TheFluxTrain
Sign in

Tutorials

Tutorial·

How a perfume brand ad was made with AI tools | TheFluxTrain

A simple, story-first walkthrough for beauty and lifestyle brands—from the initial idea and mood to the finished cut—without a full traditional crew.
Saquib Alam Saquib Alam

AI tools now make it possible to produce polished video ads without cameras, actors, or a physical studio. This tutorial presents a brief third-person workflow used to create an advertisement for an Indian fragrance brand using AI tools.

1. Write the Story

The process starts with a simple chronological story outline in ChatGPT. Instead of asking the model to invent the full concept from scratch, the creator defines the sequence of events first.

Then ChatGPT is used to:

  • expand the outline into story beats
  • extract all characters and important objects
  • generate detailed consistency descriptions for each recurring element

This structured list becomes the foundation for visual generation.

2. Generate Characters and Assets

Next, all characters and reusable key elements are generated on a white background so they can be reused across multiple scenes.

In TheFluxTrain, the creator uses Image from Text to generate each character from the prepared descriptions. Producing multiple angles (for example, close-up and side profile) improves flexibility during scene composition.

For environments and locations, scenes are generated directly as full backgrounds instead of white-background assets.

3. Create Storyboard Keyframes

After that, the creator identifies key moments in the narrative. These moments become individual video shots.

Using previously generated characters and elements, the creator combines assets into keyframe images that represent each scene. These keyframes then serve as guidance for video generation.

Perfection is not required at this stage; keyframes can be refined iteratively.

4. Generate the Videos

With keyframes prepared, the workflow moves to the video generation pipeline.

Inside TheFluxTrain, models such as Kling 3.0 are used to generate clips by providing:

  • the start frame
  • reference images
  • an action prompt

This process is repeated until all shots in the story are generated.

5. Edit the Final Ad

Finally, all generated clips are assembled in a video editor such as After Effects or VN.

The creator arranges clips in story order, adds voice narration with ElevenLabs, layers in background music, and exports the final ad cut.

This completes a full AI-native ad workflow from concept to delivery.