Recraft.ai
Comprehensive Company Research Report
📅 August 2026 | 🔒 Confidential
1. Company Overview
| Attribute | Details |
|---|
| Company Name | Recraft, Inc. (operating as Recraft AI) |
| Industry | Generative AI · AI Image Generation · Graphic Design Software |
| Founded Date | 2022 (incorporated in the UK on December 6, 2022) |
| Headquarters | 450 Townsend Street, San Francisco, CA 94107, USA |
| Countries of Operation | Global (200+ countries) · Offices in UK & US |
| Company Type | Private |
| Official Website | recraft.ai |
| Mission | To build design-led AI models that empower professional visual creation and brand consistency. |
| Vision | Not Publicly Available |
| Tagline | Not Publicly Available |
Recraft is a generative AI company that distinguishes itself by building proprietary models from scratch to produce design-led, brand-consistent visuals. Recognized as a Leader in the text-to-image platforms market by CB Insights.
2. Founders
Anna Veronika Dorogush
Founder & CEO
Date of Birth
Not Publicly Available
Place of Birth
Russia
Nationality
Russian
Education
Mathematics & Computer Science (top Russian university)
Career Before
Professional Model · ML Engineer at Google, Microsoft, Yandex
Previous Companies
Yandex (Lead ML Engineer), Google, Microsoft
Key Achievements
Co-creator of CatBoost (open-source ML library used by Cloudflare, JetBrains, etc.)
Net Worth
Not Publicly Available
Current Position
Founder & CEO, Recraft Inc.
Background: Dorogush has an unusual blend of creative and technical expertise — she worked as a professional model while earning a rigorous math/CS degree. This combination informs Recraft's "design-first" approach. Her creation of CatBoost at Yandex became one of the world's most popular ML libraries.
3. Founding Story
Why the company was started: Generic AI image generators (DALL-E, Midjourney) fail to meet designers' needs — they struggle with brand consistency, accurate logo placement, readable text rendering, and scalable vector outputs (SVG). Designers were spending hours manually fixing AI-generated assets.
Inspiration: Dorogush's unique background in both machine learning and creative industries gave her insight into this friction. The 2022 text-to-image wave (Stable Diffusion, DALL-E 2, Midjourney) inspired her to enter the space with a design-first focus.
Initial Idea & First Prototype: Recraft was built explicitly "for graphic designers." Early versions focused on vector graphics generation and giving designers more control over style and local edits.
Early Challenges: Designers would compliment the tool but say they wouldn't use it themselves — painful feedback that drove the decision to build proprietary models from scratch.
Funding Journey:
- Jan 2024: $12M Series A (Nat Friedman, Khosla Ventures)
- May 2025: $30M Series B (Accel, Madrona, RTP Global)
- Total: $42M raised
Biggest Turning Points:
- Oct 2024: Recraft V3 ("Red Panda") topped the Artificial Analysis benchmark, beating DALL-E & Midjourney without prior announcement.
- Early 2025: The "Hard Flash" style went unexpectedly viral in China, teaching the team about cultural specificity in user preferences.
4. Company Timeline
2022 Company founded in London; registered in the US.
May 2023 Emerges from stealth; releases public vector graphics generation.
Jan 2024 Raises $12M Series A (Nat Friedman, Khosla Ventures).
Mar 2024 Recraft V2 released (~20B parameters, native vector generation).
Oct 2024 Recraft V3 ("Red Panda") tops Artificial Analysis leaderboard.
Jan 2025 "Hard Flash" style goes viral in China.
May 2025 Raises $30M Series B (Accel, Madrona, RTP Global).
Feb 2026 Recraft V4 released – "ground-up rebuild" with tens of billions of parameters.
Early 2026 Mobile apps launch on iOS and Google Play.
Mid 2026 Recraft V4.1 released (three variants: General, Vector, Utility).
