Recraft - Your Full-Stack Creative OS 🎨

Plus: Founder/CEO Anna Veronika on model innovation and Recraft’s $30m Series B round...

May 13, 2025

CV Deep Dive

Today, we’re talking with Anna Veronika, Founder and CEO of Recraft AI.

Recraft is an AI-native design tool built for professional creatives— those working across branding, illustration, vector design, and marketing collateral. Founded in 2022 by Anna, Recraft’s platform serves as a full-stack creative OS: a place where designers can create, refine, and scale branded content with high fidelity and precision. The platform is purpose-built for control and consistency - offering brand-safe generation workflows, SVG outputs, and support for real-world design constraints. 

Recraft has been innovating on the model side from the start. Last October, under the name red_panda, Recraft’s V3 model beat out competitor models from Midjourney, Black Forest Labs, and OpenAI on the Artificial Analysis benchmark - placing 40 Elo points ahead of Black Forest Labs’ Flux1.1 Pro, on the text-to-image leaderboard.  The London-based team of 30 has developed one of the world’s top-performing diffusion models from scratch—built entirely in-house— and launched its brand orchestration engine, enabling designers to create and apply visual brand systems across all image types with near-instant rendering. 

This week, Recraft announced a $30M Series B round led by Accel - to grow its AI research and go-to-market efforts. Today, the platform is utilized by a wide variety of creatives - freelance brand designers, in-house teams at companies like Shopify and Salesforce, and large enterprises producing marketing and visuals at scale. 

In this conversation, Anna shares how Recraft approaches model training differently than competitors, why most AI tools don’t serve creative professionals, and why the future of design software is a fusion of AI and human.

Let’s dive in ⚡️

Read time: 8 mins

Our Chat with Anna 💬

Anna, welcome to Cerebral Valley! First off, introduce yourself and give us a bit of background on you and Recraft. What led you to co-found Recraft?

Hey there! I’m Anna Veronika, and I’m a computer scientist and engineer by background—both in education and in my career. I’ve worked at a few large companies, including Google, Microsoft, and Yandex, before starting Recraft about two and a half years ago. 

Now I’m fully focused on building Recraft. We’re creating a professional tool for designers that helps them use AI in their work, specifically within the graphic design space. Our goal is to empower designers to fully harness the potential of AI—amplifying their creativity, enhancing the impact of their work, and accelerating their process when speed matters..

How would you describe Recraft to the uninitiated developer or AI team?

There are two parts to what we’re building. One part is the models—we’re building our own models from scratch. We shipped Recraft V3 a few months ago, and it topped the public leaderboards. We’re training our models specifically for professional-grade use cases. 

The second part is the workflow. What differentiates us from every other image generation tool is that our models are fine-tuned for professional scenarios, and our workflow supports that. Examples of how we fine-tune the models for professional tasks include the control features and design features we’re building into the models. The control features let you go beyond just text-to-image—they allow you to provide visual hints to get the exact output you need. That’s critical in the professional space, where you have brand guidelines and strict requirements. 

We’ve built support for brands into the model so you can define and work within a specific brand style. There’s a lot of control over how you develop that style, and you can experiment with different inputs into the model to refine it. You can build a style from just a few examples. Then you can reweight those examples or add additional prompt engineering on top. The goal is to help you achieve the brand look and feel you’re aiming for. We’ve trained the model to understand that, so you don’t need to retrain anything—it just takes additional inputs. We support brand styles across photography, illustration, icons, and all kinds of visuals. 

We also offer control over positioning. If you’re designing a poster or marketing materials, for example, you can place elements exactly where you want them, thanks to our workflow. It’s not just text-to-image. You can position things within a poster however you need. We support brand style control, positioning control, and brand colors too. If you have a defined brand palette, you can generate an image using those colors or edit an existing image by adding elements in your brand’s specific color. That’s a frequent use case.

Lastly, there are the design features. Since our users are designers, we support scenarios that are specific to design workflows. For instance, we support generating vector art—we don’t just output raster images or photos, we can also generate SVGs. We also support AI mockups. You can turn any image into a mockup, whether it’s a T-shirt, a mug, or something else. Just drag and drop your design on top. Mockups are really important when you want to present a brand or design concept, or if you are producing designs for print. 

So overall, there are two sides to the product: control features and design features.

Tell us about your key users today. Who would you say is finding the most value in what your team is building with Recraft? 

Right now we have a pretty wide variety of users overall. A lot of people try the tool just out of curiosity—they want to test something new, generate art for themselves, or experiment. But the users who stick with us and keep coming back are mostly professionals, especially professional designers.

There are a few distinct groups within that. One group is freelance designers who work on a range of projects for different brands. They’re creating logos, brand identities, and more. Another group is design teams—these might be small teams of two or three people in startups or larger teams at bigger companies like Salesforce and Shopify.

For those larger companies, brand consistency is a big deal. They care about things like brand identity, brand colors, and having the right workflow support. We also have a lot of users in the print space—people working on merchandise. For them, features like AI mockups and vector art are essential.

So our primary user groups are design teams across a wide range of company sizes, as well as freelance brand designers. 

