MixAgent automates music post-production using audio examples and critiques
Agentic Music Post-Production
SoundArtificial Intelligence
Summary
Mixing music after recording is hard because it involves many steps and decisions to get the sound just right. The authors created MixAgent, a tool that mimics how audio engineers work by grouping tracks, choosing effects, and fixing issues based on examples and feedback. They found that simply asking AI models to improve mixes often made them worse, so they made a special dataset and trained a system called MixCritic to better suggest corrections. In tests, MixAgent improved music mixes and was rated better than mixes made by other AI models.
What this means in practice
- •For music audio engineers: Use MixAgent to automate creating and refining music mixes aligned to a reference song with interpretable adjustments.
- •For podcast producers: Improve post-production audio quality by applying group-based effects and systematic refinements derived from professional examples.
Authors
Jaywon Koo, Junghyun Koo, Marco A. Martinez-Ramirez, Wei-Hsiang Liao, Seungheon Doh, Qiyu Wu, Vicente Ordonez, Yuki Mitsufuji
Abstract
We introduce MixAgent, an agentic pipeline for music post-production that follows the workflow of audio engineers to build interpretable and editable mixing plans guided by a reference song. A grouping agent first routes related input tracks, such as drum recordings, into subgroups. A planning agent then uses retrieved examples to choose and order audio effects, such as volume control and reverb, for tracks, subgroups, and the full mix. To refine the mix, a detector identifies what needs correction, and a critic proposes numerical parameter adjustments from tracks to the full mix. In our experiments, prompting recent multimodal models to listen and revise often worsens the mix, making reliable correction a central challenge. We therefore build MixCritic, a dataset of 45,657 repair critiques derived from parameters fitted to professional mixes, and train the critic with supervised fine-tuning followed by preference optimization on rendered audio outcomes. On MedleyDB, learned refinement reduces error by 19% relative to the initial mix, with preference optimization reducing it by 8.4% over supervised training alone. Retrieved examples further improve effects selection, and a human listening test shows that MixAgent gets higher quality ratings than mixes produced by Qwen3-Omni, Gemini 3.8 Flash and GPT-4o-audio.