For [model/version], create [output] with [subject], [composition], [lighting], [reference instructions], and [ratio].
Nano Banana Prompts with Real Outputs
Use model-specific image prompts with the version, reference-image instructions, aspect ratio, and a record of what changed between runs.
Choose a task
Start from the result you need, not a vague style label.
Copy and adapt
Replace the marked variables and keep the constraints that matter.
Check the result
Compare the output with the task and record what changed.
Search by job
Find a prompt that completes a task
Model-Specific AI Prompts
Use prompts that include model versions, settings, supported inputs, and known limitations instead of generic instructions.
Nano Banana Prompts with Real Outputs
Use model-specific image prompts with the version, reference-image instructions, aspect ratio, and a record of what changed between runs.
Start with one controlled change
Run one prompt with the shown context, change one variable, and compare the result with the stated task. We keep the original attempt visible so you can understand what the prompt actually changed.
- • Check the model and version before copying.
- • Preserve details that must not change.
- • Record failures instead of hiding them.
Use the model’s native image-editing workflow. Preserve [details], change [one variable], and report any limitation.
Generate three variants for [model/version] and explain which parameter changed between each run.
Published examples include a test date, model context, and a human review path. Results vary by model, input, settings, and reference material. Use images and likenesses only with permission.