Your First Adaptation

This tutorial walks you through creating a culturally adapted TV spot using the multi-agent pipeline. By the end, you’ll have an origin script adapted for a new market, with a generated storyboard.

Note

Make sure you’ve completed Installation & Setup and have all services running via ./start.sh. Familiarity with Your First Image is helpful but not required.

Overview

The adaptation workflow has five steps:

1. Create a Brand — define voice, values, and guidelines

2. Create a Campaign — the container for your TV spot project

3. Create the origin — a VideoAdUnit with the original script

4. Create an adaptation — target a new market; the pipeline runs automatically

5. Generate a storyboard — turn the adapted script into images

Step 1: Create a Brand

Navigate to TV Spots > Brands > Add Brand.

Fill in:

Code (required)

A short unique identifier, e.g. ACME.

Name (required)

Display name, e.g. ACME Corp.

Guidelines (recommended)

Brand voice, values, visual identity, and messaging rules. This is what the brand evaluation gate checks against during adaptation. For example:

Voice: Warm, approachable, optimistic. Avoid sarcasm.
Values: Family, togetherness, authentic moments.
Visual Identity: Bright colors, natural lighting, real people (not models).
Messaging: Always lead with emotional benefit, product secondary.

Insights (optional)

Structured brand insights as a JSON array:

[
  {"heading": "Core Positioning", "points": ["Family snacking moments", "Authentic, unscripted feel"]},
  {"heading": "Visual Guidelines", "points": ["Warm color palette", "Natural lighting only"]}
]

Click Save.

Step 2: Create a Campaign

Navigate to TV Spots > Campaigns > Add Campaign.

Fill in:

Job ID (required)

A unique tracking identifier, e.g. ACME-2024-SUMMER.

Client Name (required)

e.g. ACME Corp.

Script Title (required)

e.g. Summer Moments 2024.

Brand (recommended)

Select the brand you just created. This applies as the default for all ad units in the campaign.

Product Name (optional)

e.g. ACME Chips Classic.

Click Save.

Step 3: Create the Origin VideoAdUnit

The origin represents your original TV spot script — the source material that will be adapted.

Navigate to your Campaign’s detail page and click Create Origin Ad Unit, or go to TV Spots > Video Ad Units > Add Video Ad Unit directly.

Fill in:

Campaign (required)

Select the campaign you just created.

Code (required)

e.g. ORIGIN.

Title (required)

e.g. US English Origin.

Origin or Adaptation

Set to Origin.

Language (required)

The language of the original script, e.g. en-US.

Duration (optional)

Spot length in seconds, e.g. 30.

Click Save, then add script rows.

Adding Script Rows

On the VideoAdUnit detail page, add rows using the Script Rows inline:

Shot	Timecode	Visual	Audio
01	00:00:00	Wide shot of a sun-drenched backyard. A family sets up a picnic table.	Music: Upbeat acoustic guitar
02	00:05:00	Close-up of hands opening a bag of ACME Chips. Golden chips catch the sunlight.	VO: “Some moments just call for the perfect crunch.”
03	00:12:00	Kids and parents laughing, passing the bag around the table.	VO: “ACME Chips. Made for moments like these.”
04	00:22:00	Product hero shot: bag of ACME Chips against a sunset backdrop.	VO: “ACME Chips. Crunch into summer.”
05	00:27:00	End card: ACME logo with tagline.	SFX: Chip crunch sound

Save after adding all rows.

Step 4: Create an Adaptation

This is where the multi-agent pipeline takes over. Navigate to your Campaign and click Create Adaptation, or create a new VideoAdUnit directly.

Fill in:

Campaign (required)

Same campaign as the origin.

Source Ad Unit (required)

Select the origin VideoAdUnit you just created.

Origin or Adaptation

Set to Adaptation.

Region (required)

The target market region, e.g. DACH (Germany/Austria/Switzerland).

Country (optional)

Specific country within the region, e.g. Germany.

Language (required)

Target language, e.g. de-DE.

Brand (optional)

Override the campaign brand if the adaptation uses a different brand name in this market.

Persona (optional)

Target audience persona, e.g. Budget-Conscious First-Timer.

Click Save.

The Pipeline Runs Automatically

On save, the adaptation task is automatically queued to Celery. The multi-agent pipeline executes 7 nodes:

1. Concept Analyst — extracts core themes, emotions, and narrative structure from the origin script

2. Cultural Researcher — investigates the target market’s values, communication styles, and cultural sensitivities

3. Writer — adapts the script with culturally appropriate visuals and dialogue

4. Format Evaluator — verifies visual descriptions are in English, with only voiceover and supers in the target language (including English translations)

5. Cultural Evaluator — validates cultural sensitivity and appropriateness

6. Concept Evaluator — ensures the adapted script preserves the original campaign intent

7. Brand Evaluator — checks brand voice, values, visual identity, and messaging consistency

If any evaluation gate fails, the writer automatically revises the script (up to 3 retries per gate).

Monitor progress on the VideoAdUnit detail page — the status updates through each pipeline stage.

Viewing the Results

Once the pipeline completes (status: completed), the VideoAdUnit detail page shows:

Concept Brief

JSON output from the concept analyst: core themes, emotional arc, narrative structure extracted from the origin.

Cultural Brief

JSON output from the cultural researcher: target market insights, values, communication patterns, potential sensitivities.

Evaluation History

Array of results from each evaluation gate: pass/fail decisions with reasoning.

Adapted Script Rows

The adapted script, saved as AdUnitScriptRow records. Visual descriptions remain in English; voiceover and supers are in the target language with English translations in brackets.

Step 5: Generate a Storyboard

With the adapted script complete, you can generate visual storyboard frames.

Navigate to the completed adaptation VideoAdUnit and click Generate Storyboard, or go to TV Spots > Storyboards > Add Storyboard.

Fill in:

Video Ad Unit (required)

Select the completed adaptation.

Diffusion Model (required)

Choose the image generation model. Good choices:

• Flux.1-dev — high quality, good for presentation storyboards

• SDXL Turbo — fast iterations for draft storyboards

LoRA Model (optional)

Apply a visual style to all generated frames.

Images Per Row (optional)

Number of image variations per script row. Default: 1.

Click Save. The storyboard task is automatically queued.

The system:

1. Generates image prompts from each script row’s visual description

2. Creates a DiffusionJob for each prompt

3. Queues all jobs to the Celery worker

4. Images generate sequentially (one at a time to manage GPU memory)

Monitor progress on the Storyboard detail page — it shows total jobs, completed jobs, and a progress percentage.

Once complete, the storyboard displays frame-by-frame with the generated images alongside the script rows.

What Just Happened

In this tutorial you:

• Defined brand guidelines that the pipeline evaluates against

• Created a campaign with an origin TV spot script

• Ran a 7-node multi-agent pipeline that produced a concept brief, cultural brief, and culturally adapted script

• Generated a visual storyboard from the adapted script using diffusion models

The entire adaptation — from origin script to adapted storyboard — was driven by the Django admin UI with Celery handling all background processing.

Next Steps

• Prompt Template Editing — edit the LLM prompts that drive each pipeline node

• Segments & Personas — create detailed audience personas for more targeted adaptations

• Working with LoRAs — apply LoRA styles to your storyboard frames