Guide

Image to Prompt の仕組み

This guide explains how the homepage workbench converts an image into a prompt, how to read the structured output, and how to refine results for different generators.

1

Step 1: Analyze the image as direction, not caption

The backend is instructed to recover creative direction instead of writing a plain description. That is why the structured output includes subject anchor, scene envelope, style signals, shot logic, emotion, and remix handles.

2

Step 2: Choose the prompt shape you need

The homepage mode switcher lets you reinterpret the same underlying analysis for a general prompt, Midjourney, Flux, Stable Diffusion, or raw JSON.

3

Step 3: Edit the result, don’t restart it

Use the structured view to understand which parts control the result. Small edits to style signals, shot logic, or remix handles are usually faster than writing a new prompt from scratch.

ここで試す

自分の参照画像から始めましょう

ホームページと同じワークベンチをここでも使えるので、ガイドを読みながらその場でツールを試せます。

ここに画像をクリック、ドラッグ、ドロップ、または貼り付け
PNG、JPG、WebPを最大10 MBまでサポートしています。
クリップボードから画像を直接貼り付けることもできます
出力スタイル
画像から始める

Prompt、構造化ビュー、JSONを生成します。

Use the mode chips above to reshape the output for Midjourney, Flux, Stable Diffusion, or a machine-readable JSON flow.

ガイド FAQ

生成前に知っておきたいこと

Why keep the tool on the homepage?

Because the homepage is the strongest SEO page. Putting the workbench there keeps engagement and conversion signals concentrated.

Why show structured output and JSON?

Different users want different levels of control. Some want a ready-to-copy prompt, while others need editable building blocks or machine-readable data.

最初の prompt を得た後は何をすればよいですか?

そのままコピーするか、特定の部分だけ編集するか、ワークベンチのツールから直接画像生成や動画生成へ送れます。