AI Learn
Runway · AI Video Tutorial

Build a consistent character with Runway References before video generation

Use a small reference set to lock character identity, generate controlled still variations, then move the strongest frame into an image-to-video workflow.

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

Runway References can use multiple reference images to influence a generated still before video. Consistency improves when references are clear, complementary and not overloaded with conflicting identities.

1. Prepare clean reference views

Choose a small set that clearly shows the same character’s face, hair, clothing or important design features. Avoid contradictory ages, lighting or accessories.

2. Name the invariants

Write the identity features that must remain: face shape, hairstyle, key garment, color accents and any distinctive prop.

3. Generate a new still first

Create the character in the target pose or environment as a still image. Solve identity and composition before asking for motion.

4. Compare identity systematically

Check face proportions, hairline, clothing details and repeated accessories across outputs. Keep only a frame that passes the identity check.

5. Send the approved frame to video

Use the consistent still as the input image, then prompt only the movement and camera behavior. Re-check identity during the entire clip.

Try this prompt

Use these references only to preserve the same character identity. Generate a still of the character in [new pose/scene]. Preserve face shape, hairstyle, signature clothing colors and accessory. Do not merge features from other people.

Common mistakes

Using many contradictory references; trying to fix identity only after video generation; judging consistency from clothes while the face changes; introducing a new character in the motion prompt.

Finish check

You have an approved still that matches the reference identity and a video clip that keeps the same character recognizable through motion.

开始之前

Runway References 可以用多张参考图影响静态生成,再进入视频。参考越清楚、互补且不互相冲突,身份一致性通常越容易控制。

1. 准备干净的多角度参考

选择少量、清楚展示同一角色脸型、发型、服装或关键设计特征的图片,避免年龄、光线或配件互相矛盾。

2. 写清身份不变量

列出必须保留的身份特征:脸型、发型、核心服装、颜色点缀和标志性道具。

3. 先解决静态图,再做运动

先生成角色在目标姿势或环境里的静态图,把身份和构图解决后再进入视频。

4. 系统比较身份一致性

逐张检查脸部比例、发际线、服装细节和重复配件,只保留真正通过身份检查的帧。

5. 把通过审核的帧送入视频

把一致性最好的静态图作为图生视频输入,文字只描述运动与镜头行为,并在整段视频里继续检查身份。

可以直接套用的提示词

这些参考图只用于保持同一个角色身份。生成角色处于【新姿势/场景】的静态图,保留脸型、发型、标志性服装颜色和配件,不要混入其他人的特征。

常见错误

使用大量互相矛盾的参考;视频生成后才想修身份;只看服装一致却忽略脸已经变了;在运动提示词里又引入新的角色。

完成检查

完成后,你应该有一张通过身份审核的静态图,以及一段角色在运动过程中仍然保持可识别一致性的影片。

開始之前

Runway References 可以用多張參考圖影響靜態生成,再進入影片。參考越清楚、互補且不互相衝突,身份一致性通常越容易控制。

1. 準備乾淨的多角度參考

選擇少量、清楚展示同一角色臉型、髮型、服裝或關鍵設計特徵的圖片,避免年齡、光線或配件互相矛盾。

2. 寫清身份不變量

列出必須保留的身份特徵:臉型、髮型、核心服裝、顏色點綴和標誌性道具。

3. 先解決靜態圖,再做運動

先生成角色在目標姿勢或環境里的靜態圖,把身份和構圖解決後再進入影片。

4. 系統比較身份一致性

逐張檢查臉部比例、發際線、服裝細節和重復配件,只保留真正通過身份檢查的幀。

5. 把通過審核的幀送入影片

把一致性最好的靜態圖作為圖生影片輸入,文字只描述運動與鏡頭行為,並在整段影片里繼續檢查身份。

可以直接套用的提示詞

這些參考圖只用於保持同一個角色身份。生成角色處於【新姿勢/場景】的靜態圖,保留臉型、髮型、標誌性服裝顏色和配件,不要混入其他人的特徵。

常見錯誤

使用大量互相矛盾的參考;影片生成後才想修身份;只看服裝一致卻忽略臉已經變了;在運動提示詞里又引入新的角色。

完成檢查

完成後,你應該有一張通過身份審核的靜態圖,以及一段角色在運動過程中仍然保持可識別一致性的影片。

Official references

Check current product details at the source

Related practical tutorials

Continue with another single-task workflow.

All tutorials →