The Fintech AI Ad Playbook: From JSON Prompts to a 1M-View Launch Video for <$300

Introduction: The $235 Video That Reached a Million People

Imagine spending less on a video ad than you would on a team dinner and having it reach over a million people. It sounds impossible, but it’s the new reality of AI-driven creative. As Will Leatherman, co-founder of Slash, revealed:

“We spent $235.60 on this video. Like that is all we spent. It was just nuts... It's at a million views straight up right now. So, at this cost, it's effectively $0.002 per impression, which is crazy.”

For decades, high-quality video production has been a significant bottleneck for marketing teams. The process is notoriously slow, eye-wateringly expensive, and often results in content that feels too polished or “ad-like” to perform well on authentic, fast-moving platforms like X and LinkedIn. Traditional video ads can cost anywhere from $1,000 to $50,000 to produce, locking out smaller teams and making experimentation a high-stakes gamble.

This playbook changes the game. It’s a soup-to-nuts guide detailing the exact process Slash used to create their viral AI-generated ad for the launch of their ‘Global USD’ feature. We’ll break down the strategy, the production workflow, the precise prompting techniques, and the distribution plan that turned a sub-$300 experiment into a massive marketing win. This isn't about novelty; it's about a replicable system for creating high-impact, low-cost video content that drives real business results.

Part 1: The Strategy — Entertainment, Trust, and Cultural Borrowing

Before diving into the tools and prompts, it’s crucial to understand the strategic foundation. A successful AI video isn’t just a technical achievement; it’s a marketing asset designed to solve a specific business problem.

The Struggle: Breaking Through Fintech's Content Ceiling

Fintech marketers face a unique challenge: they need to capture the attention of a younger, more skeptical audience (like Gen-Z founders) while simultaneously earning the trust of established, older decision-makers. The result is often a content strategy that defaults to the safest, most boring option: dense, educational content that underperforms on social media. Teams are caught between the fear that entertaining content will erode their credibility and the reality that their current content isn't getting seen.

The New Approach: Two-Track Content for Reach and Conversion

Instead of choosing between entertainment and education, this playbook advocates for a two-track system that does both.

1. Track 1: Entertainment for Reach. Use culturally relevant, entertaining AI-generated content to stop the scroll and earn top-of-funnel attention. This is your hook.

2. Track 2: Education for Trust. Follow up immediately with clear, concise product explainers and real demos to build credibility and guide users toward conversion.

The genius of the Slash campaign was pairing their AI-driven, The Big Short-inspired ad with a straightforward animated video that showcased the actual product. This combination is proven to work. A 2024 Trustpilot study found that combining AI creative with real product demos increases consumer trust by 25% in fintech ads.

“We also did start the launch... with like a genuine animated video showing off the real product, like the real feature set. Um, so that probably helped a little bit here.” - Will Leatherman

Culture-Borrowing: Your Unfair Advantage

Why did a spoof of The Big Short work so well for a fintech product launch? Because it leveraged a pre-existing cultural asset. The audience already understands the film's tone, characters, and context. This creates an instant connection and makes the content feel more like a clever piece of commentary than a traditional ad. Your team can create a matrix of cultural touchstones that align with your product’s story (e.g., Ocean's Eleven for security, Grand Theft Auto for breaking norms, etc.).

Part 2: The Production Playbook — A Step-by-Step Guide

This is the tactical, hands-on section that walks you through creating your video.

Step 1: Concept & Meticulous Storyboarding

Your first step isn’t opening an AI tool; it’s planning. Once you’ve chosen your cultural reference, the key is to deconstruct it.

“I basically mapped out every single frame from the movie here. So this is like the opening frame, second cut scene... it just goes through like a few different cuts throughout this scene.”

Map out your video on a shot-for-shot basis, creating a detailed storyboard. For each shot (9+ cuts is a good starting point), define:

• Camera Angle: (e.g., Medium shot, close-up, wide angle)

• Action: (e.g., Character looks at the camera, types on a keyboard)

• Setting: (e.g., A messy office, a sterile server room)

• Dialogue/Voiceover Cue: What is being said at this moment?

This pre-planning is not optional. According to 2024 creator surveys from video generation platforms like Runway, pre-mapping shots reduces generation iterations by an average of 30%, saving you significant time and credits.

Step 2: Assembling Your AI Toolchain

No single tool can do everything. The optimal workflow involves a stack of specialized AI models.

• Image Generation: Midjourney. Use Midjourney (specifically v6 or later) for its superior photorealism and control over style. Its improved SREF (Style Reference) feature is perfect for maintaining a consistent aesthetic across your base images.

• Prompt Engineering: Claude. For complex prompting, especially when dealing with structured data like JSON, Claude’s large context window (up to 1 million tokens in some models) and its ability to interpret structured inputs make it the best choice.

• Video Generation: Runway or Pika. These are leading text-to-video and image-to-video models. You’ll use this to bring your still images to life.

• Image Upscaling: Magnific AI. Before feeding images into your video generator, run them through an upscaler like Magnific. It can enhance resolution up to 4x while intelligently reducing AI artifacts, resulting in a cleaner, more professional final video.

Your workflow will look like this: Midjourney → Magnific (Upscale) → Claude (JSON Prompting) → Runway/Pika (Video Gen).