5. Products & Services
Recraft V4 Family
Recraft V4 represents a "ground-up rebuild" of the model architecture, scaling to tens of billions of parameters and optimized for total compute operations (FLOPS) per generation.
| Feature | V4 (General) | V4 Pro | V4 Vector | V4 Utility |
|---|
| Purpose | General creative generation | High-res assets | Native SVG vector generation | Catalog & e-commerce |
| Max Resolution | 1536×1024 | 3072×1536 | Scalable Vector | 1536×1024 |
| Key Use | Concepts, Social Posts | Print, Posters | Logos, Icons, Typography | Product Shots, Mockups |
Key Features:
- Native Vectorization (SVG): Generates editable SVGs via
txt2vec job types.
- Precise Text Rendering:
text_layout parameter for bounding-box text placement.
- Prompt Length: Up to 10,000 characters (10× V3).
- Color Control: Precise palettes and background colors for vector types.
- Speed: Standard ~18s, Pro ~40s, Vector ~28s.
- Export Formats: SVG, PNG, JPG, PDF, TIFF, Lottie.
Recraft Studio
All-in-one creative workspace offering three session layouts: Manual, Agentic, and Exploration modes. Combines AI generation with editing tools (background removal, vectorization, mockups) and built-in video generation.
Mobile App
iOS and Android apps providing access to Recraft's own models plus external models: Nano Banana, GPT Image, Flux, Kling, Veo, and Grok.
Pricing
- Free Plan: Limited daily credits, basic features.
- Pro Plans: Monthly/yearly subscriptions with more credits, private generations, full commercial rights.
- Credit Top-Ups: €2.00 (200), €4.00 (400), €9.00 (800), €19.00 (1600).
- Teams Plan: Collaborative workspaces for teams.
Competitor Comparison
| Feature | Recraft V4 | Midjourney | Adobe Firefly | Ideogram | WaveSpeed |
|---|
| Native Vector (SVG) | ✔ Yes | No | Via CC | No | No |
| Text Rendering | Strong | Poor | Good | Excellent | Strong |
| Video Generation | ✔ Built-in | No | No | No | Yes |
| Pricing | Freemium / $0–$20/mo | Subscription | CC Sub | $8–$96/mo | Pay-per-use |
| Target User | Designers | Artists | Adobe Users | Text-in-image | Developers |
6. Technology Stack
- AI Models: Proprietary, built from scratch. V2 (~20B params), V4 (tens of billions, ground-up rebuild).
- Infrastructure: Trained on NVIDIA Blackwell GPUs via Quantum-2 InfiniBand, using bfloat16 precision.
- Programming Languages: Python (ML), JavaScript/TypeScript (web).
- Cloud Providers: Not Publicly Available (likely AWS or GCP).
- APIs: REST API available via Replicate and fal.ai. (Custom styles not yet API-accessible.)
- Data Sources: Proprietary, copyright-compliant training data.
- Architecture: Modern transformer architecture adapted for vector/raster output (proprietary).
7. Business Model
- Revenue Streams: SaaS subscriptions + credit-based usage.
- Pricing Strategy: Freemium + subscription tiers.
- Customers: Individual creators, design teams, enterprises.
- Enterprise Solutions: Teams plan with collaborative workspaces.
- Consumer Products: Mobile apps (iOS & Android).
- Competitive Advantage: Native vector generation, precise text rendering, brand consistency tools, design-first workflow.
8. Funding History
| Round | Date | Lead Investors | Amount |
|---|
| Series A | Jan 2024 | Nat Friedman, Khosla Ventures | $12M |
| Series B | May 2025 | Accel (Madrona, RTP Global, Khosla, Elad Gil) | $30M |
Total Funding: $42M
Notable Investors: Accel, Khosla Ventures, Madrona, RTP Global, Nat Friedman, Elad Gil, South Park Commons, Basis Set Ventures, Abstract, AAL VC, Edward Lando, Akshay Kothari, Dharmesh Shah.
Current Valuation: Not Publicly Available.
