The Same Prompt, Every AI Image Model (2022 → 2026)
The same prompt through every major text-to-image model, 2021 → 2026 — the pace of generative-image progress.
7credits3sourcesAll episodes
Used For
Credit
Source
01VQGAN+CLIP sample (2021)
VulcanSphereScenic Valley in the Afternoon Artistic (VQGAN+CLIP).jpg / CC BY 4.0
02DALL·E 2 sample (2022)
DALL·E 2 — via Cameron ButlerA soaking wet astronaut holding a white umbrella walks in the pouring rain on surface of the moon.png / Public domain
03Midjourney 'Théâtre D'opéra Spatial' (2022)
Jason M. Allen / MidjourneySpace opera 1 – Midjourney – Jason M. Allen – Colorado State Fair Fine Arts Competition entrant.jpg / Public domain
04DALL·E 3 self-portrait (2023)
DALL·E 3 — OpenAIDALL·E's self-portrait.jpg / Public domain
05DALL·E 1 announcement sample (2021)
OpenAIDALL·E 1 announcement montage / © OpenAI, editorial fair use
06GLIDE paper sample (2021)
GLIDE authors, OpenAIGLIDE Fig. 1 'a corgi' / arXiv 2112.10741, editorial fair use
07Imagen sample (2022)
Google ResearchImagen 'a corgi on a bike in Times Square' / © Google, editorial fair use
Data sources & references
01Model papers — DALL·E, GLIDE (arXiv:2112.10741), Imagen (arXiv:2205.11487), VQGAN+CLIP
02Stability AI (Stable Diffusion 1.5/SDXL/3.5), Black Forest Labs (FLUX), Ideogram, OpenAI (gpt-image) — model releases
03Replicate — API platform used to run the models under a fixed protocol