QuiverAI Launches Arrow-1.0 to Generate Editable SVGs
QuiverAI, which has raised $8.3 million in funding, has launched its Arrow-1.0 model for generating editable Scalable Vector Graphics (SVGs) from text or image prompts. The tool is aimed at creative professionals who require precise vector designs for their pipelines. A video demonstration shows the model generating designs via real-time sketching.
QuiverAI's core thesis is that current generative AI models, which primarily produce pixel-based raster images, are insufficient for professional design workflows. The company is instead focused on "visual code generation," treating editable formats like SVG as a structured language that models can reason about and create with precision. This approach moves beyond generating flat images to producing assets with hierarchy and layers that can be programmatically edited and animated. The founding team is composed of AI researchers, engineers, and designers, with Joan Rodríguez, a prominent researcher in SVG generation, at the center. His previous work includes StarVector, one of the first models to generate structured, editable SVGs. The team's public members also include researchers Pascal Wichmann and Haotian Zhang, who have co-authored papers with Rodríguez. Arrow-1.0, currently in public beta, has already achieved the #1 rank on the SVG Arena leaderboard, becoming the first model to surpass an Elo score of 1500 for SVG generation. This benchmark performance indicates a significant step up in the quality and structural integrity of AI-generated vector graphics compared to previous models. The model's API allows for generation from both text and image prompts, with parameters to control the number of outputs, randomness via temperature, and specific style guidance. This level of control is aimed at developers and designers looking to integrate AI-generated assets into their pipelines. Future plans include adding capabilities for direct SVG editing and animation through natural language prompts. The company’s $8.3 million seed round was led by Andreessen Horowitz (a16z), with participation from K Fund, JME, and Mission. The round also attracted angel investors from the creative and developer tool space, including the CEOs of Webflow and Replit, signaling confidence from leaders in adjacent fields. This focus on structured output positions tools like QuiverAI as a new infrastructure layer for visual creation. Instead of a final, static image, the output is a starting point. This aligns with a philosophy of AI as a collaborator, augmenting human creativity by handling the initial generation while leaving the crucial refinement and creative judgment to the designer.
