← ClaudeAtlas

brand-static-ad-batchlisted

Use when someone asks to research a brand and batch-generate many static image ads (for example 40 ad concepts) from its real product photos with Nano Banana 2 through the fal API.
valezion7/turn-any-videos-into-skills · ★ 3 · AI & Automation · score 71
Install: claude install-skill valezion7/turn-any-videos-into-skills
## When to use The person wants a high volume of static ad creatives for one brand or product: DTC brands, agencies, performance teams. They have or can get product photos and a brand URL. ## What is new here - Run three phases in order: brand DNA, then templated prompts, then image generation. Each phase leaves a file on disk. - Nano Banana 2 on fal: `fal-ai/nano-banana-2/edit` takes the prompt plus product photos (`image_urls`), `num_images` 1-4, `aspect_ratio`, `resolution` 0.5K/1K/2K/4K, `output_format`. About $0.08 per image at 1K, so 40 ads cost about $3.20. - Known failure points: .avif inputs, quotes inside prompts, several images per call, and the script only worked once the fal docs were supplied. ## Setup 1. Check what exists: `python --version`, `python -c "import fal_client"`, and whether `FAL_KEY` is set. Never print the key's value. 2. Ask before installing or signing up. If the person agrees: `pip install fal-client`, and create a key at https://fal.ai. The key goes in the `FAL_KEY` environment variable, never in a script or a committed file. 3. Project layout: - `brands/<brand>/product-images/` - `references/ad-templates.md` (the concept templates, with placeholders) - `generate_ads.py` 4. Verify: run one template at 1K and confirm the result contains `images[0].url`. 5. Other tools can generate too (for example Higgsfield's web UI, where the original framework was done by hand). fal is one option, chosen here for a clean API. ## Steps 1. **Collec