AI solutions · Generative AI

Text-to-SVG, image-to-SVG and native vector generation

Public SVG corpora are icons and emoji. Models trained on them cannot produce a poster with a headline, a legend and a chart. SVGZO supplies layered, text-rich pages with a normalised twin of every file where the order includes one and three caption lengths.

Talk to the data team

Fields that carry the supervision

svg_raw
svg_normalized
png_448
caption_short / medium / detailed
token_len
complexity_tier

A training recipe

Pair svg_normalized with caption_detailed for text-to-SVG supervised fine-tuning. Use png_448 with svg_normalized for image-to-SVG. Stratify by complexity_tier so the loss curve is not dominated by simple pages.

Evaluation

Buyers typically run DINO, SSIM, LPIPS and PSNR for image-to-SVG; FID, CLIP text-image score and render-validity rate for text-to-SVG; alignment and overlap metrics for layout; and IoU over room polygons for floor plans. A held-out evaluation set stratified by complexity is specified in the order, so these numbers stay comparable between runs.

Why vector data trains better models →