relationships to animals. The resulting images were definitely different from my initial expectations, which I believe would require an in-depth understanding of the software to achieve. I like how the images contain brushstroke features, but maybe it was due to the images not being fully processed.

qazxsw-VQGAN+CLIP

Prompt: “A Pingpong learning to ski”

I can see the background is like a mix of ski trail and ping-pong table. I guess the creature on the right is the skiing ping-pong, but I’m not sure why it looks furry.