Free AI Workflow

How I Made the Baby STORM MV 🎬

From reference photos to Higgsfield Genjutsu — here's the complete step-by-step setup.

This is the full workflow I used to recreate the STORM-style video with a child as the main character, Indian ethnic-wear background dancers and a custom heritage environment. No gate. No fluff. Everything is below.

Jump to Workflow Follow on Instagram

AI video generation is not perfectly deterministic, so results may vary slightly from generation to generation.

What You'll Need

01A clear photo of your main child

02AI image generation / ChatGPT

03A Higgsfield account

04A short source video for Motion Transfer

Quick flowSource video → Main Character → Background Dancers → Environment → Higgsfield Genjutsu → Motion Transfer

STEP 01

Prepare the Motion Reference 🎬

The first thing you need is the short section of the original STORM video that you want to use as the motion reference.

For this workflow, the reference section was taken from approximately 4:21–4:49.

Keep the clip under 30 seconds. A 27–29 second clip works well.

This video will be used by Higgsfield as the driving reference for:

—camera movement
—choreography
—body movement
—performer positions
—spacing
—formation changes
—timing
—framing
—editing rhythm

Important

The source video controls MOTION only. Your reference images will control the appearance of the characters and environment.

STEP 02

Create IMAGE 1 — Main Character

IMAGE 1 is the main / center character. Use a clear photo where the child's face is visible and the outfit is easy to understand.

For better facial consistency, you can use multiple photos of the same child while creating the reference image.

The goal is to create one clean full-body reference: standing upright, facing camera, black background, same face, same hairstyle, same outfit, holding a milk carton, straw already attached.

IMAGE 1 Prompt

Create a photorealistic full-body image of the child from the supplied reference photos.

Preserve the child's identity as accurately as possible:
- same face
- same facial proportions
- same eyes
- same curly hairstyle
- same toddler proportions
- same Indian ethnic outfit
- same clothing colours
- same overall appearance

The child should be standing upright and facing directly toward the camera.

The child is holding one small milk carton naturally in one hand.

A short white drinking straw is visibly inserted into the milk carton.

Keep the milk carton and straw clearly connected as one object.

Show the child from head to toe.

Use a completely solid black background.

Keep the result realistic and photographic.

Do not redesign the child's face, hairstyle or outfit.

Full-body reference of the main child on a black background holding a milk carton

The cleaner IMAGE 1 is, the easier it is for Motion Transfer to understand who the lead character should be.

STEP 03

Create IMAGE 2 — Background Dancers

IMAGE 2 is not one specific character. It is a visual reference pool for all of the other children appearing in the crowd. The most important thing is diversity.

Do not include the main child's face in this image. Create five different Indian children with noticeably different faces, hairstyles, body types and outfits. They should all wear festive Indian ethnic clothing.

IMAGE 2 Prompt

Create a photorealistic full-body group portrait of five different young Indian children standing side by side against a completely solid black background.

All five children should face directly toward the camera and stand naturally with their arms relaxed.

Include both boys and girls.

Every child must have a clearly different face and appearance.

Create natural variation in:
- facial features
- hairstyles
- hair length
- height
- body proportions
- slim, average and slightly chubby builds

Dress all five children in festive Indian ethnic clothing.

Girls may wear:
- lehenga choli
- embroidered festive skirts
- mirror-work outfits
- Gujarati or Rajasthani-inspired clothing
- dupattas

Boys may wear:
- kurta pajama
- dhoti-style bottoms
- embroidered waistcoats
- festive jackets
- traditional Indian sets

Use different but visually coordinated colours.

Every child should wear stylish dark sunglasses.

Make sure all five children are visible from head to toe.

Do not include:
- the IMAGE 1 child
- the IMAGE 1 child's face
- milk cartons
- straws
- props

The five children must look like five genuinely different children rather than copies of one another.

Five diverse Indian children in festive ethnic wear and sunglasses on a black background

STEP 04

Create IMAGE 3 — Environment

Originally, a playground-style environment was tested, but a simple heritage location worked much better for the Indian ethnic visual direction.

The background should remain EMPTY. Do not put children into the environment reference. Use pale stone, ivory / cream architecture, a heritage staircase, simple composition, natural daylight and minimal decoration.

IMAGE 3 Prompt

Create a photorealistic cinematic image of an empty Indian heritage staircase.

The setting should feel authentic, elegant and understated.

Use:
- warm ivory stone
- cream and off-white tones
- naturally aged stone steps
- subtle Indian heritage architecture
- minimal architectural detailing
- soft natural daylight
- gentle shadows
- very minimal greenery

The staircase should provide a large usable area for a group of dancers.

Keep the scene uncluttered and spatially simple.

