---
license: apache-2.0
language:
- en
tags:
- gguf
- image-to-svg
- svg
- starvector
pipeline_tag: text-generation
library_name: candle
base_model:
- starvector/starvector-1b-im2svg
---

# StarVector-1B im2svg (Full GGUF Pack)

This repository contains full GGUF exports of **[starvector-1b-im2svg](https://huggingface.co/starvector/starvector-1b-im2svg)**:
- image encoder
- projection adapter
- svg decoder

## Files

| File | Quantization | Approx. Size |
|---|---:|---:|
| `starvector-1b-im2svg-full-f16.gguf` | f16 | ~5.1 GB |
| `starvector-1b-im2svg-full-q8_0.gguf` | q8_0 | ~1.5 GB |
| `starvector-1b-im2svg-full-q4_0.gguf` | q4_0 | ~0.8 GB |

## Description

These are full-model GGUF checkpoints for image-to-SVG generation.

This is **not** a chat LLM.  
The model takes an image as input and generates SVG markup as output.

## Compatibility

These GGUF files are **not compatible** with `llama.cpp`.

Use them only with **starvector-rs**:  
https://github.com/oxide-lab/starvector-rs

## Notes

- The repository contains quantized variants of the same model weights.
- Lower-bit quantization usually reduces memory usage and file size, with a possible quality trade-off.