Recraft V3 is Recraft's second-generation image model. It adds 20+ specialized style presets, native vector output, long multi-word text generation with precise positioning, and improved anatomical accuracy over Recraft V2.
import { experimental_generateImage as generateImage } from 'ai';
const result = await generateImage({ model: 'recraft/recraft-v3', prompt: 'A red balloon on a wooden table.'});Recraft V3 is Recraft's second-generation image model. It expands the V2 foundation with improvements in style control, text generation, and anatomical accuracy.
The most visible addition is the style library: 20+ specialized presets covering realistic portraits, digital illustrations, pixel art, vector graphics, and more. These presets give you direct control over visual direction without prompt engineering workarounds. Select a preset and the model applies that aesthetic consistently.
Text generation also improved in V3. The model handles long multi-word text with precise positioning within compositions. Combined with anatomically correct output (correct proportions, natural poses, realistic hands and faces), V3 produces images that need less post-production correction.
Native vector output is another V3 addition. Generate scalable vector graphics directly for logos, icons, and print assets. No need to rasterize and trace. This fits into professional design toolchains where vector formats are the standard.