Do not include:
- people
- children
- crowds
- playground equipment
- excessive decoration
- fantasy architecture
- overly ornate palace elements

The result should feel like a real heritage location photographed in natural daylight.

Use a 16:9 aspect ratio.

Empty warm-ivory Indian heritage staircase in natural daylight

Keep the environment simple. Genjutsu needs room to place the performers naturally.

STEP 05

Set Up Higgsfield Genjutsu

Open Higgsfield and select Higgsfield Genjutsu, then select Motion Transfer.

Upload the assets in this order:

REFERENCE VIDEO — the STORM source clip

↓

IMAGE 1 — Lead Character (main child)

↓

IMAGE 2 — Other Children (background pool)

↓

IMAGE 3 — Environment (empty heritage staircase)

Higgsfield Genjutsu Motion Transfer setup with reference video and three images

Make sure "Motion Transfer" is selected. Do not use Object Swap for this workflow.

STEP 06

Use the Final Genjutsu Prompt

This is the final prompt used to tell Genjutsu how each reference should behave. Copy the whole thing.

Final Genjutsu Prompt

Use the ORIGINAL VIDEO as the fixed source for camera movement, framing, choreography, body motion, timing, performer positions, spacing, editing rhythm, shot order, and total duration.

Replace the original environment and performers according to the reference images below.

==================================================
CHARACTER ASSIGNMENT
==================================================

IMAGE 1 = the ONE center / lead child.

IMAGE 2 = the visual reference pool for ALL OTHER children.

IMAGE 3 = the heritage staircase environment.

Maintain approximately the same number of performers, formations, spacing, positions, choreography, and density shown in the original video.

Track each performer according to their role in the source choreography rather than their temporary screen position.

==================================================
1. CENTER / LEAD CHILD → IMAGE 1
==================================================

Identify the single lead performer in the original video.

Replace that performer completely from head to toe with the child shown in IMAGE 1.

Match IMAGE 1 as closely as possible:
- same child identity
- same facial appearance
- same curly hairstyle
- same toddler proportions
- same height-to-body ratio
- same overall silhouette
- same green Indian ethnic kurta
- same light beige bottoms
- same overall outfit colours and styling

Do not retain the original performer's face, clothing, footwear, body appearance, or styling.

The original performer provides ONLY:
- pose
- skeletal movement
- choreography
- position
- orientation
- scale
- movement timing

IMAGE 1 determines the entire visible appearance.

There must be EXACTLY ONE IMAGE 1 child throughout the entire video.

Never duplicate the lead child among the other performers.

==================================================
2. LEAD CHILD PROP
==================================================

The lead child begins holding ONE small milk carton naturally in one hand.

A short white drinking straw is physically inserted into the carton.

The carton and straw must remain one connected object at all times.

The relationship must always remain visually clear:

CHILD'S HAND → MILK CARTON → ATTACHED DRINKING STRAW

When the choreography brings the child's hand toward his face, preserve the milk carton in his hand and keep the straw physically connected to the carton.

The prop should move naturally with his hand and body.

Do not create any additional handheld objects.

==================================================
3. ALL OTHER PERFORMERS → IMAGE 2
==================================================

Replace EVERY performer other than the IMAGE 1 lead child with young Indian children visually based on IMAGE 2.

IMAGE 2 represents a visual pool of different children.

Do not repeat one identical child across the entire group.

Create natural variation across the ensemble in:
- boys and girls
- facial features
- hairstyles
- hair length
- tied and open hairstyles
- height
- toddler body proportions
- slim, average, and slightly chubby builds
- sunglasses shapes

Every background performer should remain visually different from the lead child.

The IMAGE 1 child's face must NEVER appear on any background performer.

==================================================
4. INDIAN ETHNIC WARDROBE
==================================================

All IMAGE 2 performers wear festive Indian ethnic clothing based on IMAGE 2.

Girls may wear:
- lehenga choli
- embroidered skirts
- festive blouses
- dupattas
- mirror-work details
- Gujarati or Rajasthani-inspired styling

Boys may wear:
- kurta pajama
- dhoti-style bottoms
- embroidered waistcoats
- festive jackets
- traditional Indian sets

Use varied but visually coordinated colours and designs.

Every child wears stylish dark sunglasses.

Do not reproduce the original performers' wardrobe.

==================================================
5. CROWD CONSISTENCY
==================================================

Maintain approximately the same performer count and crowd density as the original video.

The IMAGE 1 child is the ONLY lead character.

All remaining performers are distinct IMAGE 2-style Indian children.

Do not create:
- duplicate lead characters
- repeated identical faces
- obvious cloned children
- adults
- teenagers
- original performers
- original wardrobe

When characters rotate, cross paths, become partially hidden, or leave and re-enter the frame, preserve their established identities.

