Video summary

How to Create UNLIMITED Viral AI Carousels in Claude

Main summary

Key takeaways

Technology

Summary of the Video (Technological Concepts + Tutorial Workflow)

Core idea

Even though Claude doesn’t have an image generator, you can still create Instagram/TikTok “viral AI carousels.” The workflow works by having Claude generate HTML/CSS code for a carousel. Then, that HTML can be rendered and exported/downloaded as individual images for each slide.

Why this works

Claude is being used for what it’s best at: coding and design-system consistency. Instead of generating pixels directly, it generates self-contained HTML for each slide—including fonts and styling—which can be rendered/exported into images.


Tutorial / Guide Steps Described

1. Find winning design “vibes” from existing assets

  • The creator uses images posted on their Instagram/TikTok as references.
  • These images are uploaded to Claude as reference files.
  • The prompt instructs Claude to:
    • Extract the style, design, color, fonts, etc.
    • Ignore written content
  • The output is returned as structured JSON (machine-readable data).

2. Create a Claude Project to standardize carousel generation

  • In Claude’s Projects (similar to ChatGPT Projects), create a project and optionally describe the goal.
  • Paste the previously generated JSON/style-system description into the project instructions.
  • The instructions define Claude as an “Instagram carousel design system” and include rules such as:
    • Generate a fully self-contained swappable HTML carousel
    • Ensure each slide can be exported as an individual image
    • Include a step for collecting brand details before generating
    • Ask which account/brand instance it’s for (useful for multi-account creators)

3. Generate carousel content on demand

  • The user provides newsletter/article content to Claude.
  • Claude:
    • Asks for/uses brand details
    • Verifies the color palette
    • Uses the extracted design system (e.g., “Claude’s palette” if relevant)
    • Produces a slide-by-slide breakdown
    • Outputs the carousel HTML code for multiple slides (example shows seven)

4. Export the slides

  • A carousel preview is shown (including mention of a progress bar).
  • Slides can be downloaded.
  • If downloading fails, a workaround is suggested:
    • Ask Claude to convert everything into PNGs so each slide image can be downloaded individually.

Product / Features / Analysis Highlights

  • Consistency without re-prompting: Once the design-system JSON is embedded into a Claude Project, new carousels can be created with minimal or no additional prompting while staying on-brand.
  • Better typography than image generators: The creator claims the output looks better because the code uses the correct fonts, rather than relying on an image generator that may mishandle typography.
  • Design workflow options:
    • If you already have Canva templates, it’s better to use the Canva connector inside Claude instead of rebuilding from scratch.
    • The Claude design-system workflow is positioned as a fast start for people who don’t already have templates or design systems.

Advice / Meta Guidance from the Speaker

The video cautions against over-engineering AI workflows just because the idea is trendy. If AI can do something, it doesn’t always mean you should build an elaborate system for it.


Main Speaker / Source

Krish (channel owner; “chief AI officer” as the sign-off)

Original video