One thing i have in my mind is outputting the individual strings to allow individual effects on them. The plugin does internally a form of (very rudimentary) harmonic source separation. However to do this I need to implement some machine learning.
I’m investigating anyhow in how to identify the frets being played with neural nets. That’s a precursor to the string separation.
Once I have the fret identification working, that info could be output in some form
First I’ll try and bring the plugin as it is now to the duo
Then the dwarf when I get my hands on it
Hi. Ive been busy working on a neural net inside the plugin. It will be released in some weeks/months. Here’s a video demoing the current state: #audiodsp #guitartomidi #deeplearning #music #audiotranscription #realtimeaudiotranscription #guitar #guitarmidi #automaticmusictranscription #dsp #vst #plugins | Gerald Mwangi
Any chance this will also work for Bass on deeper notes?
great to see this @jimsondrift it seems pretty promising!
Wouldn’t you be up to releasing a test version for the MOD so that the MOD community could eventually help you with feedback and bug reporting?
in principle it could work, but the model would have to be trained on bass data. thats a massive undertaking. I tried it on my bass and it doesnt work
id like to but the plugin it too compute heavy currently. Ill will produce a release for linux soon though. The issue I am facing is that I am lacking nvme storage, i have only 3tb and need more of it to train the model
and the prices are too high thanks to uncle sam and his tech bros.
I get it! Let’s hope you find a solution soon.
Maybe someone from the community can give you some help. Don’t be shy to create a thread asking
Thats a good idea. Im planning to release a Version for linux after the current training round. Then ill reach out to the community to port it to the dwarf
That will be next Week, training takes 5-10 days on 52hrs of nonstop guitar audio!!!
it would be a great practice session if you were required to actually live-play all that time 
Maybe the way to finance the project would be a guiness award of most time playing guitar non-stop and at the same time, you would train the system 
Announcing GuitarMidi-lv2 version 2.0 for linux. This version is a massive upgrade incorporating a custom neural net Release on May the 8th on github: GitHub - geraldmwangi/GuitarMidi-LV2: A concept for guitar to midi as an lv2 plugin · GitHub Stay tuned
That sounds great - recognizing individual notes and chords. 
Are you planning to include that in the DuoX as well?
No. I dont have the duox so cant test. But once its on the dwarf, ill callout for hell regarding duox.
It is quite a behemoth of plugin so it will take time
GuitarMidi-LV2 v2.0 for linux is out:
I highlighted the current drawbacks in the release notes. Happy testing
I need help porting this plugin to the mod dwarf and the other devices
Here the announcement video:
Thanks for sharing your achievement.
It works on Debian13 with guitarmidi-lv2_2.0_ubuntu2404_amd64.deb, though I mistakenly set buffer size as 128 on MOD Desktop at first.
I could successfully build from source too.
However, when I tried building the Windows version with PawPaw, the build failed because pango could not be found.
It would be great if Windows MOD Desktop users could use it as well.
By the way, what’s the difference between ubuntu2404.deb and ubuntu2604.deb?
no functional difference. Its just that the build for ubuntu 26.04 depends on more recent system libs (glibc).
Does it work on linux for you
Great! Excited to use it once it’s available for Mod Duo X. Thanks for your great work!
Only the Ubuntu 24.04 build works on Debian 13.
The Ubuntu 26.04 build doesn’t work.
and how is the performance for you. is it detecting anything?