TimeCues Studio

Explore TimeCues Studio: An Open-Source Tool for Music Annotation

If you're labeling music for research or training models, TimeCues Studio offers a clean, open-source way to do it on a beat-locked timeline.

3 min readMachine Learning
Explore TimeCues Studio: An Open-Source Tool for Music Annotation
ACM Multimedia 2026: TimeCues Studio - a free, open-source tool for annotating music structure [R]

TimeCues Studio is exactly the kind of tool that makes us optimistic about the future of data annotation, and it deserves a close look from anyone who has ever struggled to align music with a timeline. This open-source web app, born from PhD research at ACM Multimedia 2026, tackles a genuinely painful problem: annotating music on a beat-locked timeline while simultaneously testing algorithms against your own labels. The practical implications are immediate for researchers building music-analysis models, developers syncing visuals to audio, or teams training music-to-dance systems. It is a reminder that the most useful innovations often come from people who actually do the work, not from large companies with marketing budgets. For more on how open-source tools are reshaping specialized workflows, see how Explore open-source decision models that run locally and empower your AI agents and Explore how DiffusionEngine unifies diffusion research in one extensible PyTorch framework are following a similar pattern of community-driven progress.

What makes TimeCues Studio stand out is its refusal to treat annotation as a black box. Under your marks, you see separated stems, a spectrogram, and a three-band waveform that colors lows, mids, and highs differently. That design choice means kicks and hi-hats jump out visually, saving hours of guesswork. Detectors like allin1 and MSAF run on the same timeline and get scored against your labels, so you can immediately see where an algorithm succeeds or fails. This is not a toy demo; it is a working research tool that lets you work alone or in teams, export to Audacity labels, MIDI, or JSON, and run it with a single Docker command. The MIT license means there is no gatekeeping, no vendor lock-in, just a well-documented tool that anyone can adapt.

The most telling detail is the demo. You can try it without signing in, no email required. That level of accessibility reflects a deeper philosophy: annotation tools should empower the user, not extract data from them. It is a stark contrast to the usual approach of locking features behind accounts and paywalls. For anyone building music-to-dance models or video-sync pipelines, this is a concrete resource that removes friction from a tedious but essential step. The paper is on arXiv, the code is on GitHub, and the creator is openly asking for feature requests and bug reports. That is the kind of iterative, transparent development that actually moves a field forward.

The open question is how this tool will scale beyond individual research projects. Will it attract a community of contributors who extend it to handle larger datasets or more complex annotation schemas? The foundation is solid, but sustained growth depends on whether other researchers adopt it and feed improvements back into the repository. That is the moment to watch, because a tool this well-crafted deserves to become a standard, not a footnote.

From Machine Learning

Hey all, I wanted to share TimeCues Studio with you. It's part of my PhD work (ACM Multimedia 2026), and I thought you might find it useful. It's an open-source web app for annotating music on a beat-locked timeline and comparing algorithms against your annotations.

If you need to annotate songs (sections, cues, spans, riffs, lyrics and loops), you're building a music-analysis model and want to test it against real labels, you need labeled music for training music-to-dance or video-sync models, or you're syncing visuals to audio, TimeCues Studio might be useful to you.

Read the original at Machine Learning