ANTIMATTER project is a NEUTRINO derivation that improves the usability of NEUTRINO voices. It support a full fledged editor, where it delivers a new way of editing songs that makes NEUTRINO vocal editing competitive with major players like SynthenizerV and Cevio. With the uniformity limit of the model, regardless, it still break the limit of what you can do with MIDI or MusicXML.

Defying the limits of most major synth provider, AMADEUS editor has the capability to incoporate the close suport for multilingual singing, starting with the close prototype that embeds cephome - the Vietnamese transcoder in its heart. Built by one who loves Vietnamese vocaloid-producing, it first delivers the ease of translation, moving into the field of cheating the language barrier and voice models to deliver absolutely best pronounciation quality. The engine is constantly being tested against children songs in the textbooks due to nature of *MusicXML* format to deliver classical music, but it will soon be a real deal to modern youth music.

With the rebuilding, the voice stack can be observed, tuned as soon as the engine yields out. It allows modification that defying the limit of the voice, while its playback experience possibly has the gap to tune vocal mixes without leaving for the DAW.

AMADEUS also built with performance in mind, providing a smooth editing experience and playback with vanishing delay, smooth as hell rendering with graphics acceleration, without the burden of browser. It is widely compatible (yet), with a curated set of platform that plays, whenether you are on a crappy Windows or a swift Wayland desktop. Moreover, it supports more acceleration patches to model that support running on OPENVINO, allowing marginal performance boost on CPU by collapsing up the model for more throughput.

Overall, we made a music editor so we have total control of what we can do with the engine that can streamline our language, and making it having actual joy while using it. We hope you enjoy AMADEUS as much as we do.

---

System requirements:

- CPU: Only desktop processor is supported (x86_64), which does not includes ARM processor
+ Intel: Haswell microarchitecture (4th generation Core processors, 2013) and newer
+ AMD: Excavator microarchitecture (2015) and newer. This includes all subsequent families like Ryzen, EPYC, and Zen-based Athlon
- OS:
+ Windows 10 or Windows Server 2016+
+ Linux kernel 3.2+, glibc 2.17+
- RAM: At least 4GB of free memory to load inference models. Might work with system with at least 8GB, although we recommend atleast 16GB system. 
- Disk: At least 800MB for application if most component are installed. Voices model size are not included. Please note that project file are also cost a fortune because of compressed lossless cache.
- Acceleration:
+ Graphics acceleration: Support DX12 or Vulkan or OpenGL
+ CUDA enabled system is prefered, minmum version is CUDA 10. More than 4GB of VRAM is recommended to use this acceleration.
+ Intel system is prefered for OPENVINO acceleration. NPU acceleration is yet supported (it is marginal *un*gain and suffers much incompatibility).
+ Base inference runtime might work with above CPU configuration with no acceleration.
