All News DISPATCH COMPUTE

Recraft V4: image generation with design taste

Recraft V4 moves beyond standard AI image generation by prioritizing design-centric compositions and vector outputs. Filmmakers and designers can now generate high-quality assets with accurate text and infinite scalability.

Replicate

Recraft V4 is now available via Replicate, marking a shift in how AI models handle graphic design and vector art. While most image generators focus on photographic realism, this model prioritizes art direction, layout, and the technical requirements of professional designers. For creators who need more than just a raster image, the ability to generate editable SVGs directly from a prompt changes the utility of AI in a production pipeline.

What's new

Recraft V4 introduces several technical improvements that address common pain points in AI-assisted design. The most significant update is the model's ability to generate vector graphics (SVG) that remain fully editable in software like Adobe Illustrator or Figma. Unlike standard upscalers or tracers, this produces clean paths and points rather than just a high-resolution grid of pixels.

Text rendering has also seen a major upgrade. The model can now place specific, legible text within an image or design, following the requested font style and placement with high accuracy. Additionally, the model offers better control over composition and brand consistency through a feature that allows users to define a specific set of colors and styles before generation (see the provider's announcement).

How it fits your workflow

For filmmakers and production designers, Recraft V4 serves as a fast-track for creating in-world assets. If you need a logo for a fictional brand, a poster for a background set piece, or specific UI elements for a screen replacement, this tool replaces the need for rough sketching or fighting with Midjourney’s unpredictable text. Because the output can be an SVG, a motion designer can take the file directly into After Effects and animate individual layers without manual rotoscoping or recreating the asset from scratch.

Storyboard artists can use the model to maintain a consistent visual style across a series of frames. By setting a specific art direction, you ensure that characters and environments retain their look throughout the pre-visualization process. While tools like DALL-E 3 have improved text, Recraft V4 is more aligned with the needs of a graphic designer who requires control over the final file format and geometric precision.

Editors and social media managers can also use the model to generate thumbnails and promotional graphics. The ability to dictate layout and text placement means less time spent in Canva or Photoshop fixing AI hallucinations and more time on the actual creative edit.

What it costs / how to try it

You can run Recraft V4 on Replicate using their standard hardware-based pricing model, which charges based on the compute time used for each generation. Detailed pricing per second of execution is available on the Replicate model page.

Read the original announcement on Replicate ↗

Help keep this running

Your tip funds servers, models, and the time it takes to ship new tools faster. Set any amount below — every bit helps.