Kyutai Releases MuScriptor: An Open-Weight Decoder-Only Transformer for Multi-Instrument Music Transcription to MIDI
Kyutai and Mirelo released MuScriptor, an open-weight decoder-only Transformer that transcribes multi-instrument music into MIDI files. The model uses a three-stage pipeline and was trained on 170,000 real recordings plus 1.45 million synthetic MIDIs.
- MuScriptor is an open-weight decoder-only Transformer from Kyutai and Mirelo for transcribing multi-instrument music into MIDI.
- The model was trained on 170,000 real recordings and 1.45 million synthetic MIDIs, using a three-stage pipeline.
- Benchmarks show MuScriptor outperforms YourMT3+ in instrument conditioning and transcription accuracy.
- An interactive demo is available to test the model's capabilities directly.
Kyutai, in collaboration with Mirelo, has introduced MuScriptor, an open-weight decoder-only Transformer designed to transcribe complex multi-instrument music recordings directly into MIDI format. The model leverages a three-stage processing pipeline to handle the intricacies of mixed audio, ensuring accurate transcription across instruments. Trained on a dataset comprising 170,000 real-world recordings and 1.45 million synthetic MIDI files, MuScriptor aims to bridge the gap between raw audio and editable musical notation.
The release includes benchmarks comparing MuScriptor against YourMT3+, highlighting its performance in instrument conditioning and transcription accuracy. Kyutai has also provided an interactive explainer demo to showcase the model's capabilities, allowing users to test its functionality firsthand. This open-weight approach lowers the barrier for researchers and developers to experiment with and build upon the technology, potentially accelerating innovation in music AI applications.
The model's focus on multi-instrument transcription addresses a longstanding challenge in music processing, where separating and transcribing individual instruments from mixed audio has historically been difficult. By releasing MuScriptor under an open-weight license, Kyutai is contributing to the broader AI and music communities, fostering collaboration and further research in this niche yet impactful domain.
Provides an open-weight tool for music transcription, enabling new applications and research in AI-driven music processing.
Offers a potential solution for music production, licensing, and AI-assisted composition tools.
Serves as a practical example of applying Transformer models to real-world audio processing tasks.
Expands accessibility to advanced music transcription technology through open-source release.
- MIDI
- Musical Instrument Digital Interface, a protocol for communicating musical information between devices.
- open-weight
- A model whose parameters are publicly available, allowing users to inspect, modify, or fine-tune the model.
AI ToolsHow We Test an AI Product Without Burning Credit
OpenAI's Browser Isn't Dead, It Just Moved To The ChatGPT App - Engadget
OpenAI Will Shutter Its ChatGPT Browser Atlas in August - CNET
AI ToolsOpenAI staffer maps out which of GPT-5.6 Sol's five reasoning levels fits which task complexity
Astellas Deploys Structure Prediction AI "Boltz-2 NIM" to Democratize and Accelerate Drug Discovery - NVIDIA
Florida lawsuit against OpenAI heads to federal court in Fort Pierce - Yahoo
A Florida lawsuit against OpenAI will proceed in federal court in Fort Pierce, marking a significant legal escalation in privacy and data use disputes involving AI technologies.
News outlets urge a judge to sanction OpenAI in a high-stakes AI copyright fight - Times Leader
A coalition of news publishers has formally requested a judge to impose sanctions on OpenAI for alleged copyright violations in its AI training practices.
Musk tells Tesla staff to switch to Grok — a model he admits is worse - Electrek
Elon Musk has directed Tesla employees to use Grok for internal tasks, even though he acknowledged the model performs worse than alternatives.
BusinessRed Hat will support your RHEL forever now - for a price
Red Hat now offers a paid Long-Life Add-On to extend support for specific RHEL releases indefinitely, ensuring long-term stability for enterprises.
BusinessApple Is Suing OpenAI for Allegedly Stealing Hardware Secrets
Apple has filed a lawsuit against OpenAI, accusing the company of encouraging poached employees to steal confidential hardware designs and supplier details.
AI in battlefield intelligence: Expanding the speed of decision-making - Federal News Network
The U.S. military is integrating AI tools to speed up battlefield intelligence and decision-making processes, according to Federal News Network.