Describe this reference image as a reusable prompt. Separate visible facts from uncertain guesses and format it for [model].
Photo to Prompt Workflow
Turn a photo into a structured prompt for a new image. Compare the generated description with the reference and edit uncertain details.

Imagen examples
Five visual directions, with the prompt behind each one
Select an image to inspect a copy-ready prompt for that direction. Compare framing, light, and setting before changing the prompt variables.

Copy-ready prompt
Illustrative directionDescribe the reference image as a reusable prompt. Separate observable facts from uncertain guesses. Include subject, composition, lighting, materials, color, and style. Format the result for [model]. Variation 1 — Wide editorial composition: keep the core subject and intent stable while changing only this visual direction. Preserve important details, avoid unreadable text and accidental logos, and frame the result for practical reuse.
This image is an illustrative category example; the prompt is a reusable direction and is not a verified run record.
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
Image to Prompt Generator
Upload a reference image and turn visible details into an editable prompt draft. Review the output before using it in an image model.
Photo to Prompt Workflow
Turn a photo into a structured prompt for a new image. Compare the generated description with the reference and edit uncertain details.
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.
Extract subject, composition, lighting, materials, color, and style from the image. Mark details that cannot be verified.
Create a structured prompt from this image and include three editable variables for subject, mood, and framing.
Know what this prompt can and cannot do
- Best for
- 从参考图提取可编辑的主体、构图、光线、材质和风格信息。
- Not for
- 把不可见信息当作事实,或要求模型准确恢复原始提示词。
What to edit before you run it
- • 可观察事实
- • 不确定推断的标记
- • 面向目标模型的可编辑变量
- • 把事实和猜测分开
- • 保留三到五个可替换变量
- • 用原图复核描述是否过度推断
- • 把品牌或文字猜错
- • 遗漏镜头与光线关系
- • 风格词过多导致主体不稳定
This page currently uses an illustrative example; no verified output record is published yet. We will add the model, version, run date, reviewer, and result notes when a run is documented. Results vary by model, input, settings, and reference material. Use images and likenesses only with permission.