Skip to content
repository radar
Voice, Vision & Multimodal

🖌️OmniSVG

ACTIVE STEADY

OmniSVG/OmniSVG

[NeurIPS 2025] OmniSVG is the first family of end-to-end multimodal SVG generators that leverage pre-trained Vision-Language Models (VLMs), capable of generating complex and detailed SVGs, from simple icons to intricate anime characters.

⭐ Popular: 3k stars, gaining about 6 a week

View on GitHub ↗

repo profile

vintage 2025 1 year old
language Python
license Apache-2.0

momentum

total stars 3k
stars added last week +6
commits / week 0 steady
issues closed 14% ~16d to close, median
contributors 7
release cadence n/a
last activity 5mo ago

durability

backing community / independent
openness permissive

since we covered it

monthly average + 85/mo (+7%/mo) · + 1k total since PR#6

why it's a big deal

  • Generates SVGs from text prompts or images, covering the range from simple icons to detailed anime characters, so designers and developers can produce editable vector art without drawing paths by hand.
  • Ships three open model variants (an 8B and 4B at 17GB and 8GB, plus a legacy 3B), letting users pick a size that fits their hardware.
  • Releases the MMSVG-2M dataset of 2M annotated SVGs and the MMSVGBench evaluation set of 600 held-out samples, giving the field shared training and benchmarking resources.

under the hood

  • Built end-to-end on pre-trained Vision-Language Models, handling both text-to-SVG and image-to-SVG as multimodal generation tasks.
  • Represents SVGs as its own token format rather than raw XML, with the README timing generation per OmniSVG SVG token versus XML code token.
  • Runs on Python with PyTorch 2.3.0 and CUDA 12.1, and uses the Cairo library to render generated SVGs.

Radar summary, generated from the project's public sources

star history

PR#6 · 1k3k now Apr 2025Aug 2026
  1. PR#6 1k 2025-04-16
  2. now 3k + 1k since first covered

curve is sampled from GitHub's star history, plus our own daily readings since we covered it; the dashed stretch is before we first covered it, the solid line since. figures at coverage are the numbers we printed then (approx.), current count is live.

understory

Better known than its recent output, coasting a little on attention.

-37 understory score output 0 · clout 37
Aug 2025 Jul 2026
  • output, commits & releases
  • clout, star velocity

output = commits & releases; clout = star velocity, both 0 to 100 monthly indices; the gap where output runs above clout is the understory. The understory →

covered in

  • PR#6 2025-04-16 below the radar

    A multimodal generator that uses VLMs to create detailed SVGs, from simple icons to complex anime characters

similar projects

compare these →
  • 💡 PaddleOCR

    34× the stars

    Turn any PDF or image document into structured data for your AI. A powerful, lightweight OCR toolkit that bridges the gap between images/PDFs and LLMs. Supports 100+ languages.

    88k ACTIVE
  • Unsloth

    28× the stars

    Local UI to run and train LLMs and diffusion models, including Qwen3.8, Kimi K3, MiniMax-H3, Gemma 4, DeepSeek-V4, FLUX and more.

    72k ACTIVE
  • 🎙️ Real-Time-Voice-Cloning

    23× the stars

    Clone a voice in 5 seconds to generate arbitrary speech in real-time

    60k ACTIVE

comments

Sign in with GitHub to add your blip on OmniSVG.

loading comments…