9. Leadership Team
- CEO: Anna Veronika Dorogush (Founder)
- Head of AI: Pavel Ostyakov
- CTO / COO / CFO / VP Eng / Head of Research / Board: Not Publicly Available
10. Financial Information
| Metric | Value | Status |
|---|
| Annual Recurring Revenue (ARR) | Not Publicly Available | Est. ~$5M+ |
| Revenue | Not Publicly Available | — |
| Profit / Loss | Not Publicly Available | — |
| Market Cap | N/A (Private) | — |
| Employee Count | ~55 | Est. May 2026 |
| Growth Rate | Not Publicly Available | — |
11. Competitors
Midjourney
Strengths
Artistic quality, strong brand recognition.
Weaknesses No vector, poor text, no video.
Adobe Firefly
Strengths
Creative Cloud integration, licensed training data.
Weaknesses Vector only via CC, slower speeds.
Ideogram 2.0
Strengths
Excellent text-in-image rendering.
Weaknesses No vector, subscription-only.
WaveSpeed
Strengths
600+ models, video, pay-per-use, full API.
Weaknesses No native vector, less design-focused UI.
12. SWOT Analysis
Strengths
- Top-ranked on design benchmarks
- Native vector + precise text rendering
- $42M from tier-1 VCs
- Founder with deep tech+creative expertise
- Mobile presence (iOS + Android)
Weaknesses
- Single proprietary model (no Flux, etc.)
- Video features less mature than specialists
- API limitations (custom styles not exposed)
- Small team vs. Adobe/OpenAI
Opportunities
- Enterprise design workflow expansion
- Video generation growth
- Third-party integrations
Threats
- Intense competition (Firefly, Canva AI, open-source)
- Rapid tech change — competitors may close the gap
- IP regulation / training data lawsuits
13. AI & Innovation
- Research Areas: Controllable generation, style consistency, vectorization, text rendering, design-first architectures.
- Patents: Not Publicly Available.
- AI Models: V2 (~20B), V3 ("Red Panda", leaderboard-topping), V4/V4.1 (ground-up rebuild, tens of billions of params).
- Open-Source: Founder created CatBoost (open-source); Recraft itself has no open-source projects.
- Breakthrough Innovations: Native SVG vector generation (
txt2vec), precise text rendering with bounding boxes (text_layout), "design-first" AI architecture.
14. Partnerships
- Technology Partners: Available on Replicate and fal.ai.
- Enterprise Customers: Not Publicly Available.
- Government Collaborations: Not Publicly Available.
- Strategic Alliances: Not Publicly Available.
15. Global Presence
- Offices: San Francisco, CA (HQ) · London, UK.
- Countries Served: 200+ countries.
- International Expansion: The "Hard Flash" style went viral in China, highlighting cultural resonance.
- Localization: Supports prompts in English, French, Spanish, Portuguese, German, Italian, Russian, Chinese, Japanese, Korean, Vietnamese, Thai, Arabic.
16. Marketing Strategy
- Branding: "Design-first" positioning as a professional designer's AI tool.
- Content Marketing: Community gallery for inspiration.
- SEO: Targets keywords like AI image generation, vector graphics, logo design.
- Social Media: Active on LinkedIn and Twitter/X.
- Community: Gallery + referral program (currently disabled).
- Developer Relations: API hosting on Replicate; public API documentation.
17. Company Culture
- Work Culture: Founder-driven, technically focused, explicitly anti-grind. Dorogush: "The biggest lesson was that grinding isn't everything. We have to be excellent at what's mission-critical."
- Diversity: Solo female founder/CEO in the male-dominated AI space.
- Remote Work: Likely supports remote work given global offices.
- Values: Excellence in core tech, humor (e.g., 6-fingered hand in the UI).
18. Awards & Recognition
- Artificial Analysis Leaderboard: #1 Ranking (Recraft V3, ELO 1172).
- CB Insights: Named a Leader in text-to-image platforms.
- Expert Collection: Included in CB Insights' AI Expert Collection.
19. Challenges & Controversies
- AI Copyright: A concern for all generative AI companies. Recraft emphasizes copyright compliance and commercial safety.
- Competition: Facing well-funded incumbents with enormous distribution (Adobe, Canva, OpenAI).
- API Limitations: Users report that not all features (e.g., custom styles) are accessible via API.
- Content Moderation: Prompts are filtered for unsafe or sensitive content per Terms of Use.