Pinterest Affiliate Marketing Has Changed Forever with AI

You are analyzing a downloaded reference video to reverse-engineer it for a new AI-generated video in the same style.

Niche: {niche}
Video file path: {file_path}

Using ffmpeg to sample frames and a transcription API (e.g. AssemblyAI) for the full spoken transcript, do the following:
1. Extract and read the full transcript of the video.
2. Sample frames throughout the video and describe what is visually happening at each key moment (hook, transitions, product reveal, ending).
3. Identify the video's structure: the hook (first 3 seconds), the pacing, the on-screen character (if any), and the product being promoted.
4. Write a condensed 30-second concept for a new video that follows the same hook and structure, adapted to my niche.
5. Write a detailed character design prompt for an AI presenter that fits this niche and concept (appearance, setting, wardrobe, tone).
6. Write three follow-up angle prompts for the same character (3/4 turn, full side profile, rear view, close-up portrait), worded so they can be pasted immediately after the first character image in the same chat to keep the character consistent.
7. Identify and name the specific product featured or implied in the reference video.

Return the transcript, the visual breakdown, the condensed concept, the character prompt, the angle prompts, and the product name, clearly labeled.

Here is a product photo pulled from an e-commerce listing.

Remove all text, dimension callouts, badges, watermarks, and marketing overlays from this image. Do not alter the product itself in any way, keep its exact shape, color, materials, proportions, and position in the frame untouched. The background can be simplified if it contains overlay graphics, but keep it realistic and neutral. Output a clean, photorealistic product image suitable for use as a reference image in a video generation tool.

Using the video concept, hook, pacing, and character details you identified earlier, write a complete video generation prompt for an AI video tool that:

1. Opens with the same hook and pacing as the reference video, adapted to this product.
2. Describes the scene, camera framing, and character actions in concrete, filmable detail, not abstract adjectives.
3. Names the on-screen AI character exactly as it is set up in the video tool, so the tool recognizes it as the character reference rather than plain text.
4. Naturally introduces the product using this confirmed title and description: {product_title} — {product_description}.
5. Ends with a clear call to action appropriate for the platform (e.g. "tap the tag to shop").
6. Is written as one continuous prompt block, 30 seconds of spoken/visual content, ready to paste directly into the video tool.

Keep the prompt specific enough that the generated product closely matches the real reference photos, explicitly instruct the model to closely match the attached product reference images rather than inventing its own version of the product.