How to Create Facebook Ad Images and Videos From a Product URL Using AI

Most ad platforms give you one AI model for one task. Write copy here. Generate an image there. Stitch it together yourself. Hope it works.

KillScale's Creative Suite takes a different approach. Multiple specialized AI models work together in a single pipeline — each handling the task it's best at — to take you from a product URL to running ads on Meta in minutes. Not hours. Not days. Minutes.

This article walks through exactly how it works: which models do what, how the image and video pipelines differ, what the Creative Suite menu looks like, and how the monitoring loop feeds performance data back into your next round of creative.

The AI Models

Each model in the pipeline was chosen for a specific strength. No single model does everything well — but together, they cover the entire creative production workflow.

Claude is the backbone of the pipeline. Claude Sonnet analyzes product URLs, extracts product details, writes ad copy, and studies reference ad styles. Claude Opus acts as the creative director — it crafts rich, detailed image prompts with specific composition, lighting, and mood direction. Two tiers working together: Sonnet for analysis, Opus for creative vision.

Gemini 3 Pro Image generates ad images. It takes your product photo, a reference ad, and Claude's detailed creative prompt, then produces a new image using your product in the style of the reference — your product, their proven format.

Veo 3.1 generates video clips on Google's Vertex AI. It produces 8-15 second clips in two quality tiers: Standard (720p, fast) and Premium (1080p). Clips can be used as raw footage or layered with text overlays and branding in the editor.

GPT-5.2 handles video prompt engineering. When you're creating video ads, GPT-5.2 takes your product description and generates structured scene prompts, UGC scripts with dialogue, and segments scenes into base clips with extensions — all optimized for what Veo executes best.

The Image Ad Pipeline

The image pipeline is the most common workflow. It takes a product URL, researches competitors, generates copy and images, and outputs ready-to-launch ad creative.

Here's what happens at each step:

The Video Ad Pipeline

The video pipeline follows a similar pattern but uses different models for the visual generation step. Instead of a static image, you get 4-12 second video clips that can be edited with text overlays and branding.

The key differences from the image pipeline:

• Claude Sonnet analyzes the product. It extracts product features from the scraped URL and builds a structured profile used downstream for prompt generation.
• GPT-5.2 writes video prompts. It takes the product profile and generates structured scene descriptions, UGC scripts with dialogue, and segments scenes into 8-second base clips with 7-second extensions — all optimized for what Veo executes best.
• Veo 3.1 generates the footage. Running on Google's Vertex AI, it produces 8-15 second clips in two tiers: Standard (720p, fast) and Premium (1080p). You can generate from a product image as reference or from text prompts alone.
• The Overlay Editor adds branding. Built on Remotion, it lets you add text overlays, logos, transitions, and call-to-action frames on top of the generated video. Preview in real-time before exporting.

The entire video pipeline — from URL to finished, branded video ad — takes about 5-10 minutes. Compare that to traditional video production: days of shooting, editing, revisions, and rendering.

The Creative Suite Menu

The Creative Suite isn't a single page — it's a full workspace with seven sections, each handling a different part of the creative lifecycle. Here's what you get:

Ad Studio

This is where the image pipeline lives. Paste a product URL, research competitors, generate copy and images. Everything starts here for image-based ad creation.

Video Studio

The video pipeline. Analyze your product, generate video prompts with GPT-5.2, create clips with Veo 3.1, and edit them with text overlays and branding in the built-in editor.

AI Tasks

Every session you start in Ad Studio or Video Studio is automatically saved. AI Tasks shows all your saved sessions so you can revisit generated copy, regenerate images with different styles, adjust and refine, and pick up exactly where you left off.

Ads

View all your currently running ads with performance data. See which creative is performing and which is fatiguing. Every ad is scored on four dimensions: Hook, Hold, Click, and Convert.

Media

Your complete media library — every image and video you've generated or uploaded. Browse, search, and select media for new campaigns without re-generating.

Copy

Your saved ad copy library. Every headline, primary text, and description you've saved from AI generation or imported from existing ads. Search by angle, filter by performance, and reuse proven copy across campaigns.

Monitor, Score, Iterate

Creating ads is half the story. The other half is knowing what's working so you can make more of it. The Creative Suite closes this loop with automated scoring and a performance feedback cycle.

Once your ads are live, the system automatically scores every creative on four metrics:

• Hook Score (video only) — Are people stopping to watch? Measured by thumbstop rate.
• Hold Score (video only) — Are they watching through? Combines hold rate and completion rate.
• Click Score — Are they clicking? Weighted blend of CTR and CPC.
• Convert Score — Are they buying? Based on ROAS benchmarks.

Each score is 0-100. Green (75+) means performing well. Amber (50+) is average. Orange (25+) needs attention. Red means it's time to replace that creative.

The Performance Dashboard adds campaign-level verdicts — Scale, Watch, Kill, or Learn — based on your ROAS thresholds. And the Trends page shows 30-day performance trajectories so you can spot fatigue before your ROAS craters.

The loop closes when you take what you learned — which creative scores highest, which angles convert, which styles hook attention — and feed that knowledge back into your next round of AI generation. Use your top-performing ad as the new reference. Let the data tell you what to make next.

Session Persistence: Pick Up Where You Left Off

One underappreciated feature: every AI session is automatically saved. If you generate ad copy on Monday, you can come back on Thursday, open that session in AI Tasks, and generate images for it. No re-entering product URLs. No re-running competitor research. Everything is preserved.

This matters because creative production isn't always a straight line. Sometimes you generate copy, then come back the next day to do images. Sometimes you want to try a different image style on existing copy. Session persistence makes all of that seamless.

Credits, Not Per-Model Billing

The Creative Suite runs on a unified credit system. You don't pay per model or per API call — you get a pool of credits that work across everything.

• Image generation: 5 credits per image
• Video generation (Standard): 20 credits (30 for extended)
• Video generation (Premium): 50 credits (75 for extended)
• Ad copy, analysis, prompts: Included in your plan (no credit cost)

Every plan includes 500 credits. Trial accounts get 25 credits to test the system. If you need more, you can purchase additional credit packs. No surprises, no per-API-call billing, no "you used 47 cents of Veo and 12 cents of Gemini" confusion.

Why Multiple Models Instead of One

There's a reason we don't use a single model for everything. Each model has a specific strength that no other model matches:

The tradeoff is complexity — orchestrating four models is harder than calling one API. But you don't see any of that complexity. The Creative Suite handles the orchestration. You paste a URL and click buttons.

The Full Picture

Zoom out and the Creative Suite is a closed-loop system:

1. Research what's working (competitor ads, your own performance data)
2. Generate new creative (copy, images, video) informed by that research
3. Launch directly to Meta without leaving the platform
4. Monitor performance with automated scoring and verdicts
5. Iterate using performance data to inform the next round of generation

Every step feeds into the next. The more you use it, the better your inputs become, because you're constantly learning what works for your specific product and audience.

This isn't AI replacing your creative instincts. It's AI handling the production so you can focus on strategy. You decide the angle. You pick the reference. You choose the style. The AI executes at a speed and cost that would be impossible with traditional creative production.