HOMR GUI is a desktop graphical front-end for liebharc/homr, an end-to-end Optical Music Recognition (OMR) system built on vision transformers. Screenshot or drag in an image of a music score, and it converts it into MusicXML and MIDI in a single click — no command line required.
- Graphical Interface — a minimalist PyQt6 desktop app; no more command line.
- One-click conversion — grab a score region with
Win + Shift + Sor drag & drop an image; the score is recognized automatically and exported as.musicxmland.mid. - CPU / GPU mode switch — run on CPU (
onnxruntime) or GPU (onnxruntime-gpu+ CUDA). - China-friendly downloads — model downloads are proxied through a mirror for faster speeds in mainland China.
- Drop straight into your DAW — once converted, simply drag the generated MIDI file into your DAW.
HOMR GUI requires Python 3.11–3.15.
-
Clone the repository:
git clone https://github.com/Quackone/homr_gui.git cd homr_gui -
Install dependencies:
# CPU mode (recommended for most users) pip install -r requirements.txtFor GPU acceleration, you also need CUDA, and install
onnxruntime-gpuin place ofonnxruntime:pip uninstall onnxruntime pip install onnxruntime-gpu
Alternatively, using Poetry:
poetry install # CPU poetry install --with gpu # GPU
Launch in GPU mode:
python homr_gui.pyOr in CPU compatibility mode:
python homr_gui_cpu.pyOn Windows you can also use the bundled launchers:
启动.bat— normal (GPU) launchCPU兼容模式.bat— CPU-safe mode for maximum compatibility调试启动.bat— debug launch with console output
How to use:
- Press
Win + Shift + Sand select the score region, or drag & drop an image onto the window. - The AI engine recognizes the score automatically.
- The results (
.musicxmland.mid) are saved next to the source image. - Drag the
.midfile into your DAW and start editing.
- liebharc/homr — the upstream OMR engine by Christian Liebhardt.
- GUI by Quackone.
This project is licensed under the GNU AGPL v3 license.
HOMR GUI 是基于 liebharc/homr 的桌面图形界面封装与增强版。HOMR 是一个基于视觉 Transformer 的端到端光学乐谱识别(OMR)系统。截图或拖入一张乐谱图片,即可一键转换成 MusicXML 和 MIDI,无需命令行。
- 图形界面 —— 基于 PyQt6 的极简桌面应用,告别命令行。
- 一键转换 —— 用
Win + Shift + S框选乐谱区域,或直接拖入图片,自动识别并导出.musicxml和.mid。 - CPU / GPU 模式切换 —— 支持 CPU(
onnxruntime)与 GPU(onnxruntime-gpu+ CUDA)两种模式。 - 国内下载优化 —— 模型下载走镜像代理,国内下载速度更快。
- 直接拖入 DAW —— 转换完成后,把生成的 MIDI 文件直接拖进 DAW 即可。
HOMR GUI 需要 Python 3.11–3.15。
-
克隆仓库:
git clone https://github.com/Quackone/homr_gui.git cd homr_gui -
安装依赖:
# CPU 模式(推荐大多数用户使用) pip install -r requirements.txt如需 GPU 加速,还需要 CUDA,并用
onnxruntime-gpu替换onnxruntime:pip uninstall onnxruntime pip install onnxruntime-gpu
或使用 Poetry:
poetry install # CPU poetry install --with gpu # GPU
以 GPU 模式启动:
python homr_gui.py或以 CPU 兼容模式启动:
python homr_gui_cpu.py在 Windows 上也可以直接双击自带的启动脚本:
启动.bat—— 常规(GPU)启动CPU兼容模式.bat—— CPU 安全模式,兼容性最好调试启动.bat—— 调试启动,带控制台输出
使用方法:
- 按
Win + Shift + S框选乐谱区域,或将图片拖入窗口。 - AI 引擎自动识别乐谱。
- 识别结果(
.musicxml和.mid)会保存在原图片旁边。 - 把
.mid文件拖进 DAW 即可开始编辑。
- liebharc/homr —— 上游 OMR 引擎,作者 Christian Liebhardt。
- GUI 作者:Quackone。
本项目遵循 GNU AGPL v3 协议。
