REAPER Audio Tag

REAPER Audio Tag is a small REAPER action for fast clip-level audio inspection. Select one audio item, run the action, and get compact local PANNs Cnn14 tags inside REAPER.

Install it with ReaPack, download the ONNX model from inside the REAPER window once, then analyze audio locally without installing Python or leaving your DAW. After analysis, Write Tags to Project can save the result into the selected item notes and create a matching project region.

Current REAPER Audio Tag report window with top cues, tag chips, timing, and CPU/GPU status.

What You Need

• REAPER with ReaPack.
• ReaImGui from ReaPack.
• This repository imported as a custom ReaPack repository.

You do not install Python, create a venv, or choose a model file manually. The plugin uses a self-contained backend installed by ReaPack. The ONNX model is downloaded explicitly on first run and stored in REAPER's data folder.

The model is large: about 327 MB.

Model

This project uses the PANNs Cnn14 AudioSet tagging model by Qiuqiang Kong, Yin Cao, Turab Iqbal, Yuxuan Wang, Wenwu Wang, and Mark D. Plumbley. The release model is an ONNX export of the upstream Cnn14_mAP=0.431.pth checkpoint, converted for local REAPER use.

More details and the maintainer-only conversion workflow are in docs/model.md.

Install

1. Install ReaPack.
2. In REAPER, open Extensions -> ReaPack -> Import repositories.
3. Add this repository URL:

https://github.com/dennech/reaper-audio-tag/raw/main/index.xml

4. Open Extensions -> ReaPack -> Browse packages.
5. Install REAPER Audio Tag.
6. Install ReaImGui: ReaScript binding for Dear ImGui if it is not already installed.
7. Restart REAPER if ReaPack asks you to.

First Run

1. Select exactly one audio item.
2. Run REAPER Audio Tag.
3. Click Download Model.
4. Wait for the download and checksum verification to finish.
5. Run REAPER Audio Tag again, or click Analyze Selected Item if the window is still open.

The model is downloaded from this project's GitHub Release assets:

cnn14_waveform_clipwise_opset17.onnx
sha256 deb65c5a2d291b3ce4ebf2360af71072b789ba11a4214ef77406b89ab97333aa

The model is stored here:

REAPER/Data/reaper-panns-item-report/models/

Platforms

• macOS Apple Silicon and Intel: CoreML is tried first, then CPU fallback.
• Windows x64: DirectML is tried first, then CPU fallback.
• If GPU acceleration is unavailable, analysis still runs on CPU.

The first CoreML run on macOS can be slower because macOS compiles and caches the model.

Public Actions

• REAPER Audio Tag
• REAPER Audio Tag - Debug Export

There is no separate setup or configure action in the public package.

Notes

• FFmpeg is not required for the current public flow.
• Analysis is local after the model has been downloaded.
• The current version analyzes one selected audio item at a time and produces clip-level tags.
• Write Tags to Project writes a visible region over the analyzed item and stores the full tag list in the item notes. It does not rename tracks.

Troubleshooting

See docs/troubleshooting.md.