2 Followers
0 Following
300 Posts

I write code and play games and stuff. My old username from reddit and HN was already taken and I couldn't think of anything else I wanted to be called so I just picked some random characters like this:

>>> import random >>> ''.join([random.choice("abcdefghijklmnopqrstuvwxyz0123456789") for x in range(5)]) 'e0qdk'

My avatar is a quick doodle made in KolourPaint. I might replace it later. Maybe.

日本語が少し分かるけど、下手です。

Alt: [email protected]

Short video of a bilaterally gynandromorphic Green Honeycreeper near Manizales, Colombia, 20 May 2022. (Video John Murillo)

A recording of an extremely rare half male, half female Green Honeycreeper (Chlorophanes spiza)... #bird #nature #ornithology #video

https://figshare.com/articles/media/DSCN2268_MOV/23739894

DSCN2268.MOV

Short video of a bilaterally gynandromorphic Green Honeycreeper near Manizales, Colombia, 20 May 2022. (Video John Murillo)

figshare

This is a composite I stitched together from 12 screenshots taken from episode 5 of Penguindrum. I pretty much completely rewrote the tool I used for my last composite to make this. My old program could only handle solving two images; the new version solves a graph of correspondences one pair at a time to attach as many images as it can to a pinned starting image. I spent all weekend writing this specific editor, and much of my free time over the last couple weeks has gone into my broader art tool project.

I haven't done much in the way of artistically changing the piece. Other than the final cropping (which I did in the GIMP), it's as true to the imagery from episode 5 as I could get it. Notably though, the lower portion of the image and the upper portion of the image (with the reapers) are not from one continuous shot in the episode; part of my motivation for stitching this was that I wanted to see how it all looked when put together. I was surprised to find that the top portion with the reapers was actually scaled differently than the lower portion. I've cropped it for the main post, but if you'd like to see what it looks like without the final crop, you can see that image here.

The "lovers" are clearly a parody of The Kiss by Klimt -- https://en.wikipedia.org/wiki/The_Kiss_(Klimt) -- but I have no idea what the heck is going on with the cherubs in the middle...

Over on Hacker News, user z991 had the fun idea of hooking up Dall-E 3 and GPT4-Vision in a loop and they made a Show HN thread about it: https://news.ycombinator.com/item?id=38432486

This is the image output from part way through one of their results that I enjoyed. You can see the rest of the chain here: https://dalle.party/?party=RqpIijhH

Prompt -- note, multiple parts! Initial prompt to Dall-E, another for GPT4 to generate prompts for Dall-E to make more images from, and one for what was fed to Dall-E later in the chain.

Original prompt to Dall-E:

A truck going down a highway at full speed

GPTV4 Prompt:

Write a prompt for an AI to make this image. Just return the prompt, don't say anything else. Replace everything with a paperclip.

Five steps deep in the chain, this is the prompt GPT4 generated:

Create an image of a surreal landscape where everything is made of paperclips. There should be winding roads, mountain ranges, and forests, with the sun setting in the background. The elements, including the trees, mountains, and roads, should all have a metallic paperclip texture and appearance. The overall scene should have a warm, golden hue to simulate sunset lighting.

There's a lot of other fun combinations like "Corgi Madness" in the thread as well.

Show HN: A Dalle-3 and GPT4-Vision feedback loop | Hacker News