Browse prompts by model
A prompt is written against a model. Run it on a different one and you get a different picture, which is why every prompt here names the one it was tested on.
GPT Image
Strong at following a long instruction literally and at legible text in an image.
OpenAINano Banana
Built for editing an existing image and for holding a face across edits.
GoogleGemini
General image generation inside a general assistant, useful for quick iterations.
GoogleMidjourney
The most art-directed look of the group, at the cost of literal instruction following.
MidjourneyFlux
Photographic output with room to run it yourself rather than through one vendor.
Black Forest LabsVeo
Video with sound, and the steadiest camera moves of the video models here.
GoogleRunway
Video generation alongside an editing suite, rather than generation on its own.
RunwayKling
Long clips and dependable motion, with a usable free level.
KuaishouChoosing a model for a prompt
A prompt that depends on holding a face needs a model built for editing an existing image. A prompt that depends on readable text in the picture needs one that renders type. A prompt that is mostly a mood does better on a model with a strong house look and worse on one that follows instructions literally.
The model named on a prompt is the one it was run on, not the only one it will work on. Moving a prompt to another model usually means adjusting a clause or two rather than rewriting it.