AI Learn
Midjourney · AI Image Tutorial

Midjourney: Control Midjourney with image prompts and references

Use a reference image to influence content, composition or colors while keeping the text prompt responsible for the final scene.

Reviewed September 19, 2026Independent tutorialFeatures may vary
Not official support. AI Learn is an independent education site and is not affiliated with Midjourney. Check the provider’s current documentation for plan-specific limits, pricing and feature availability.
Before you start

Image prompts guide rather than copy an image exactly. Reference behavior and weighting can differ by Midjourney version.

1. Choose a reference for a specific reason

Decide whether the reference is mainly for composition, subject silhouette, color, mood or another visual property.

2. Attach the image in the Create interface

Upload or select the image and attach it as the appropriate reference/input type for the current interface.

3. Write a complete text prompt anyway

Describe the final image you want. Do not write only instructions like “make this better,” because the reference is inspiration, not an edit specification.

4. Adjust influence carefully

If the current version exposes image-weight or related reference controls, change them gradually and compare how strongly the result follows the reference.

5. Match crop and composition when possible

A reference with a similar aspect ratio and framing gives the model a cleaner structural signal than a radically different crop.

Try this prompt

Use the attached image only as a reference for [composition/color/mood/subject]. Final scene: [description]. Preserve [specific property] but change [specific property]. Do not copy any text, logo or incidental background detail from the reference.

Common mistakes

Assuming the model will reproduce the reference exactly; using a reference without saying what matters; relying on a mismatched crop; escalating weight instead of improving the text prompt.

Finish check

You should be able to say which part of the result came from the text and which visual property was stabilized by the reference.

开始之前

Image Prompt 的作用是引导,而不是精确复制。参考图行为与权重范围可能因 Midjourney 版本不同。

1. 先明确参考图到底参考什么

先决定这张图主要提供的是构图、主体轮廓、配色、氛围还是其他视觉属性。

2. 在 Create 界面加入参考图

上传或选择图片,并根据当前界面把它作为合适的参考/输入类型加入提示词。

3. 仍然要写完整文字提示词

描述最终想要的画面,不要只写“把这张图变好”。参考图是引导,不是自动等于精确修改指令。

4. 逐步调整参考影响程度

如果当前版本提供 image weight 或类似控制,逐步改变,不要一次拉到极端,并比较结果跟随参考图的程度。

5. 尽量让参考图裁切与目标画幅接近

参考图与最终画幅、构图越接近,模型收到的结构信号通常越清晰。

可以直接套用的提示词

只把这张参考图用于[构图/颜色/氛围/主体]。最终画面:[描述]。保留[具体属性],改变[具体属性]。不要照搬参考图中的文字、Logo 或无关背景细节。

常见错误

以为模型会精确复制参考图;不说明到底要参考什么;参考图裁切与目标画幅完全不匹配;文字提示词有问题时只会不断提高参考权重。

完成检查

完成后,你应该能说清:哪些内容由文字提示词定义,哪个视觉属性是被参考图稳定下来的。

開始之前

Image Prompt 的作用是引導,而不是精確複製。參考圖行為與權重範圍可能因 Midjourney 版本不同。

1. 先明確參考圖到底參考什麼

先決定這張圖主要提供的是構圖、主體輪廓、配色、氛圍還是其他視覺屬性。

2. 在 Create 介面加入參考圖

上傳或選擇圖片,並根據目前介面把它作為合適的參考/輸入類型加入提示詞。

3. 仍然要寫完整文字提示詞

描述最終想要的畫面,不要只寫“把這張圖變好”。參考圖是引導,不是自動等於精確修改指令。

4. 逐步調整參考影響程度

如果目前版本提供 image weight 或類似控制,逐步改變,不要一次拉到極端,並比較結果跟隨參考圖的程度。

5. 盡量讓參考圖裁切與目標畫幅接近

參考圖與最終畫幅、構圖越接近,模型收到的結構信號通常越清晰。

可以直接套用的提示詞

只把這張參考圖用於[構圖/顏色/氛圍/主體]。最終畫面:[描述]。保留[具體屬性],改變[具體屬性]。不要照搬參考圖中的文字、Logo 或無關背景細節。

常見錯誤

以為模型會精確複製參考圖;不說明到底要參考什麼;參考圖裁切與目標畫幅完全不匹配;文字提示詞有問題時只會不斷提高參考權重。

完成檢查

完成後,你應該能說清:哪些內容由文字提示詞定義,哪個視覺屬性是被參考圖穩定下來的。

Official references

Check current product details at the source

Related practical tutorials

Continue with another single-task workflow.

All tutorials →