==================================================
6. ENVIRONMENT → IMAGE 3
==================================================

Replace the original environment with the heritage staircase shown in IMAGE 3.

Maintain:
- warm ivory and cream stone
- realistic Indian heritage architecture
- lightly sunlit atmosphere
- simple uncluttered composition
- natural stone staircase
- subtle architectural detailing
- minimal greenery

Keep the environment realistic and photographic.

Do not introduce unrelated structures or extra background people.

The location must remain spatially consistent throughout the video.

Integrate all performers naturally using correct:
- perspective
- scale
- ground contact
- shadows
- lighting
- depth
- occlusion

==================================================
7. CHARACTER CONSISTENCY
==================================================

IMAGE 1:
Always the exact same ONE lead child in the green kurta outfit.

IMAGE 2:
A diverse ensemble of distinct Indian children in festive ethnic clothing and sunglasses.

Fast choreography, camera movement, overlap, rotation, or motion blur must not change character identity.

Dancing changes pose and movement only.

Appearance remains consistent.

==================================================
8. SOURCE VIDEO MOTION LOCK
==================================================

Preserve the ORIGINAL VIDEO as closely as possible for:

- camera movement
- framing
- lens perspective
- shot composition
- choreography
- body motion
- performer positions
- performer spacing
- formation changes
- orientation
- timing
- shot order
- editing rhythm
- total duration

Do not redesign the choreography.

Do not introduce new dance movements.

Do not randomly reposition performers.

The ORIGINAL VIDEO determines MOTION.

IMAGE 1 and IMAGE 2 determine CHARACTER APPEARANCE.

IMAGE 3 determines ENVIRONMENT.

==================================================
FINAL PRIORITIES
==================================================

1. Preserve the original camera movement, choreography, framing, performer positions, timing, formations, shot order, and duration.

2. EXACTLY ONE lead performer uses IMAGE 1.

3. Preserve the IMAGE 1 child's face, curly hairstyle, toddler proportions, green Indian ethnic kurta outfit, light beige bottoms, and overall appearance.

4. All other performers use IMAGE 2 as the visual reference and must have different faces from the lead child.

5. Preserve realistic diversity across the ensemble.

6. The lead child's milk carton remains clearly visible and physically connected to its drinking straw.

7. Replace the environment with the warm ivory heritage staircase from IMAGE 3.

8. Do not restore the original performers or wardrobe.

9. Maintain character identity and environment continuity throughout the entire video.

10. Original video = MOTION.
Reference images = APPEARANCE AND ENVIRONMENT.

STEP 07

Generate

Once everything is uploaded:

Double-check Motion Transfer is selected.
Check IMAGE 1, IMAGE 2 and IMAGE 3 are in the correct order.
Paste the final prompt.
Start the generation.
Wait for Higgsfield to finish processing.

Generation time varies depending on video length, queue, server load and complexity. For a roughly 28-second workflow like this, expect the generation to take some time rather than finishing instantly.

Troubleshooting

If Higgsfield Flags the Generation

Sometimes the source clip itself can trigger Higgsfield's safety system. In this workflow, the original source footage contains a gesture close to the performer's mouth.

If a generation is flagged:

—avoid repeatedly mentioning restricted objects in the prompt
—describe only the positive action you want
—keep the milk carton and straw clearly connected
—simplify the prompt if necessary
—if the source clip itself continues triggering the safety system, trim or replace the problematic opening moment

Clean prop wording

The lead child begins holding one small milk carton naturally in one hand.

A short white drinking straw is physically inserted into the carton.

The carton and straw remain one connected object.

When the choreography brings the child's hand toward his face, preserve the milk carton in his hand and keep the straw physically attached to the carton.

Do not create any additional handheld objects.

If Higgsfield shows "credits refunded", the rejected generation was not charged.

Things That Improve Your Chances

Use a sharp IMAGE 1 with the full body visible

Keep IMAGE 1 on a plain black background

Use multiple reference photos when creating IMAGE 1 if identity accuracy matters

Make IMAGE 2 children visibly different from one another

Never include the main character's face in IMAGE 2

Keep IMAGE 3 empty

Avoid overly complicated environments

Preserve the source video for motion only

Keep all reference roles clearly defined

Do not overload the prompt with unnecessary instructions

AI video generation is probabilistic, so one generation may differ from another

About Free Generations

Higgsfield occasionally provides free generations or promotional allowances. Availability can differ between accounts, promotions, models and plans.

Do not assume every new account will automatically receive a free Genjutsu generation. Always check what your account shows before generating. If the button displays a credit amount, the generation may use credits.

Want more AI content workflows like this?

I share practical AI workflows, content experiments and marketing ideas through Think Social.

Follow on Instagram Explore Think Social

Made by Think Social.