Base44 Superagents-AI agent that does it all
Mini toolsSVG T2I
SVG-T2I extends the SVG framework to perform diffusion in VFM representation space, using a frozen vision encoder plus a trainable SVG autoencoder and DiT. By operating on VFM features, it unifies visual understanding and generation, supporting image editing, retrieval and multimodal alignment. The project fully open-sources the autoencoder, diffusion models, training and evaluation code and pre-trained checkpoints.
New
Image
Gen 4
Overview
SVG-T2I is a text-to-image diffusion model that generates directly in Visual Foundation Model feature space (e.g. DINOv3) instead of pixels or VAE latents, achieving high-fidelity images and competitive GenEval and DPG-Bench scores.
About Kuaishou Technology
Chinese short-video & live-streaming platform (Kwai) with e-commerce and ads.
View Company ProfileTools using SVG T2I
No tools found for this model yet.
Last updated: February 25, 2026
