EN
Back to the archive

The archive · Developer & Business Tools · Product decision · 2022–2025

Recraft bets brands need controlled AI imagery; V3 tops benchmark, $30M Series B

London AI design tool with vector output and brand style controls; $42M raised, 4M users, V3 beat DALL-E and Midjourney on Artificial Analysis.

Recraft

The betThat professionals will pay for AI that obeys brand rules — vector output, style controls, exact logo placement — and a lean in-house model can beat image-gen giants.Scaling

What the business is

AI design platform generating brand-consistent raster and vector images — logos, icons, marketing assets — with style controls, powered by its own Recraft V3 foundation model.

Starting capital$12M Series A led by Khosla Ventures and Nat Friedman (Jan 2024); $30M Series B led by Accel (May 2025)

How it started

Anna Veronika Dorogush — creator of the CatBoost ML library and a former Yandex, Google and Microsoft researcher — founded Recraft in 2022 with its main office in London. She bet professional designers, not consumers, were the underserved market for generative imagery, and launched on Product Hunt in May 2023.

What happened

Recraft reached 300,000 users in eight months and raised a $12M Series A led by Khosla Ventures and Nat Friedman in Jan 2024 to build its own foundation model instead of fine-tuning Stable Diffusion. It grew to 2.5M users including teams at Netflix and Airbus, then released V3 (codenamed red_panda), which topped Artificial Analysis' text-to-image leaderboard in Oct 2024, beating DALL-E and Midjourney. In May 2025 it raised a $30M Series B led by Accel.

How it ended up

Still scaling: 4M+ users and $5M+ ARR as of May 2025, with brand-control features (Advanced Style Remixing, Infinite Style Library) shipped alongside the Series B.

Background

Recraft is a generative AI design platform, founded in 2022 by Anna Veronika Dorogush — creator of the CatBoost machine-learning library and a former Yandex, Google and Microsoft researcher — with its main office in London. It generates brand-consistent raster and vector images, icons and marketing assets with style controls, powered by its own foundation model.

The bet was that professional designers were underserved: consumer tools like Midjourney produce striking rasters that can't be edited, resized or kept on-brand. Recraft launched vector-first on Product Hunt in May 2023, grew to 300,000 users in eight months, and raised a $12M Series A led by Khosla Ventures and Nat Friedman in Jan 2024 to build its own model rather than fine-tune Stable Diffusion.

Recraft V3, codenamed red_panda, topped the Artificial Analysis text-to-image leaderboard in Oct 2024 — beating DALL-E and Midjourney — and in May 2025 the company raised a $30M Series B led by Accel, reporting 4M+ users and $5M+ ARR, with customers including teams at Amazon, Salesforce, Netflix and Airbus.

With its Series B, Recraft shipped Advanced Style Remixing and an Infinite Style Library aimed at brand teams, positioning itself between image generators (Midjourney, Adobe Firefly, Black Forest Labs) and design platforms like Canva, while keeping a deliberately small team and a Discord community as its growth channel.

What has to be true

  • Professional designers need editable, on-brand assets, and generic generators can't reliably place logos or hold a style across a campaign.
  • Vector-first output was a technical edge competitors ignored, making Recraft's images usable in real design workflows.
  • Building its own model freed Recraft from Stable Diffusion's limits and let a small team top a public benchmark against giants.
  • A lean team plus a low-friction Discord bot turned community into an acquisition channel twice the size of its X following.

What can be applied

In generative AI, differentiation can come from the output contract, not model size: vector output, style control and brand fidelity made a small team's tool the one professionals pay for.

Aftermath

As of May 2025, Recraft was scaling with 4M+ users and $5M+ ARR, $42M raised in total, and new brand-consistency tools (style remixing, Infinite Style Library) launched with the Accel-led Series B. It continues to compete with Midjourney, Adobe Firefly, Stable Diffusion and Black Forest Labs while aiming squarely at professional branding workflows.

Sources

spotted an error? The archive wants to know.

Your turn

You just read one. Describe what you are building, and see who is betting on the same thing.

Free account · 3 free questions · no card

Related cases