Skip to content
PI Square
Book

Generative content

On-brand images, video and copy in volume, checked against your brand rules before anyone sees them.

approved assetscontent credentialsAalogobrand kitsceneproductexact photostudiobudgetbrandcampaignchannel2991:12994:52999:1629916:9every format · prices set from the recordlogookcolouroktypeokclaimsokproductokcheck the brandkids jacketwinter, 4 channelsbriefapprovea person approvesstorywebstore
  1. 1Brief
  2. 2Search approved assets
  3. 3Brand kit and prompts
  4. 4Generate, product kept exact
  5. 5Every format
  6. 6Check the brand
  7. 7A person approves
  8. 8Stored and published
Try it

A generative content system lets marketing teams make campaign images, video and copy with AI models, inside the rules of each brand. PI Square builds studios where every brand has its own rules and its own isolated environment, and every asset is scored against those rules before a person approves it.

  1. A brief

    the product, the audience, the channel

  2. Your cloud

    Brand rules applied

    layered prompts for each brand

  3. Images, video and copy made

    refined over several rounds

  4. Scored against the rules

    before a person sees it

  5. A person approves

    then it goes to your channels

What it's for

  • Retail

    Campaign images and video for several brands, each in its own style.

  • Consumer goods

    Product imagery for every channel, with the product placed exactly as photographed.

  • Financial services

    Campaign copy drafted inside the wording rules your compliance team sets.

  • E-commerce

    Product descriptions written from the product record, in your tone.

How we keep it safe

  • Approved assets are searched first, so nothing is generated when something you own already fits.
  • The product stays exactly as photographed; only the scene around it is generated.
  • Each brand runs in its own cloud project, so no brand's rules or assets reach another.
  • Two checks on every asset: fixed rules in code, then a model scoring it against the brand.
  • Prices and legal claims come from your records, never from the model.
  • A budget on each request caps the rounds of refinement and the spend.
  • Every approved asset is stored with content credentials, recording that AI made it and who approved it.

Built with

  • Gemini image models
  • Veo
  • Vertex AI
  • Next.js
  • FastAPI
  • Cloud Run
  • Firestore

Where we've built this

  • A retail group with two businesses and four brands: a studio where marketing teams make campaign images and video, with each brand in its own isolated cloud project. The first phase is live.

How an engagement runs

  1. Understand

    We look at how assets are made today, what each one costs, and which brand rules must hold.

  2. Prove

    One brand and one channel, with your team approving every asset.

  3. Embed

    Your team owns the brand rules, review process and asset workflow. Additional brands, channels and formats are introduced when useful.

The full approach

Questions about generative content

Will AI-made assets stay on brand?
Only if the rules are written down. We turn each brand's rules into layered prompts, score every asset against them, and send it to a person for approval. Anything that fails the rules is made again or dropped before anyone sees it.
Can it use our own product photography?
Yes. Your product photographs can be placed into generated scenes, so the product appears exactly as photographed, with the label and colours unchanged.
How will people know an image was made with AI?
Every approved asset carries content credentials, the C2PA standard: a signed record that AI made it, with which tool, and who approved it. Some platforms strip that record on upload, so your library keeps the original as the proof.
Who owns what the system makes?
Ownership follows your contract with us and the model provider's terms, which we go through with you during Understand. Your brand assets, prompts and outputs are kept in your own cloud.
How do you enforce brand separation in multi-brand businesses?
We can separate brand assets, prompts and permissions using dedicated projects or other access boundaries suited to your setup. The design includes tests for unintended access across brands and a review process for material that needs to be shared.

The other five

Tell us what you're working on

Keep this general; leave out sensitive or confidential information.

Or email nikhil@pisquare.ai