Model Product Shoot

Places a styled model and a product into a reference scene. The reference image drives the composition, framing, and lighting — the model and product are swapped in while preserving everything else.

Four images passed to Nano Banana in this order:

1. Reference image — composition, framing, lighting
2. Product image — the product to feature
3. Model headshot — model identity reference
4. Styled model image — clothing and body reference

Output saved to ./brands/[brand-name]/model-product-shoot/[output-name]/.

Step 1 — Select brand

Scan ./brands/ for subfolders that contain a brand-identity/visual-guidelines.md file.

• One brand found: use it automatically and confirm: "Using brand: [name]"
• Multiple brands found: list them and ask which to use
• None found: tell the user to run /brand first

Step 2 — Create reference folder and collect reference image

Create the shared composition references folder:

mkdir -p brands/[brand-name]/model-product-shoot/composition-references

List any images already in the folder. If empty, tell the user:

"I need a reference image — this is a photo showing the kind of scene, framing, and lighting you want. It could be a model holding a product on a track, applying something post-workout, or any lifestyle scene that fits the vibe. The model and product in the reference don't matter — I'll swap them out.

>

Drop it into: brands/[brand-name]/model-product-shoot/composition-references/

>

Name it clearly — e.g. track-hold.jpg, postrun-apply.jpg. Let me know when it's in."

If images are already present, list them and ask which to use.

Wait for the user to confirm before proceeding.

Step 3 — Select product

Ask: "Which product should the model be holding or interacting with?"

Check brands/[brand-name]/brand-identity/product-images/ for a match. Show what was found:

Found:
  ✓ freeze-roll-on.jpg

Not found:
  ✗ [product name] — not in product images

If not found:

1. Check brands/[brand-name]/brand-identity/products.json for the product and its image URL
2. Try to download it:

   curl -L "[product-image-url]" -o "brands/[brand-name]/brand-identity/product-images/[slug].jpg"

1. If the download fails or the product isn't in products.json, create a product upload folder and ask the user:

   mkdir -p brands/[brand-name]/model-product-shoot/product-uploads

"I couldn't find that product image. Drop it into: brands/[brand-name]/model-product-shoot/product-uploads/ and let me know when it's in."

Step 4 — Select model and styled image

Scan brands/[brand-name]/models/ for subfolders containing headshot.png and fullbody.png.

Present available models:

Available models:
  • sofia
  • marcus

Ask: "Which model would you like to feature?"

Then scan brands/[brand-name]/styled/ for output folders that reference the selected model. List the styled outputs found:

Styled outputs for sofia:
  • sofia-allblack-athletic (sofia-allblack-athletic_v1.png)
  • sofia-puresport-cap (sofia-puresport-cap_v2.png)

Ask: "Which styled outfit would you like to use?"

Use the most recent versioned file (highest version number, or the unversioned file if only one exists).

If no styled outputs exist for the selected model, tell the user:

"No styled outputs found for [model]. Run /style first to dress the model."

Step 5 — Aspect ratio

Ask: "What aspect ratio? Default is 3:4. Other options: 1:1, 4:5, 2:3, 9:16, 16:9."

Use 3:4 if the user doesn't specify.

Step 6 — Name the output

Derive a slug from the model name, product, and reference:

• sofia-rollerbottle-track
• marcus-creatine-postrun

Ask the user to confirm or rename.

Step 7 — Write model-product-shoot-spec.json

Create the output folder and write the spec:

brands/[brand-name]/model-product-shoot/[output-name]/
  model-product-shoot-spec.json

{
  "output_name": "sofia-rollerbottle-track",
  "brand": "puresport",
  "product_name": "Freeze Roll On",
  "product_image": "brands/puresport/brand-identity/product-images/freeze-roll-on.jpg",
  "composition_reference": "brands/puresport/model-product-shoot/composition-references/track-hold.jpg",
  "model_name": "sofia",
  "model_headshot": "brands/puresport/models/sofia/headshot.png",
  "styled_image": "brands/puresport/styled/sofia-allblack-athletic/sofia-allblack-athletic_v1.png",
  "aspect_ratio": "3:4"
}

Step 8 — Generate

Run from the project root:

python3 skills/references/generate-model-product-shoot.py brands/[brand-name]/model-product-shoot/[output-name]

Uploads images in order: composition reference → product → headshot → styled image.

Calls fal-ai/nano-banana-2/edit at 2K in the chosen aspect ratio.

Cost: ~.12 per image.

Step 9 — Present the result

Tell the user the exact path where the file was saved, e.g.: brands/[brand-name]/model-product-shoot/[output-name]/[output-name]_v1.png

Ask: "Happy with this? Or would you like to regenerate, swap the reference, or try a different product?"

Once the user is happy, suggest the next step:

"Great shot. If you'd like to animate this into a short video clip, run /add-motion."

To regenerate (same spec):

python3 skills/references/generate-model-product-shoot.py brands/[brand-name]/model-product-shoot/[output-name]

Each run saves as the next version — [output-name]_v1.png, [output-name]_v2.png, etc. Previous outputs are never overwritten.

To try a different reference: update composition_reference in the spec and re-run. To swap the product: update product_image and product_name and re-run. To change the outfit: update styled_image and re-run.

Notes

• Image order matters: composition → product → headshot → styled model. The script handles this automatically.
• The reference image is the strongest driver of the output — choose it based on the scene and energy you want, not the product or model in it
• Multiple brands have their own model-product-shoot/ subfolder — independent workflows can run side by side
• Composition references are shared across all outputs for that brand — reuse them freely