Marketing Professionals’ AI Version of Canva?
In 2024, a mysterious image model named “red_panda” surpassed OpenAI’s DALL-E and Midjourney in ratings on the AI analysis platform Artificial Analysis, attracting industry attention and prompting the question: who exactly created this model?
New model on top of Artificial Analysis Image Arena: red_panda. It beats Flux 1.1 pro, ideogram v2, and midjourney v6.1 byu/Gothsim10insingularity
Recraft Emerges Behind the Model
The mysterious image model named red_panda has drawn attention on the AI analysis platform Artificial Analysis, and over time, the startup behind it has surfaced: Recraft. On May 5th, U.S. time, it announced the completion of a $30 million Series B funding round, led by Accel, with participation from Khosla Ventures and Madrona.
Recraft’s founder and CEO Anna Veronika Dorogush revealed that the initial model code-named “red_panda” was inspired by early users continuously generating adorable panda images. Recraft’s AI model is particularly adept at generating promotional images such as posters or brochures for brands, making it akin to a “marketing AI version of Canva.” It currently boasts an annual recurring revenue (ARR) of $5 million and has 4 million users.
Recraft: An AI Tool Tailored for Designers
Recraft is a generative AI tool specifically designed for brands and graphic designers, capable of generating and editing design elements, icons, 3D images, and illustrations while maintaining brand style. Its uniqueness lies in its ability to generate not only raster graphics but also scalable, easily editable vector images, which are crucial for designers. For example, if a designer needs to place a company logo on a curved mug, Recraft allows for easy generation and editing based on the brand’s color palette, making it stand out for its versatility.
Recraft can generate not only raster graphics but also scalable, easily editable vector images.
Image / Recraft Official Website
After just 8 months since its establishment in 2023, Recraft has garnered the favor of 300,000 users. In the Series A funding round of 2024, it secured $12 million led by Khosla Ventures and former GitHub CEO Nat Friedman. The starting point of all this began with Anna Veronika Dorogush’s sister.
“She is a designer, and during the design process, I noticed one thing: she spent a lot of time on tracing. I immediately realized that this should be automated,” Anna Veronika Dorogush told the media outlet Maddyness. From her sister’s work, she identified three key points:
- Vector graphics are crucial in design.
- Designs need to have a consistent style.
- Designers typically need to generate a “set” of designs rather than just a single one.
Anna Veronika Dorogush explained that the main limitation of most generative AI currently is the inability to post-process and edit, which is why Recraft specifically focuses on generating vector graphics and also requires the model to produce a consistent set of images for easier designer workflow.
Testing: Commands in Traditional Chinese, Mockup Performance Lacking
After testing the free version of Recraft, the operation and user interface are quite similar to Canva. The difference lies in that users do not need to start from scratch; they can simply upload images or input commands for the AI to generate the desired images and begin designing.
All command inputs on Recraft can currently be done in Chinese. Recraft impressively handles long commands, although it performs better in English. However, the Mockup feature, commonly used for poster design or product promotion, does not perform well on Recraft, for example, printing logos on faces.
Journalist tested Recraft; Traditional Chinese can also be used, but the Mockup feature is not ideal.
Image / Startup Gathering Custom
In comparison, the free version allows for the generation of only 2 images per month, while the basic, advanced, and professional plans at $10, $27, and $48 a year offer more generation limits and advanced features, including finer image editing and color style conversion.
Image Generation Speed Twice as Fast as OpenAI’s DALL-E 3
Anna Veronika Dorogush has a background in mathematics and engineering, having worked at Microsoft, Google, and Yandex. The open-source machine learning algorithm CatBoost, used for search engines and recommendation systems, was developed by her.
Recraft founder Anna Veronika Dorogush has worked at Microsoft, Google, and Yandex, establishing a strong technical and practical foundation.
Image / Anna Veronika Dorogush LinkedIn
Confident in her technical abilities, Anna Veronika Dorogush chose to develop her own foundational model instead of using open-source models to build the product. This decision has proven competitive, as Recraft received positive reviews on the AI analysis platform Artificial Analysis, and its image generation speed is 7 seconds, twice as fast as OpenAI’s DALL-E 3. Anna Veronika Dorogush admitted that she could have started her business earlier, saying, “I have market sensitivity, product capability, and can manage a team… but I always felt I needed to learn more before starting on my own. Now I believe this is wrong; the best way to learn is to do what you want to do, at least that’s how it is for entrepreneurship.”
This article is a collaborative reprint from: Startup Gathering
The original draft of this article was written by AI.
Sources: TechCrunch, TechBlast, Maddyness, VentureBeat