the customer

Tasteful visuals at AI scale

‍

Lummi began as an internal experiment to generate websites with AI – but poor, generic images held everything back. So the team built their own curated library of AI-generated visuals, guided by taste, not templates. The result was a platform where creators could find expressive, high-quality images that felt more natural than traditional stock.

“Great images aren’t just about resolution,” says Pablo Stanley, Co-founder of Lummi. “They’re about restraint – knowing what not to show.” Today, Lummi powers bold visual creation for thousands of users across the web.

the challenge

Building at the speed of creativity

‍

As Lummi evolved from an internal tool to a public product, the team faced growing demands:

• Fast turnaround: Users expected instant results, whether customizing formats or stripping backgrounds.

• High image quality: AI-generated images needed to match – or exceed – the polish of traditional photography.

• Creative tooling: Teams wanted more than static downloads; they needed interactivity, video, styling options, and metadata – all at scale.

• Bulk processing: Lummi’s differentiator was its ability to preview transformations across dozens of images simultaneously.

• Small team, big vision: With limited developer resources, Lummi had to find ways to deliver advanced features without bottlenecks.

Cost inefficiencies emerged as traffic grew, straining budgets and requiring a more scalable model.

“The real challenge wasn’t building an AI model,” says Stanley. “It was finding a way to filter, frame, and elevate the results so they actually looked great.”

the solution

Fast, flexible visual tooling for a human-first AI platform

‍

Lummi partnered with Imgix to integrate dynamic image processing directly into its platform. This gave them the infrastructure to offer premium visual experiences – from metadata to manipulation – without bloating their development backlog.

Now, all of Lummi’s images are processed and optimized through Imgix, enabling features such as:

• Background removal: Rolled out in one hour, now one of Lummi’s most viral features

• Smart cropping and aspect ratio adjustment: Powers Reframe, their second most-converting product feature

• Bulk editing and grid-wide previews: Enables dynamic browsing experiences at scale

• Metadata extraction: Tags, color palettes, lighting, and blur hashes all sourced dynamically

The team even uses Imgix for experimental features like image-to-video transformations – a growing demand among Lummi’s social-first creators.

“We’re constantly looking for ways to offer pro-level tools with minimal friction,” says Mariate Vallenilla, Product Manager at Lummi. “The result is a seamless visual experience where powerful features just work – no heavy lifting required.”

the results

More conversions, less complexity

‍

By offloading complex image infrastructure to Imgix, Lummi has turned vision into reality – without ballooning its dev resources.

• 94% average compression: Large, AI-generated originals are compressed from ~980KB to ~57KB on average, enabling lightning-fast experiences.


• 5.3M+ monthly image requests: Lummi’s platform scales effortlessly, whether handling homepage galleries or plugin integrations.


• Reduced time to launch: Background removal and watermarking shipped in hours, not weeks.


 • Higher conversion rates: Reframe, powered by Imgix, became the second highest-converting feature across the product.


Rock-solid performance: Not a single outage reported, even as usage surged.

These outcomes reflect not only technical gains, but also a faster path from concept to value for Lummi’s users.