Walk us through Recraft’s new release. Which use-cases should designers experiment with first, and how easy is it for them to get started? 

Our new release builds on what we’ve already done with brand consistency - but adds a whole set of new tools you can use to create or scale a brand identity.

First, there’s now an infinite library of AI styles that you can explore. You can browse through a wide range of styles—photorealism, illustration, vector art—and search for very specific looks. There are also style recommendations to help you find something aligned with the aesthetic you need. You can visually identify a style and then immediately start using it.

We’ve added new tools for creating styles as well. You can now mix images with weights, or combine styles with other styles. There’s a way to experiment and see how different changes in style affect the resulting images. We also released a new style model. Now there are two: one that captures the basics like texture, color, and linework, and another that captures more nuanced aspects like art direction, composition, and contextual elements. So you’ve got several models to choose from when working with style. 

There’s a lot you can do when building your brand identity, and it’s all in one place. You just open the style creation panel and use a few simple controls that unlock a lot of flexibility. Recraft has the only model with true brand consistency, meaning you can create as many illustrations as you want in the same consistent style, or generate a full photo shoot with consistent lighting, color, and styling. That’s really the core of what we launched in this release.

There are a number of teams pushing the boundaries on AI x creativity and design. What would you say sets Recraft apart from a product or technical perspective? 

In terms of technology, we are building our own models from scratch. The models we are building are extremely high quality, and that's an important differentiation for designers because designers do not compromise on quality.

The second one is the control and design capabilities I mentioned. Other image generation companies don’t provide those because their users don’t care that much about them. Control and consistency are not the things that are most important for general exploration, but they are very important specifically for design scenarios—and we are training our models to support that. So that is a technical innovation that other models do not provide. Those are the major differences: the quality, the control features, and the design features that we’ve built into the model.

Now, how we’re different from other design companies: we’re building an AI-native tool. It’s built from scratch with all the AI scenarios in mind. We offer an infinite canvas for image generation, editing, mockups and graphic designs—powered entirely by AI. When we build features, we’re aware of what AI is capable of.

Could you share how Recraft’s team splits AI research and product development? What are the innovations behind some of the architectural decisions you’ve made with V3? 

We do have a machine learning and research team that’s training the models for the product. It’s one team—it’s a research team doing research, and the results of that research are the product we’re building.

We’ve come up with a number of innovations. Some features we were the first to develop—figuring out how to solve something with diffusion models. In other cases, we just have unique ideas or novel ways of approaching problems that we use inside our models. All of that is research, but as a startup, we don’t do research just for the sake of it. We do research because we have a goal: to advance technology to solve the specific needs our users have.

In our case, the most important priorities are continually increasing quality—so users can consistently get the results they actually want—and developing new ways to control the model. Coming up with new interfaces and controls is really important in the AI era, so people can express more clearly to AI what it is they’re trying to achieve.

2025 is the “Year of the AI Agent”. How do agents factor into Recraft’s product roadmap, and are we going to see a design agent anytime soon? 

One of the things we do is provide a design tool that’s actively used by designers. We also offer an API because we train our model from scratch, and the model is very high quality with a bunch of unique features that aren’t available anywhere else—like support for brand consistency, which is really important for some users.

We provide this functionality via API, and some companies are using it to scale image generation for ads, marketing materials, websites—there are a few large categories where people need image generation at scale. So we have users tapping into that through our API.

Now, one way to use the API is by calling it through code. Another way, which we’re planning to release very soon, is through an agentic interface. So there’s the design tool we’re building, and then there’s the API access to the model itself, which can be used either directly or eventually through agents like MCP.

You’ve just announced a $30m Series B round led by Accel and a number of your previous investors. What are your plans for this next phase of Recraft? 

The new funding is going to be used first for research and advancement of the product. As I said, we see our goal as providing full control to users over the results they're getting, which requires new technology and new models. That’s one big area we’re working on. The second is putting more resources into go-to-market and marketing. That’s the next priority. Thirdly, we’ll also be expanding the team and hiring additional people to support product development.

Overall, the main goal for us next year is to build the right product and improve the technology so that our users can solve their tasks with very high quality in Recraft.

Lastly, tell us a bit about the team at Recraft. How would you describe your culture, and are you hiring? What do you look for in prospective team members joining the Recraft? 

We’re a team of 30, and most of us are based in London working from the office. We have an in-office culture. As you can see from public benchmarks and leaderboards, with the team and resources we had, we were able to build a very high-quality model that topped the charts—even against companies that had significantly more funding and larger teams at the time. That’s still true today.

The reason we were able to train such a strong model with limited resources comes down to the level of talent on the team. We have a very high hiring bar and a super strong engineering culture, and that’s really been the foundation for Recraft’s technical success.

Now, with the new funding, we’ve started hiring across the board. We’re growing the machine learning team, backend and frontend engineering, and investing more into go-to-market as well. So we’re hiring across a bunch of roles and are definitely open to applications.

Conclusion

Stay up to date on the latest with Recraft, learn more here.

Read our past few Deep Dives below:

If you would like us to ‘Deep Dive’ a founder, team or product launch, please reply to this email ([email protected]) or DM us on Twitter or LinkedIn.