Year/Date of Release: 2026.10
Version: 2.0.0
Developer: DamRsn
Developer Website: damrsn.github.io/NeuralNote
Format: STANDALONE, VST3
Bit Depth: 64-bit
Crack: Not required
System Requirements:
OS: Win 10+
CPU: x64-compatible CPU
GPU: Supports GPUs with Vulkan (highly recommended for Medium and Large models)
Storage (SSD/HDD): Approximately 210 MB for Small, 620 MB for Medium, or 2.7 GB for Large models
Internet: Required for the initial download of models and for checking updates
Recognition can be performed using either the CPU or the GPU
- 3.36 GB
NeuralNote is a plugin that converts audio recordings into MIDI using artificial intelligence and deep learning technologies.
It allows you to extract notes from audio, recognize multiple instruments simultaneously, and transfer the resulting MIDI parts into your DAW for further editing.
- Audio-to-MIDI: Converts audio to MIDI with automatic note detection.
- Multi-track recognition: Processes entire mixes, separating notes by instrument.
- New “MuScriptor” model: This second version utilizes a more powerful recognition model—developed by Kyutai and Mirelo—replacing the previous Spotify Basic Pitch model.
- Built-in synthesizer: Allows you to audition recognized parts, with individual volume controls, as well as Mute and Solo functions.
- Local processing: Audio is analyzed directly on your computer; no recordings are sent to an external server.
- GPU and CPU support: Computations can be performed using either your graphics card or processor.
- Three model sizes: Small, Medium, and Large—letting you choose the optimal balance between speed and quality.
- MIDI export: Drag and drop finished MIDI parts directly into your DAW project, or save them as files. All three models are included in this release.
Updated: Additional models included.
Notes :
Copy the models to the following path: (C:\Users\UserName\AppData\Roaming\NeuralNote\models),
or install them yourself (NeuralNote downloads models in GGUF format from the
DamRsn/muscriptor-gguf repository on Hugging Face).
MuScriptor is available in three sizes: Small, Medium, and Large. Larger models offer potentially higher recognition quality, but they also require more time and computational resources to run.
