ClaudeSalesRevenue growthAI score · 64/100AI-assisted edit · no obvious prompt-injection riskX

Claude Studied Ad Patterns to Build a 9:16 Vertical Video Skill, Cutting a 4-Min Testimonial into an Ad

Input: team's existing ads + 4-min raw testimonial; Process: Claude analyzes patterns → ChatGPT generates scenes → Gemini generates music → AI tracks subject; Output: publish-ready 9:16 vertical video ad

Some experienceStepsWorkflowTip
Who
Marketing teams/content creators with raw interview/testimonial footage who need quick social media vertical ads
You get
A publish-ready 9:16 vertical video ad with visuals, music, subject tracking and captions in about 90 minutes
You need
ChatGPT · Gemini

Constraints

Use Claude on desktop to analyze ad structures and build a dedicated skill, ChatGPT to generate/iterate scene visuals, and Gemini (multimodal) to generate soundtrack. Requires raw video footage and team's existing ads as reference. No additional third-party tools. Cost conditions unknown.

What they did

Mike Stelzner's team first recorded a batch of well-done event ads. He then used Claude to analyze their common structure (cold open on face, five-beat arc, super short captions, fight scenes, end card). He asked Claude to create a new skill dedicated only to 9:16 vertical ads. For visuals, Claude struggled with palm tree scenes, so he switched to ChatGPT and iterated until the second mock looked great. For music, he used Gemini's multimodal capability to generate high-energy percussion. Since Sean Cannell moves around a lot, he trained the AI to track him. Finally the AI extracted key points from the raw clip, tracked Sean, and assembled the final ad.

How it went

Author reported the process took about 90 minutes and produced the final ad shown in the video. He noted he now has a system to feed raw video footage to AI and process it on his desktop. (author_report)

Worth knowing

Specific prompt texts not provided; steps are post-hoc operational summaries, not verbatim copyable instructions. The exact interaction method for 'training AI to track subject' is not detailed. Output quality depends on raw footage quality and team's existing ad reference samples. Music generation relies on Gemini's multimodal capability; specific parameters not disclosed. Author did not disclose subscription fees or API costs for each tool. The finished ad in the video serves as visual evidence, but this site has not independently verified the production workflow.

Steps

  1. Analyze team's existing ads with Claude to extract common structure

    Prepare your team's previously recorded event ads and have Claude study them to identify common elements. According to the author's report, Claude recognized: cold open on the face, five-beat arc, super short captions, fight scenes, and end card.

  2. Have Claude create a new skill dedicated to 9:16 vertical video ads

    Based on the analyzed ad structure, ask Claude to create a new skill dedicated solely to producing 9:16 vertical video ads. The skill's only job is to process footage according to this structure and output vertical ads.

  3. Generate and iterate scene visuals with ChatGPT

    When Claude struggles with certain scenes (e.g., palm trees), switch to ChatGPT for visual generation. Prepare prompts describing desired scenes, then iterate based on results: the first generation may be imperfect (the author noted it looked 'too much like Hawaii'), so continue adjusting until satisfied.

  4. Generate high-energy percussion soundtrack with Gemini

    Use Gemini's multimodal capability to generate background music for the ad. Specify the desired style; the author ended up with high energetic percussion. Pair the generated music with visual assets.

  5. Train AI to track subject movement and assemble final ad

    Since the subject in the video (e.g., Sean Cannell) moves around a lot, train the AI to track the specific person. After providing raw footage, the AI will automatically extract key points, track subject movement, and composite all elements (visuals, music, tracking, captions) into the final 9:16 vertical ad.

Prompt you can copy

Reference prompt · Editorial suggestion。Confirm the destination, input variables and output format before use.

This block is the prompt. Ads are never placed here.

No prompt needed — just follow the steps.

This workflow relies on interactive conversation with Claude/ChatGPT/Gemini; no fixed copyable prompt is available. Test the structure analysis with your team's existing ads first before processing target footage.

Source

让 Claude 研究团队优秀广告后建一个只做 9:16 竖版广告的技能，把 4 分钟客户证言剪成广告 · Mike_Stelzner · X · 2026-10-15T10:00:00.000Z

Comments & feedback

Tried this method? Share results, problems or improvements. Comments appear after moderation. Please avoid sensitive information.

Loading…

    Related