MAGDA - Free Open Source DAW with Integrated AI (v0.9.0)

Explore how Machine Learning and AI can expand musical creativity while keeping the human in the creative workflow. This forum is dedicated to respectful dialogue where diverse perspectives are welcomed.
RELATED
PRODUCTS

Post

0.9.0 is out - the media database release.

The headline: MAGDA can now index your sample folders and presets and search them by sound.

The Library is a new database icon in the right-panel browser. Right-click a folder in Disk mode, Index this folder, tag the scan if you want, and it runs in the background. Audio is analysed for tempo, key, family, shape, duration, and tags (from the file and the folder names around it). MAGDA presets get indexed through their metadata, so .mps and plugin-native presets share the index.

Search:
  • Filters work as soon as a folder is indexed - family, shape, key, tempo, tags, no model required.
  • Download the optional Sample Analyzer model from Settings, AI Settings and the search bar takes plain language. Describe a sound, press enter, results rank by similarity.
  • Right-click any result and Find similar sounds to re-rank by closeness to that sound.
Other stuff:
  • Drum grid - rename rows, save and load kits, factory kits included
  • Drummer agent - select a track with a drum grid and the AI assistant switches to drummer mode; ask for a pattern, it writes into the kit
  • External audio editor - right-click an audio clip, Edit in External Editor, MAGDA passes the source to the app set in Preferences, edits come back
  • MIDI editor - per-clip vertical zoom on the piano roll and drum grid, a fullscreen toggle, and a quick view switch between the two
  • Renamable macros and modulators, and drum grid pad envelopes are modulation targets
MAGDA is now also on MuseHub.

Video walkthrough:
Devices catalogue: magda.land/devices
Download: magda.land
Source: GitHub

Free and open source under GPL v3.
MAGDA - Free Open Source DAW with Integrated AI | GitHub

Post

0.10.0 is out - the first of two mixing releases.

The headline: two new analysis devices, a post-FX area in every chain, gain staging, and a stack of mixer improvements.

Analysis tools:
  • Oscilloscope - live waveform that locks to the trace so it stays still, with a timebase and a trace colour
  • Spectrum Analyzer - log-frequency plot with peak hold, an adjustable slope tilt, and smoothing
  • Drop either anywhere in a chain, pop it out into its own always-on-top window, and they show up as mini views on the mixer strips
Post-FX area - a new stage in the chain that runs after the main effects and before the fader, on every track and the master. The natural home for meters and analyzers, but any effect works there too.

Mixer:
  • A toggle rail to show only the rows you want - sends, I/O, monitor, oscilloscope, spectrum, FX chain
  • A mini FX chain on every strip, with the exact parameters you choose
  • Mini scopes on the strips that expand for controls and pop out for the full view
  • The fader now sits over its peak meter, with a dB readout
Gain staging - set a target peak, play a section, stop. MAGDA trims each device's output to hold your headroom, only attenuating, carrying the gain forward, and marking every move. Optional AI pass can weigh limiters and saturators instead of a flat target.

Customisation - keyboard shortcuts are remappable, and mouse/trackpad gestures (scroll, zoom, track height) are configurable. Bindings are saved between sessions.

Also:
  • Live MIDI input lights up the piano roll and drum grid as you play
  • Arrow-key auditioning in the sample browser, with left/right transport
  • A restyled mixer in the Session view
  • Auto-hide arrangement scrollbars, new footer icons, and the usual fixes
Video walkthrough:
Devices catalogue: magda.land/devices
Download: magda.land
MuseHub: musehub.com
Source: GitHub

Free and open source under GPL v3.
MAGDA - Free Open Source DAW with Integrated AI | GitHub

Post

DAW looks cool and will try it out, but wanted to ask you a couple of questions, if you dont mind:

1. Is every folder path that writes something to the system disk configurable or stated in the preferences, or are there some hidden automatic ones: something like bitwig's index and cache folders that produce hundreds of thousands of folders and files or melodyne's separations etc... I hope everything is transparent in the folder structures.

2. Are all processes inside the DAW non-destructive(changing original samples when pitching/stretching or editing). If there are some destructive commands, are there dialogs to let you know something will change?

3. Is the program messing with the OS power schemes or anything else OS related(installing additional software/drivers included)?

4. Does the AI functions train itself in any form on a user material/sound and methods and ways of working etc... or sending data somewhere?

5. Is the ASIO support ready, i saw in earlier posts it is not?

thanks,

Post

LSN wrote: Thu Jun 11, 2026 9:41 pm DAW looks cool and will try it out, but wanted to ask you a couple of questions, if you dont mind:

1. Is every folder path that writes something to the system disk configurable or stated in the preferences, or are there some hidden automatic ones: something like bitwig's index and cache folders that produce hundreds of thousands of folders and files or melodyne's separations etc... I hope everything is transparent in the folder structures.

2. Are all processes inside the DAW non-destructive(changing original samples when pitching/stretching or editing). If there are some destructive commands, are there dialogs to let you know something will change?

3. Is the program messing with the OS power schemes or anything else OS related(installing additional software/drivers included)?

4. Does the AI functions train itself in any form on a user material/sound and methods and ways of working etc... or sending data somewhere?

5. Is the ASIO support ready, i saw in earlier posts it is not?

thanks,
Thanks for the thoughtful questions.

1. MAGDA's user-writable folders are exposed in Preferences: Data, Presets, and Render. There is also a small waveform peak cache used for drawing waveforms, which is safe to delete and rebuilds automatically. The goal is to keep folder structure transparent, not to silently create huge hidden databases.

2. Audio editing is intended to be non-destructive: pitching, stretching, clipping, warping, and normal clip editing do not alter the original sample files. Operations that create audio, such as render, bounce, freeze, or export, write separate files. If you do want destructive editing, you can open the sample in an external editor of your choice; since MAGDA never modifies the source files itself, the two workflows coexist cleanly.

3. MAGDA does not change OS power schemes or install extra system drivers or background services.

4. MAGDA does not train on user projects, samples, prompts, or workflow data. We do use LoRA training internally for MAGDA's own agents/models, but that is based on our own/internal training material, not user material. Cloud AI calls only happen if the user configures a provider/API key; local AI can run offline.

5. Yes, ASIO support is fully working. It was only missing in the very first 0.1.0 release; the current release is 0.10.2, and the audio stack has moved a long way since then.
MAGDA - Free Open Source DAW with Integrated AI | GitHub

Post

Thanks for the thorough answers - seems transparent and safe. Where can I report bugs or give suggestions, if possible?

Post

LSN wrote: Fri Jun 12, 2026 11:34 am Thanks for the thorough answers - seems transparent and safe. Where can I report bugs or give suggestions, if possible?
https://github.com/Conceptual-Machines/ ... ore/issues :tu:
MAGDA - Free Open Source DAW with Integrated AI | GitHub

Post Reply

Return to “Machine Learning and AI for Music Creation”