---
title: flux-high-res
canonical_url: "https://www.modelscope.cn/models/MusePublic/flux-high-res"
md_url: "https://www.modelscope.cn/models/MusePublic/flux-high-res.md"
repository: MusePublic/flux-high-res
chinese_name: "FLUX.1-dev HighRes"
last_updated: 2025-06-24
license: "Apache License 2.0"
pipeline_tag: text-to-image-synthesis
tasks:
  - text-to-image-synthesis
base_model:
  - AI-ModelScope/FLUX.1-dev
base_model_relation: finetune
parameters: 11.9B
tensor_type:
  - BF16
library_name:
  - lora
  - safetensors
  - pytorch
frameworks:
  - Pytorch
supports_inference: txt2img
downloads: 4324
stars: 93
tags:
  - Checkpoint
  - text-to-image
---

# flux-high-res

> flux-high-res - MusePublic 在 ModelScope 开源的模型。flux.1-dev HighRes 是一个基于 flux.1-dev 微调得到的模型。通过在数百万张 2048 分辨率左右各类尺寸的图片上进行微调，该模型在 2048、2560 等分辨率下的图像生成质量显著优于原始的 flux.1-dev。同时相比使用 flux.1-dev 生成 1024 分辨率图片并使用 Real-ESRGAN 等模型进行图像超分辨率，flux.1-dev HighRes 生成的原生 2k…

MusePublic/flux-high-res 是 ModelScope 魔搭社区上的 11.9B 参数text-to-image-synthesis模型，采用 Apache License 2.0 许可，基于 AI-ModelScope/FLUX.1-dev 构建，并支持在线推理（txt2img）。

- **Repository**: MusePublic/flux-high-res
- **License**: Apache License 2.0
- **Tasks**: text-to-image-synthesis
- **Parameters**: 11.9B
- **Base model**: AI-ModelScope/FLUX.1-dev
- **Online inference**: txt2img
- **Tags**: Checkpoint, text-to-image
- **Downloads**: 4324
- **Stars**: 93
- **Last updated**: 2025-06-24

Source: https://www.modelscope.cn/models/MusePublic/flux-high-res

---

# 模型介绍

flux.1-dev HighRes 是一个基于 flux.1-dev 微调得到的模型。通过在数百万张 2048 分辨率左右各类尺寸的图片上进行微调，该模型在 2048、2560 等分辨率下的图像生成质量显著优于原始的 flux.1-dev。同时相比使用 flux.1-dev 生成 1024 分辨率图片并使用 Real-ESRGAN 等模型进行图像超分辨率，flux.1-dev HighRes 生成的原生 2k 分辨率图片在细节、纹理、边缘上的质量显著更优。

flux.1-dev HighRes is a model fine-tuned from flux.1-dev. By training on millions of images with resolutions around 2048p and various aspect ratios, it achieves significantly better image generation quality at resolutions like 2048 and 2560 compared to the original flux.1-dev. Furthermore, when compared to generating 1024p images using flux.1-dev and then applying super-resolution models such as Real-ESRGAN, flux.1-dev HighRes produces native 2K-resolution images with substantially better quality in terms of details, textures, and edges.

转发模型请明确注明来源：https://www.modelscope.cn/models/MusePublic/flux-high-res