MuScriptor: Open-Source Multi-Instrument Transcription
kyutai · hf · 2026-07-15
MuScriptor is an open-weight model designed for multi-instrument music transcription, aiming to solve the issue where existing methods only work well on single instruments and generalize poorly to real-world, complex mixes.
The authors' approach includes:
- Pre-training on synthetic data, followed by fine-tuning with real-world music audio.
- Conducting reinforcement learning-style post-training to enhance real-world applicability.
- Introducing instrument presence conditioning, allowing transcription results to be customized by specific instruments.
The paper/project claims that MuScriptor performs effectively on real-world recordings across various musical genres.
More from Multimodal
- Fake BTS of a dam disaster made with Dreamina Seedance for 29 cents — Standard_Spell5873 · 2026-10-27
- Redditor struggles with outfit swap: Qwen Image 2.1 breaks body proportions — Existing-Box-7866 · 2026-10-27
- Moss Image Turbo 3-Step LoRA Tested: Fast Generation at 6-8 Steps — JadedScorpion · 2026-10-27
- Magnific teases October 28 release: 'We're making it move' — cuenca · 2026-10-27
- Midjourney --sref demo: retro Americana Batman art series — gizakdag · 2026-10-27
- AI video is a new medium, not cinema or animation — and it's perfect for depicting dreams — Kyrannio · 2026-10-27