AI Storytelling Workshop

WAN Image & Video × Gemini

A step-by-step workshop guide for creating AI-generated stories with characters, scenes, and videos — Google Gemini handles the planning, WAN brings it to life.

Gemini — Plan WAN — Generate

Workshop Overview

In this workshop you will create a short animated story using two AI tools: Google Gemini for writing and planning, and WAN for image and video generation. Follow the steps below in order — each step builds on the one before it.

1. Story → 2. Characters → 3. Scenes → 4. Char Images → 5. Scene Images → 6. Scene Videos
Phase 1 · Pre-Production

Plan with Google Gemini

Gemini · Free

Prepare the Story

Use Gemini to draft the full narrative of your story. This gives you a clear roadmap before generating any visuals.

  • Describe the genre, tone, and target length to Gemini
  • Ask for a story with a clear beginning, middle, and end
  • Request the output broken into distinct scenes
  • Keep it concise — 6 to 10 scenes is a good range for a workshop
Prompt example: "Write a short fairy tale about a robot who discovers a garden. Break it into 8 scenes. Each scene should have a one-paragraph description and one line of dialogue."
Gemini · Free

Design the Characters

Define each character in detail so that both Gemini (for script) and WAN (for images) understand them consistently.

  • List every character that appears in the story
  • For each character, specify: name, appearance, clothing, key visual traits
  • Write descriptions in plain visual terms (e.g. "A small blue robot with round eyes and a silver antenna")
  • Keep descriptions consistent — these will be used as image prompts later
The character descriptions you write now become the reference prompts for WAN in Step 4. Make them vivid and specific.
Gemini · Free

Set the Scene Descriptions

Turn each scene from your story into a detailed visual description that an image generator can understand.

  • For each scene, write: setting/location, lighting, mood, which characters are present, and what they are doing
  • Include camera direction if helpful (e.g. "wide shot of a forest", "close-up on the robot's face")
  • Note the dialogue or narration that accompanies each scene
Good scene descriptions bridge the gap between storytelling and image generation. Think like a film director writing shot notes.
Phase 2 · Production

Generate with WAN

WAN Image

Generate Character Images

For each character, generate images repeatedly until you get a look you are happy with. This is an iterative process.

  • Use the character descriptions from Step 2 as your image prompts
  • Generate multiple variations of each character — try 3–5 generations per character
  • Tweak the prompt each time: adjust lighting, expression, angle, or style keywords
  • Save the best image for each character — this becomes your character reference
Don't settle on the first result. Small prompt changes (adding "cinematic lighting", "portrait style", "full body shot") can dramatically change the output.
WAN Image

Generate Scene Images

Now generate the background or full scene image for each scene in your story.

  • Use the scene descriptions from Step 3 as your prompts
  • Generate multiple versions per scene until the composition feels right
  • Pay attention to how the scene image will frame the characters
  • Save the best scene image for each story beat
Scene images set the mood. If a scene is meant to feel tense, try adding keywords like "dramatic shadows", "low key lighting", "stormy sky".
WAN Video

Generate Scene Videos

Combine everything: use each scene image as the starting frame, add the dialogue for that scene, and attach the character reference image so WAN knows who is speaking.

  • For each scene, provide WAN Video with: the scene image (from Step 5), the dialogue/narration text to be spoken, and the character reference image (from Step 4)
  • Generate each scene video one at a time
  • Review the output — check lip-sync, character consistency, and timing
  • Assemble the scene videos in story order to create your final piece
If a character doesn't look consistent across scenes, go back to Step 4 and refine the character reference image. A strong reference image is the key to visual continuity.
Pro Tips

Iterate, Iterate, Iterate

AI generation is rarely one-shot. Budget time to generate 3–5 versions of each image before picking the best one. Small prompt tweaks make big differences.

Keep a Prompt Journal

Write down which prompts worked and which didn't. This saves time when you need a similar result later and helps you learn the tools faster.

Consistency is Everything

Use the same character reference image across all scenes. If you change a character's look mid-project, regenerate all affected scenes.

Work Backwards from the Deadline

Video generation takes the longest. Plan to finish Steps 1–5 with enough time left for multiple rounds of video generation and review.