Google generate + Finder remove

big-O behavior change

the rapid adoption of generative ai-tools is fucking nuts.

i’m not talking about just the ‘end-users’ that downloaded the commercial version of ChatGPT in iOS.

i’m talking about businesses.

those companies create apps that a lot of people use, and lately, they’ve been heavily promoting ‘their’ ai-features.

it feels a little reactionary, but also, this shit is moving fast.

it might have been too early to invest much, but now is the time to invest at least some.

making an image

making an image is now the same process as using google, except starting with the word generate (and having the feature enabled, but this will eventually become default)

• initial request
• initial response of generated image(s)
• select from response(s) to download or
• adjust && generate new images

example with google:

1. input in searchbar: generate { image request }
2. view results
3. download from results

for adjustments:

1. request image
2. view results
3. select from results
4. create new variants based on selection ?
5. create new prompt based on selection ?

adjustments?

• variants produced
• select image
• adjust prompt based off selected-image, then generate variants
• create additional variants based on selected-image

removing the background

there are native features is OSX and iOS (stickers) that allow you to remove the background (or select a foreground object) easily.

• select image(s)
• more-actions 👉 remove background

what's the big deal?

there’s a lot of large systems that are being radically impacted at a speed that is very difficult to wrap your head around.

a short while ago, removing an object from the background was tedious & had poor results.

this tedious process was also blocked behind expensive software that required expensive hardware to run it.

now, Finder has a ‘remove background’ action built-in.

before, it would be hard to get placeholder images.

they were either:

• really, really shitty and distracting
• or you’d find some that are almost* really good

and spend time going down photo-picking-rabbit-holes.

ability to communicate with visuals

there’s a point where a gray-box communicates enough.

there’s also a point where the perfect image can convey not only what the image is in thatmoment, but what it will become.

instead of waiting for someone to make an image that you want to see, you’re able to go and make it.

i think that there’s a couple patterns people use when the interact with generative ai.

1. they’ve got a clear vision of what they want to see outputted
2. they have no idea what they want to see outputted

there’s a balance between the two, but i find that being ‘open’ is the best approach.

if there’s a specific type of image you’re trying to produce, then it might be worth digging around some prompt-forums to see how people have produced similar styles.

i’m surprised by something* every single time an image is generated.