Step 3: The Breakthrough — Character Consistency via JSON Micro-Measurements

One of the biggest challenges in AI video is maintaining character consistency across different scenes. A character’s face can subtly (or drastically) change from one shot to the next, breaking the illusion. The solution is to move beyond simple descriptive prompts and use highly detailed, structured data.

“The biggest breakthrough I had with this is realizing that the best way to get a consistent character was to give like literal measurements of the facial features of the person... nose, slim, straight, four centimeters long with five degree upturn. Eyes large are almond 3 centimeters by 1.5 centimeters... there is extreme level of detail that is necessary to get the same output if you do this correctly.”

This technique involves creating a JSON object that defines your character with machine-like precision. You feed this same JSON block into your prompt for every single scene featuring that character. Research from Anthropic backs this up, showing that structured JSON inputs improve output consistency by 40% in generative tasks.

Here’s a sample JSON structure you can adapt:

{
  "character": {
    "name": "Alex Chen",
    "age": 28,
    "ethnicity": "East Asian",
    "facial_features": {
      "face_shape": "oval",
      "eyes": {
        "shape": "almond",
        "color": "dark brown",
        "size_cm": {"width": 3, "height": 1.5},
        "eyebrows": "straight, well-defined"
      },
      "nose": {
        "shape": "slim, straight",
        "length_cm": 4,
        "upturn_degrees": 5
      },
      "mouth": {
        "lips": "medium thickness",
        "expression": "neutral, slight smile"
      }
    },
    "hair": {
      "style": "short, textured, side part",
      "color": "black"
    },
    "accessories": [
      {"item": "silver hoop earring", "location": "left ear"}
    ]

This level of detail forces the AI to render the same person every time.

Step 4: Generation, Iteration, and Problem-Solving

With your storyboard and JSON prompts ready, it's time to generate.

1. Generate Base Images: In Midjourney, use your prompts (including the JSON block) to create the key still frames for each shot in your storyboard.

2. Upscale: Run your best images through Magnific AI.

3. Generate Video: In Runway or a similar tool, use the upscaled images as a starting point (image-to-video) or use your detailed text prompt (text-to-video).

Pro Tip: Sometimes, an image-to-video prompt might get flagged for a policy violation (e.g., perceived violence or NSFW content). If this happens, switch to a text-to-video prompt for that shot, but be sure to include strong scene constraints and your character JSON to maintain consistency.

Step 5: The Time & Cost Budget

This entire process is designed for speed and efficiency. Based on the Slash case study:

• Initial Video Cost: <$300 (covering tool subscriptions).

• Initial Video Time: 18-24 hours for the first project.

Crucially, this is a one-time learning curve. Once you have your character JSON and prompt templates, the time required for future projects plummets.

“Now that I've learned all like through the hard way how to do all this, it is going to be so easy to create all these kind of next projects... I'm sure I can make it twice as good in probably half the time.”

According to Gartner, while an initial AI project can take 20-40 hours, reusing prompts can reduce future generation times to just 20-30 minutes per shot.

Part 3: The Distribution Playbook — Launching for Viral Impact

A great video is useless if no one sees it. A coordinated, multi-channel launch is what separates a fun experiment from a strategic win.

The Strategy: An Influencer-Powered Surge

The goal is to create a concentrated burst of conversation. Instead of just posting the video, Slash engineered a launch event.

“We did like a huge launch... we had a bunch of influencers posting about it too. And the launch was extremely successful. I think we got 1.1, 1.2 million impressions on it.”

Pairing your launch with an influencer campaign can boost organic reach by up to 5x. Provide a small group of relevant influencers with early access and a simple messaging kit. Coordinate their posts to go live within the same 1-2 hour window as your main post to maximize initial velocity.

Asset Sequencing for Maximum Impact

Don't just drop the video. Guide your audience through a narrative.

1. Teaser Phase (24h before): Post a short, intriguing clip or a single compelling frame on X and LinkedIn. Hint that something big is coming.

2. The Launch (Day 0): Post the full AI ad. This is your entertainment hook.

3. The Educational Follow-Up (1-2h after launch): In a thread below the main post, share the real product animation. Explain the feature in simple terms.

4. Founder's Narrative (4-6h after launch): Have your founder or a key executive post a thread about why you built the feature and the behind-the-scenes story of creating the ad. This adds a human element and builds authenticity.

Measuring What Matters: The KPI Ladder

Track your success with a clear hierarchy of metrics:

• Top-Level (Reach): Views & Impressions (Goal: 1M+)

• Mid-Level (Engagement): Click-Through Rate (CTR) from the AI ad to your educational thread/product page.

• Bottom-Level (Conversion): Demo sign-ups or product activations attributed to the campaign.

By tracking these, you can directly connect your creative efforts to business outcomes and calculate a ridiculously low Cost Per Impression (CPI), just like Slash's $0.002.

Conclusion: Your New Creative Engine

AI-generated video is no longer a futuristic novelty; it is a practical, powerful tool available to every marketing team, regardless of size or budget. By combining a smart, entertainment-first strategy with a meticulous, technically-sound production process, you can create content that not only captures attention but also builds trust and drives growth.

The playbook laid out here is your roadmap. It demystifies the process, removes the financial risk, and provides a replicable system for turning ideas into high-impact video assets in under 24 hours. The barrier to entry for world-class video creative has been obliterated. The only question left is: what will you create?