F13 emerges from stealth with $5M to build editable vector graphics AI for designers

F13, a Berlin-based startup, emerged from stealth with a $5M pre-seed round led by Credo Ventures and Point Nine Capital to build AI models that generate precise, editable vector graphics. The company claims its technology better understands 2D space and delivers visuals with correct proportions and accurate data, a gap it says existing image models fail to fill reliably. For builders who need programmatic generation of charts, diagrams, or brand assets, this could mean the difference between a disposable raster image and a production-ready SVG that can be edited.

What F13's model does differently

The model accepts text, images, or existing graphics as prompts and outputs vector graphics that remain fully editable. That means you can refine individual lines, shapes, or labels after generation instead of starting over. F13 states its proprietary model outperformed general-purpose image and code models from larger AI labs in internal tests, though no independent benchmarks are available yet.

Vector graphics are defined by coordinates and curves rather than pixels, so they scale infinitely and keep each element selectable. This makes them ideal for charts, editorial infographics, scientific diagrams, maps, and brand-aligned layouts where data accuracy and proportion matter more than visual flash. Co-founders Gregory Janik and Ingimar Tomasson bring prior exits and a team that includes researchers from Meta and Porsche and designers who have worked for Versace, Gucci, and Nike.

Practical implications for builders

Early access customers can start using F13 now via an API and a web app to create and edit shareable designs. The funding will cover go-to-market, compute, data acquisition, and hiring as the company develops models for more complex design tasks. A public launch is planned for later this year, and a waitlist is open on the company site.

For teams building agentic workflows or automated content pipelines, an API that returns editable vector graphics could slot into existing design toolchains. Instead of relying on raster outputs from DALL-E or Stable Diffusion and then manually tracing them, you could generate editable brand assets or data visualizations directly. The key question is whether the model's precision holds up at scale and across diverse inputs.

What's still unverified

All performance claims come from F13's own tests. No independent benchmarks or third-party validation were provided in the announcement. The company's descriptions of accuracy and proportion may evolve as the product matures. Additionally, pricing and latency details are not yet public, so it is unclear how the API compares to existing vectorization or generation tools. Builders should treat the current claims as directional and evaluate through the early access program before committing to production use.