I'm not sure I understand the appeal of sites that are literally a less-than-a-sentence prompt I could have typed myself.
dwaltrip · 2h ago
But you didn't, and they did :)
It made me smile, I'm glad they built it.
dmd · 2h ago
Well, no, I have many times typed exactly that prompt, because making a hundred billion dollars worth of computer go "brr" to make ascii art is amusing! I just don't understand what the point is of a dedicated site.
Life is complete, I now have a beautiful 2D ascii elephant!
Random idea: now animate it.
Run the initial generate, keep that result, provide it back to another llm call: “user requested an elephant. you drew this. (object here) generate the next frame of an animation of this.”
Iterate the prompt 2-3 times for cool animated ascii art :)
4m1rk · 4h ago
I like that! Though it's not very smart or consistent at creating ascii art.
mysterydip · 4h ago
Not quite sure it's working properly. Just saw a shrek face labeled "pikachu", and a column with nothing else labeled "windmill".
4m1rk · 4h ago
Yeah, it sometimes doesn't feel like it. I tried different prompts but it's not that smart at generating ascii art.
Simorgh · 5h ago
Is it using a model to generate the ASCII art?
It’s quite clever.
4m1rk · 5h ago
It's gpt-4.1
Simorgh · 5h ago
Got it — it’s interesting how a text model can generate compelling visual representations that are essentially style-matching.
Matching styles using vision models is not that easy, especially if you want to capture the core subject of the prompt.
It made me smile, I'm glad they built it.
Random idea: now animate it.
Run the initial generate, keep that result, provide it back to another llm call: “user requested an elephant. you drew this. (object here) generate the next frame of an animation of this.”
Iterate the prompt 2-3 times for cool animated ascii art :)
It’s quite clever.
Matching styles using vision models is not that easy, especially if you want to capture the core subject of the prompt.
ASCII Art for "bambi":
/ \