How to Use a Reference Video and Text Prompt to Generate a New AI Video

Some of the most controllable AI video workflows combine two forms of direction: a reference video and a written prompt. The reference video provides movement, timing, camera behavior, and composition, while the text prompt explains what the new version should look like. For creators searching for an AI reference video generator, this approach offers a practical balance between preserving useful structure and introducing new characters, styles, environments, or visual ideas.

Why Combine Video and Text

A reference video already answers many questions that a text prompt would otherwise need to describe. It shows how quickly the subject moves, where the camera is positioned, how the shot is framed, when actions happen, and how the scene progresses over time. The text prompt can then focus on the transformation.

Instead of asking AI to invent both motion and appearance, the creator can preserve the motion language of the original clip while directing the model toward a different visual result.

Choose the Right Reference Clip

The best reference clip is not always the most complex one. Clean movement, visible subjects, and understandable camera motion usually provide a stronger foundation. If the goal is to replace a person, choose footage where the body and face are visible for most of the clip. If the goal is to recreate a cinematic action, use a shot with clear staging rather than rapid cuts.

The reference should contain the motion you want to keep. Anything unnecessary can become additional information the AI has to interpret.

Write the Prompt Around Keep and Change Instructions

A useful prompt clearly separates preserved elements from transformed elements. Start by stating what should remain consistent, such as the original body movement, timing, pose sequence, camera angle, framing, or background layout. Then explain what should change, such as the character identity, outfit, art style, props, lighting, or environment.

For example, a creator can instruct the model to keep the dance movement and camera position from the source clip while replacing the performer with a specific character shown in attached reference images. This is more precise than simply asking for the clip to be restyled.

Add Strong Visual References

Reference images can help define details that the video alone does not contain. Character portraits can guide facial identity and hairstyle. Full-body artwork can clarify clothing and proportions. Separate images can establish accessories, products, or props.

Moescape AI is useful here because creators can generate or refine these visual assets before beginning the video workflow. A character concept can be developed through image generation, then combined with a reference clip and prompt for a more directed transformation.

Refine the Output Through Controlled Iteration

After generating the first version, review which elements were preserved correctly and where the model drifted. If the motion changed too much, strengthen the instruction to preserve timing and pose. If the character identity is inconsistent, use clearer or more consistent reference images. If important props disappear, describe their placement and appearance more directly.

Treat each generation as a step toward the final shot rather than expecting one prompt to solve every detail.

A Strong Workflow for Controlled AI Video

An AI reference video generator becomes much more useful when creators understand the role of each input. The video controls motion and structure. The prompt controls intent. Reference images control visual identity.

Moescape AI brings these elements into a broader creative workflow, making it possible to move from character design and visual references to transformed video within the same platform. For creators producing AI cosplay, anime edits, character scenes, promotional content, or stylized remakes, this approach offers a clear path from reference footage to a new and original result.