OmniSVG
omnisvg.net
- Free plan
- macOS
- Linux
- Web
- Self-hosted
What OmniSVG is
OmniSVG converts natural-language descriptions into scalable SVG graphics and can turn supported image files into editable SVG. The image converter accepts PNG, JPG, BMP, and WebP uploads up to 5 MB. Generated vectors can be previewed, downloaded, or copied as code for editing in an SVG-compatible tool. The free demo requires no account, displays the remaining daily generations in the browser, and allows five generations per day. The research project supports text-to-SVG, image-to-SVG, and character-reference generation. It combines a pretrained Qwen-VL vision-language model with an SVG tokenizer and introduces MMSVG-2M, a dataset of two million annotated SVG assets. The code is licensed under Apache License 2.0. For local use, the repository provides Gradio deployment and installation instructions using Python 3.10, Cairo, and CUDA 12.1-tested dependencies. Listed model downloads range from 7.69 GB to 17.2 GB, while reported inference memory use is 17 GB for OmniSVG1.1_4B and 26 GB for OmniSVG1.1_8B. The project also links an online demo and a community-contributed ComfyUI plugin.
Who it suits
OmniSVG suits web developers, UI/UX designers, content creators, and people making print or merchandise graphics. Its local deployment instructions may also interest users who want to run the project themselves.
Good to know, and what to watch
- Creates SVG graphics from text descriptions.
- Converts PNG, JPG, BMP, and WebP images.
- Free demo needs no account.
- Generated SVGs can be downloaded or copied as code.
- Code is licensed under Apache License 2.0.
- Image uploads are limited to 5 MB.
- Free demo allows five generations per day.
- Listed model downloads require 7.69 GB to 17.2 GB.
- Reported inference memory reaches 26 GB.
Plans
Every plan and priceFree
Free plan
Getting started
- Open the OmniSVG web demo
- Enter a natural-language description or upload a supported image
- Generate the SVG and preview the result
- Download the SVG or copy its code
- For local use, follow the repository’s Python 3.10, Cairo, and CUDA 12.1-tested Gradio setup
- Optionally use the listed Hugging Face Spaces demo or ComfyUI plugin
Limits to check
The free demo permits five generations per day and accepts images up to 5 MB. Local inference requirements can include 17 GB or 26 GB of GPU memory, depending on the listed model.
The Club's review
OmniSVG turns text prompts and raster images into editable SVG artwork, with a browser demo and a self-hosting route. It is best for developers, designers, and creators who need vector output and can work within a small daily quota or provide substantial GPU hardware. The main appeal is free, code-friendly generation; the main trade-off is that it is not a full production design suite.
Overview
Natural-language descriptions become scalable vector graphics that you can preview, download, or copy as SVG code for refinement in compatible software. The image converter handles PNG, JPG, BMP, and WebP files up to 5 MB, making it useful for modest source assets but unsuitable for larger raster artwork.
The online demo needs no account and shows your remaining daily generations in the browser. That removes signup friction for occasional work, while the five-generation daily allowance limits repeated exploration and batch conversion.
Key features
Three generation modes
OmniSVG supports Text-to-SVG, Image-to-SVG, and Character-Reference SVG generation. Prompt-based creation suits new icons and illustrations; image conversion preserves a path to editable vectors from supported bitmap files; character references add a workflow for maintaining a visual subject across generated graphics. The underlying system combines a pretrained Qwen-VL vision-language model with an SVG tokenizer, and its MMSVG-2M dataset contains two million richly annotated SVG assets.
Editable export
Results remain usable as vector assets rather than flattened previews. You can download SVG files or copy their code into an SVG-compatible editor, which is valuable when a developer or designer needs to adjust paths, colors, or layout after generation. SVG is the export format, so users needing other native formats will need another conversion step.
Local and community options
The repository provides a Gradio deployment path using Python 3.10, Cairo, and dependencies tested with CUDA 12.1. It also lists a community-contributed ComfyUI OmniSVG plugin. Local operation gives technically capable teams more control than the demo, but the model packages are large: OmniSVG1.1_8B is 17.2 GB, OmniSVG1.1_4B is 7.69 GB, and OmniSVG-3B is 8.49 GB. Reported inference memory is 26 GB GPU RAM for the 8B model and 17 GB for the 4B model, so ordinary laptops and lightly provisioned servers are poor candidates.
