Prompt Engineering > Image Gen Workflow

Image Generation for Video Workflow

You've reached the final and most critical step in preparing for advanced video creation. Vidu's "Reference-to-Video" feature is powerful, but it follows a simple rule: garbage in, garbage out. The quality and consistency of your reference images will directly determine the quality and consistency of your final video. Let's learn the professional workflow for creating them.

Why Image Generation is a Separate Skill

Creating a good *video* prompt is different from creating a good *image* prompt. For video, your prompt focuses on action and camera work. For your reference images, the prompt must focus entirely on creating a perfect, static visual asset. This requires a different mindset and a dedicated workflow.

1

Step 1: Design Your Character

Use your favorite AI image generator (e.g., Midjourney, DALL-E 3) to design your main character. Your goal here is to create a single, perfect "hero image."

Prompting Tips for Your Hero Image:

  • Be Hyper-Descriptive: Detail their hair, eyes, facial structure, clothing, and any key accessories.
  • Specify a Simple Background: Use terms like on a plain white background or studio portrait to isolate the character. This makes it easier for Vidu to understand who the subject is.
  • Aim for a Neutral Pose: A simple standing pose, facing forward, with a neutral expression is ideal for a primary reference.

Example Image Prompt:

Full body studio portrait of a female cyborg knight, intricate silver armor with glowing blue circuits, short white hair, determined expression, on a plain grey background, photorealistic, 8k.
2

Step 2: Create a "Character Sheet"

A single image is good, but a "character sheet" (or "turnaround") is better. This involves creating multiple images of the *exact same character* from different angles. This gives Vidu more data to work with, leading to better consistency.

To do this, you need to use a feature in your image generator that ensures consistency. The two most common methods are:

  • Using Seeds: Many generators allow you to reuse the "seed" number of a previous generation. By using the same seed and slightly modifying the prompt (e.g., changing "front view" to "side view"), you can create variations of the same core character.
  • Character Reference Features: Newer image generators (like Midjourney's --cref) have dedicated character reference features. You can provide the URL of your hero image and the AI will ensure all new generations match that character's appearance.

Example Character Sheet Prompts (using the same seed/cref):

1. ...character prompt..., front view.

2. ...character prompt..., side view.

3. ...character prompt..., back view.

4. ...character prompt..., dynamic action pose.

3

Step 3: Build Your Asset Library

Repeat the process for any key objects or backgrounds. Do you need a specific "cyberpunk hoverbike"? Generate a clean image of it on a plain background. Need a "glowing enchanted forest"? Generate a beautiful landscape shot. These will become the assets you upload to Vidu's "My References" library.

Your Workflow is Complete

You now have a complete, end-to-end workflow:

  1. Image Generation: Create high-quality, consistent reference images for your characters, objects, and scenes.
  2. Upload to Vidu: Upload these assets to your "My References" library in Vidu AI.
  3. Video Generation: Select your references and write a video prompt that focuses on action and camera work.

By separating the task of asset creation from video creation, you gain immense control and can produce professional-quality narrative content.

You've Mastered the Roadmap!

Congratulations! You have completed the Vidu AI Prompt Engineering roadmap. You have all the skills needed to create incredible AI videos. Now, it's time to put it all into practice.

Go to Vidu.com and Start Creating →