🤷🏼 Problem:

Everyone from online sellers to visual artists must create appealing images.

This takes a lot of time to create and often requires design skills.

🔧 Solution:

Image generation AI enables users to create custom images for a fraction of the cost.

🎲 Key Players:

• Product Image Generation:

  • Ecomtent - Product photos and Amazon infographics

  • Fashionlabs - Product photos

  • Fotographer.ai - Marketing material product photos

  • Dabble

• Interior Design

  • Interior AI - Generating 45K MRR with a 99% profit margin

  • Collov

  • Spacely

  • Room AI

• Headshots and Images of Humans

  • Photo AI - Making 50K ARR with one person

  • Aragon - Creating professional-looking headshots from casual photos

  • HeadshotsByAI - Same as above

  • HeadshotPro - Same as above

• Logo Designs

  • Logo Pony

  • Brandmark

🙌 What Sets Top Companies Apart:

• They offer tools where the work occurs. E.g. if the user does the work in the browser, a tool is delivered as a Chrome extension.

• They ship very fast. The velocity in the space is too high for lame ducks.

• They place a special focus on retention by

  • Focus on onboarding, support, and customer success. Optimize pricing to incentivize long-term usage, etc.

• They understand their ICP well and verticalize ruthlessly. This also helps their retention.

• They track the right KPIs:

  • Measuring adoption by looking at the share of paying users.

  • Look at the ratio of weekly over monthly active users (WAU-MAU ratio) to determine stickiness.

  • They find the right north star metric. E.g. a product might deliver tons of value but is only used once a month.

• They become defensible through:

  • Marketplace dynamics.

    • E.g. ElevenLabs has a marketplace for creator voices

  • Using the shareability of images to grow organically

    • E.g. people using comic versions of themselves as profile images on social media

🎯 Opportunities:

Idea 1) Rendering For Kitchen Studios

• Problem: Kitchen studios use design software to plan kitchens with their customers. The renderings don’t look very lively. This makes it hard for customers to imagine the final product.

• Solution: Create niche architecture rendering software for kitchen studios.

• What we like:

  • Kitchen studios plan dozens of kitchens every day with their customers → Creates value repeatedly

  • A kitchen is a high-ticket item. If better images increase conversion, this creates a lot of value

  • Kitchens are a high-value niche within the interior design niche. This lets you understand your customers well and find long-tail keywords with cheap CPAs.

• What we don’t like:

  • Kitchen studios require a longer sales process

Idea 2) Improve Listings For Realtors

• Problem: Realtors create tons of listings. Sometimes the apartment is not empty. Sometimes it is still under construction. In other cases, construction might not even have started yet.

• Solution: An image generation tool that solves one or more of the cases above and helps realtors post better images and increase sales.

• What we like:

  • Creates value repeatedly, high ticket item, …

  • Chance of creating virality. Images will need a disclaimer that they have been generated. Other realtors will see that and thereby learn about your product.

• What we don’t like:

  • The realtor market is very fragmented. There are many small and a few big players.

Idea 3) Niche Info Graphic Generator

• Problem: Online merchants create infographics for each product they sell. This requires not just an image but also design skills.

• Solution: A niche image generation tool e.g. in fashion or sportswear like caspa.ai

• What we like:

  • Creates value repeatedly

  • Niching down allows us to understand the ICP very well

  • This can probably be sold through ads and SEO and doesn’t require in-person sales meetings

• What we don’t like:

  • Low-ticket item. At a price of ~€20/month, it will be hard to acquire customers

  • The space is quite crowded

⚡️ Potential Risks:

• Hordes of AI tourists lead to high signup rates but low retention

• Free users can generate significant costs. Going viral can put you in the hole.

💡 Takeaways:

• Building on community platforms like Discord can be an advantage. The sharing of generated images, prompts, etc. can be great to increase stickiness and build community around the product.

  • Midjourney is a prime example of this

• How can image generation apps build moats?

  • The ML model is rarely the moat

  • Moats are built the same way traditional software businesses built moats.

  • Understanding the user/problem better than others

  • Focus on SEO

  • Build a brand

  • Exploit marketplace dynamics and product-led growth

• A general framework for finding niche image generation ideas:

  • Identify high-value use case

  • Look for repeating use cases

  • See if you can exploit viral dynamics. Images are inherently sharable. This can help us.

  • Try to spot marketplace dynamics. We can e.g. help influencers create and sell their own images. This incentivizes them to promote your tool.

📚 Additional Resources

1. The GenAI 100: AI Apps That Stick (Podcast)

2. Hacker News Post on Pieter Levels’ Image AI App

3. List of Generative AI Startups