Pricing
Free — free (billed Free demo; no account required; five free generations per day; image uploads up to 5 MB). This is a practical way to evaluate the workflows or produce occasional assets. It gives up sustained iteration and high-volume processing because the quota resets daily and the upload ceiling excludes larger images. No paid tier or seat-based option is provided, so teams needing expanded capacity must consider local deployment or another service.
Platforms
OmniSVG is offered on the web and can be self-hosted on Linux or macOS. The web route is the fastest starting point; self-hosting is aimed at users who can install the stated Python, Cairo, and CUDA stack and supply the required GPU memory.
Who it's for
Web developers can copy SVG code directly into projects, while UI/UX designers and content creators can turn prompts or small raster references into editable graphics. Print and merchandise users may find the scalable output useful for artwork that must resize cleanly. It is less suitable for teams that need generous cloud quotas, large-image ingestion, or a conventional multi-tool design workspace.
Pros and cons
- Pros: Text, image, and character-reference generation in one project; editable SVG download and code export; no-account demo; Apache License 2.0 code; web and local deployment choices.
- Cons: Five generations per day on the demo; 5 MB input limit; SVG is the sole export format; local models require multi-gigabyte downloads and up to 26 GB of GPU memory.
Alternatives
Browse AI Vector Generators for a wider comparison. Choose AnnoMotion instead if you want a freemium web, Windows, or API service with credit-based plans. Amnt is a better fit when daily cloud generations, downloads, and agent features matter. Pick GoToVector for a lightweight web workflow with watermarked free sheets and one-off purchases. SVG.new suits users who want a built-in vector editor alongside daily vectorization. Wix Logo Maker is aimed at logo creation across Android, iOS, and web. Breme is another freemium web option. Clearly Logo Maker is a macOS and web alternative. LOGO.com is a web-based logo service with PNG-focused free output.
Verdict
Choose OmniSVG if you want free, editable SVG generation from prompts or small raster files and are comfortable with a five-per-day demo or technical self-hosting. Look elsewhere if you need high-volume online production, larger uploads, broader export formats, or hardware-light deployment.
The verdict
OmniSVG offers text-based vector creation, image conversion, and character-reference generation, with both an online demo and local deployment instructions. The daily demo limit and substantial listed model and memory requirements are worth considering.
Questions people ask
Is OmniSVG free?
Yes. The listed Free plan costs nothing and requires no account for the demo.
What image formats can it convert?
The converter accepts PNG, JPG, BMP, and WebP files up to 5 MB.
What output does it produce?
It produces editable SVG graphics that can be previewed, downloaded, or copied as code.
Can it run locally?
Yes. The repository provides local Gradio deployment instructions using Python 3.10, Cairo, and CUDA 12.1-tested dependencies.
What license covers the code?
The OmniSVG code is licensed under Apache License 2.0.
How many daily generations are included?
The free demo lists five generations per day.
What the maker publishes
- Community integration
- The repository lists a community-contributed ComfyUI OmniSVG plugin. source
- Dataset
- The project introduces MMSVG-2M, a multimodal dataset containing two million richly annotated SVG assets. source
- Editing and export
- Generated SVGs can be previewed, downloaded, or copied as code for editing in an SVG-compatible tool. source
- Generation modalities
- The research project supports Text-to-SVG, Image-to-SVG, and Character-Reference SVG generation. source
- Image conversion
- The image converter accepts PNG, JPG, BMP, and WebP files up to 5 MB and outputs editable SVG. source
- License
- The OmniSVG code is licensed under Apache License 2.0. source
- Local deployment
- The repository provides local Gradio deployment and installation instructions using Python 3.10, Cairo, and CUDA 12.1-tested dependencies. source
- Model architecture
- OmniSVG uses a pretrained Qwen-VL vision-language model together with an SVG tokenizer. source
- Model downloads
- The repository lists OmniSVG1.1_8B at 17.2 GB, OmniSVG1.1_4B at 7.69 GB, and OmniSVG-3B at 8.49 GB. source
- No account
- The free demo does not require an account and displays the remaining daily generations in the browser. source
- Online demo
- The repository links to an online Hugging Face Spaces demo. source
- Performance limit
- The repository reports 26 GB GPU memory usage for OmniSVG1.1_8B inference and 17 GB for OmniSVG1.1_4B. source
- Project affiliations
- The project lists Fudan University and StepFun as the authors’ affiliations. source
- Text generation
- OmniSVG converts natural-language descriptions into clean, scalable SVG graphics. source
- Use cases
- The site lists web developers, UI/UX designers, content creators, and print or merchandise users as target audiences. source
Checked against the maker's pages on 30 Sept 2026.