Product Poster Generator (English)

Guided workflow for creating AI-generated product posters via giggle.pro image-to-image API. Always uses image-to-image mode — the user's product photo is required as the reference image.

Workflow

Stage 1: Collect Requirements

Ask the user for these items (combine into one message):

1. Product image ⚠️ Required: local file path to the product photo (e.g. /Users/xxx/product.jpg)
2. Product info: product name + one-sentence key selling point
3. Poster type: E-commerce launch / Brand event / Social media promotion
4. Visual style: Clean minimal / Bold & vibrant / Luxury premium / Tech futuristic / Playful youthful
5. Aspect ratio: E-commerce portrait 3:4 / Square 1:1 / Mobile portrait 9:16 / Wide landscape 16:9

Optional:

• Brand color (hex or description, e.g. "brand blue #1E3A8A")

If the user hasn't provided a product image path, do not proceed to generation — ask for it first.

If the user has already provided enough context, skip questions and proceed directly to Stage 2.

Stage 2: Compose Prompt & Generate

Step 1: Build the prompt using this BASE structure:

[POSTER TYPE] for [PRODUCT NAME]. [KEY SELLING POINT].
Style: [AESTHETIC], [COLOR PALETTE], [MOOD].
Visual: Keep the product as the hero element, [COMPOSITION — e.g. centered / rule-of-thirds], enhance with [BACKGROUND/SCENE — e.g. clean gradient / lifestyle scene / abstract elements].
Text layout: Main title "[PRODUCT NAME]", tagline "[SLOGAN]", clean typography with strong hierarchy.
Technical: Professional commercial photography quality, high contrast, sharp details, no watermark.

Read references/prompt-templates.md for type-specific templates, style keywords, and color palette references.

Step 2: Choose model and aspect ratio

Poster Type	Recommended Model	Default Ratio
E-commerce launch	nano-banana-2	3:4
Brand event	nano-banana-2	9:16
Social media	nano-banana-2-fast	9:16

Use nano-banana-2 for best image-to-image fidelity. Use nano-banana-2-fast for quick previews.

Step 3: Run generate_poster.py

generate_poster.py handles base64 conversion, task submission, polling, and URL extraction in one call. It prints the final plain URL(s) to stdout.

SKILL_SCRIPTS="<absolute_path_to_skill>/scripts"

poster_url=$(python "$SKILL_SCRIPTS/generate_poster.py" \
  --image "<product_image_path>" \
  --prompt "<composed_prompt>" \
  --model nano-banana-2 \
  --aspect-ratio <ratio>)

Step 4: Present results

$poster_url is already a clean, complete URL. Present it to the user as plain text:

Poster generated ✨

$poster_url

Let me know if you'd like any adjustments!

Critical URL constraints:

• Output $poster_url verbatim — do NOT wrap it in markdown [text](url) syntax
• Do NOT truncate, ellipsize (...), or reformat the URL
• Do NOT use [View image](url) or any link-text wrapper
• The URL is a long signed AWS CloudFront link — it must remain complete and unmodified

Environment Setup

Requires GIGGLE_API_KEY environment variable:

export GIGGLE_API_KEY=your_api_key

API key available at https://giggle.pro/ account settings.

Script Reference

scripts/generate_poster.py — all-in-one poster generation script (recommended).

• --image <path> → local product image (handles base64 conversion internally)
• --prompt <text> → generation prompt
• --model → nano-banana-2 (default), nano-banana-2-fast, seedream45, midjourney
• --aspect-ratio → 1:1, 3:4, 4:3, 16:9, 9:16, 2:3, 3:2, 21:9
• stdout: plain URL(s), one per line
• stderr: progress messages

References

• references/prompt-templates.md — Type-specific prompt templates, style keywords, color palette guide.