#6969th of 70 in Open modelsFor image, video and design, image and design
star-vector
Vision-language model that converts images and text into SVG code
- Stars
- 4.6k
- License
- Apache-2.0
- Last commit
- Nov 2025
- Language
- Python
4,339 on 21 Apr 20264,639 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Before you pick it
- No commits in a year. The last commit was on 2025-11-07; fixes and security patches may not come. Source (opens in a new tab), checked .
Overview
StarVector is a multimodal model built on StarCoder that generates SVG code directly from an image (image2SVG) or a text prompt (text2SVG). The repo has inference code, 1B and 8B checkpoints on Hugging Face, training scripts (DeepSpeed for 1B, FSDP for 8B), the SVG-Stack dataset and the SVG-Bench evaluation suite. A Gradio demo runs on the HF backend or a vLLM fork.
Who it is for: Researchers and developers vectorizing icons, logos and diagrams into SVG code
Strengths
- Outputs SVG code using primitives like ellipses, polygons and text, not only path curves
- Ships 1B and 8B checkpoints plus SVG-Stack (2.1M train) and SVG-Bench datasets
- Includes training and evaluation scripts, with HF and vLLM generation backends
- 8B has the highest listed DinoScore on all five SVG-Bench sets in the README table
Weaknesses
- Not trained on natural images or illustrations; suited to icons, logos, diagrams, charts
- vLLM backend needs a separate StarVector fork of vLLM
- Demo needs three processes (controller, web server, worker) started by hand
- Reference training commands assume 8 GPUs
What it needs
- GPU required
- Docker
- Needs PyTorch, Hugging Face Transformers, Gradio, vLLM fork (optional), DeepSpeed and Accelerate (training)
- Models: starvector-1b-im2svg, starvector-8b-im2svg, Hugging Face Transformers, vLLM
- port 7000
Quick answers
Is star-vector open source?
Yes. Its code is on GitHub under the Apache-2.0 license.
Can you run star-vector with Docker?
Yes. Its repository has a Docker setup: a Dockerfile, a Compose file or a published image.
Does star-vector need a GPU?
Yes, it needs a GPU.
Also in Open models
See all 70| Rank | Project | Score |
|---|---|---|
| 1 | ruviewWiFi CSI sensing on ESP32 for presence, breathing, heart rate and pose | 94 out of 100 |
| 2 | layaSingle-pass multilingual classifier for typed choice, score and yes/no decisions | 88 out of 100 |
| 3 | rf-detrReal-time transformer models for object detection, segmentation and keypoints | 79 out of 100 |
| #4 | timesfmPretrained time-series foundation model for zero-shot forecasting from Google Research | 78 out of 100 |
| #5 | yolov5PyTorch YOLOv5 object detection, segmentation and classification models with training code | 76 out of 100 |