So this week everybody and their mother turned their profile picture into a Studio Ghibli character, and yes, fine, I did it too. My cat Biscuit is now rendered as a soft-eyed anime familiar sitting in a sunlit window with little dust motes floating around her, and I have looked at it probably forty times since Wednesday.
For anyone who missed it (unlikely, it ate the internet for about four straight days): OpenAI rolled native image generation into GPT-4o inside ChatGPT, and within about 48 hours the entire timeline turned into a Ghibli filter factory. Family photos, wedding photos, that one guy's mugshot, all of it got the Miyazaki treatment. It was genuinely everywhere. Sam Altman was even joking on Twitter (I refuse to call it X, sorry) about the GPUs melting from demand, and by midweek the free tier was getting rate-limited because so many people were hammering the thing at once. I tried it around 11pm on a Tuesday and got queued behind what felt like the entire population of the internet.
Here's my actual problem with it, and its a small one, not a moral crusade: Hayao Miyazaki has been on the record for years, like actually years before any of this was a product, saying he finds AI-generated animation an insult to life itself. Thats not me paraphrasing loosely, its close to his exact words from a documentary clip that circulates every time this argument comes up. And every time it comes up, everyone posts the clip, agrees for about six seconds, and then goes back to uploading their dog photos anyway. I did the same thing. I know the guy would hate this. I made the cat picture anyway. Theres no tidy way to resolve that so Im not going to pretend I resolved it.
What I will say is the actual output is impressive in a way that's hard to be cynical about once you're looking at it. It's not the old diffusion-model mess where hands have eleven fingers and text is soup. The model actually holds onto small details across the image, it renders readable text now, and it seems to genuinely understand what you're asking for instead of just vibing in the general direction of your prompt. That's a real jump. I've been messing with AI image tools since the DALL-E 2 waitlist days and this is the first time one has made me go "oh, huh" out loud, alone, in my apartment, which is a low bar for entertainment I realize but there it is.
The copyright and style-imitation question isn't going away though, and I don't think "well the output isn't a literal frame from a real Ghibli film" settles it the way some people online are acting like it does. Style isn't nothing. A studio spends forty years building a specific visual language and then a product ships that reproduces it on command for free, and the fact that its legally murky doesn't make it feel less like using someone's labor without asking. I don't have a clean answer here. I just don't think the "its transformative, actually" argument some tech twitter accounts were making is as airtight as they wanted it to be.
Anyway. Aside from the Ghibli stuff, my actual week was mostly spent fighting with my router, which decided sometime Monday that it no longer wanted to hold a stable 5GHz connection for more than about twenty minutes at a stretch. Ended up hardwiring my desktop with an ethernet cable I had to dig out of a drawer I haven't opened since we moved apartments two years ago, so that's a fun little archaeology project I didn't ask for. If your internet has been acting weird lately too, check your router's channel width settings before you go buy a new one, mine had auto-selected some insanely wide channel that was stepping on every neighbor's wifi in the building. Fixed it in about four minutes once I actually looked.
Biscuit, for the record, does not care that she's been immortalized as anime. She spent the better part of today asleep on a heating vent and knocked a pen off my desk twice, once seemingly on purpose while maintaining eye contact with me the entire time.