top of page
Search

Fadr Launches DrumGPT: Text-to-Instrument AI Drum Generator

  • Writer: Ezra Sandzer-Bell
    Ezra Sandzer-Bell
  • 1 day ago
  • 5 min read

Updated: 7 hours ago

Fadr launched their new VST and web application DrumGPT near the end of April 2025. The plugin is a companion to their popular SynthGPT plugin, both centered on the use of text prompts to generate never-before-heard virtual instruments.


They are the first and only company to offer text-to-instrument synthesis in the DAW, along with bonus tools for stem extraction and remixing. We've shared a detailed summary of their SynthGPT plugin in a previous article here.



Each text prompt generates a collection of sixteen sounds, each mapped to a MIDI key in a chromatic sequence. They can be triggered by a standard MIDI keyboard, finger-drummed with instruments like the Ableton Push, or programmed by hand inside your DAW's piano roll.


Fadr provided us with an NFR copy of DrumGPT, shared technical specs about the software, and covered our time spent testing the application. All opinions in this article are my own.

Table of Contents



The Basics: How to use Fadr DrumGPT


DrumGPT home screen

DrumGPT's home screen, shown above, marks the entry point where all of the text prompting happens. You just describe the kind of drum kit that you're looking for and Fadr's AI model will generate 16 fully original one-shots.


DrumGPT Library screen

Each of the kits you generate will be stored permanently inside the Library tab. You can click the heart button on any kit to earmark it. Filter out the mediocre kits by hitting the "favorites" switch to only show the kits that you liked. Alternatively, you can hit the trash icon to delete any model permanently.


DrumGPT ADSR and trigger mapping screen

You'll spend most of your time in the Drum tab, shown above. This view can be broken into roughly three sections annotated above. The bottom strip highlighted in orange are your trigger pads, the green strip are your ADSR and parameter controls, and the top section holds the waveform with an envelope curve.


Seasoned music producers will look at this interface and understand what all of it means. The knobs are used to shape your sound's attack, decay, sustain, and release, known as ADSR parameters for short.


As you turn the ADSR knobs left or right, the envelope curve (the red line) will adjust automatically. You can click and drag the envelope to change the knob positioning as well. They are dual-directional in that sense.


The pitch knob controls the wavelength (horizontal length) of the waveform. As you raise the pitch, the wavelength shortens and the frequency of the drum hit gets higher. Vice versa, lowering the pitch will cause the wavelength to elongate.


Each drum hit has its own dedicated parameters, so changes to one sound don't impact the others.


Fine tuning one-shots to get the sound you want

Adjusting ADSR and pitch in DrumGPT

Your text prompts in DrumGPT should place you somewhere in the ballpark of what you're looking for. In my experience, the pitch and ADSR parameters help close that gap and arrive at the target quickly.


The example above highlights a scenario where I prompted for "spacey synthwave drums" and dialed in the snares to achieve the sound I wanted. I liked the "Snare 2" sound from this kit but needed to tune it to match the key of my song, so I dialed up the pitch.


I also wanted to get that characteristic vaporwave snare sound, so I maxed out the release parameter. These changes are automatically saved by Fadr, so when I come back to the Library and load the kit next time, it's tuned up the way I want it.


Using DrumGPT with a step sequencer


Once you've tuned your kit in DrumGPT, it's time to make music. Here's a quick walkthrough on how you can use the plugin in combination with a MIDI step sequencer. I've used Logic Pro 11 for this demo but the same principle will apply in any DAW.


Using DrumGPT with a step sequencer in Logic Pro 11

Logic's step sequencer can be accessed by right clicking on the MIDI track and selecting the second option from the dropdown menu: Create Pattern Region. This will open up a programmable drum machine where you would otherwise typically see a piano roll.


Each drum lane in the DAW's pattern sequencer will be assigned to a pitch. Click on the pitch name to access a dropdown. DrumGPT defaults to the 16-pitch range of C2 through D#3. So you'll need to decide which drum hits you want to expose to the sequencer and assign them to each lane accordingly.


Make things easier on yourself and rename those lanes to the drum type (snare, hi hate, kick, etc) so you don't have to guess what's what.


Combine DrumGPT with SynthGPT + AudioCipher

Combining DrumGPT with SynthGPT and AudioCipher in Logic Pro

Now that you've got a drum pattern in place, consider loading Fadr's other plugin, SynthGPT. It's included automatically with your DrumGPT purchase and covers all the tonal instruments you'll need for generating chords and melodies.


In the example above, I've loaded an instance of SynthGPT v2 on a second MIDI track and prompted for "vaporwave synth keys". I've used my AudioCiphers text-to-MIDI VST (sold separately) to generate a chord progression from the word "chill". We could have used any word but this one set the right intention.


AudioCipher is a great compliment to Fadr's products, as they handle the virtual instruments and we generate melodies and chord progressions from words, using a centuries-old encryption technique called music cryptograms.



Ethical model training and audio quality specs


DrumGPT's AI audio model was ethically trained. The company paid artists to license the one-shot audio assets used in their dataset. They were not sourced from creative commons, or scraped from the internet.


The plugin outputs audio at 44.1Khz sample rate, and supports playback at any sample rate. My personal experience with the kits were that they sounded high quality and lacked some of the troublesome high end noise that other AI diffusion models often introduce.


Is DrumGPT a ChatGPT Plugin?


No, DrumGPT is a DAW plugin, not a ChatGPT plugin. The service does not load inside of OpenAI. It's loads inside a DAW as a VST3 on Windows and Mac, or as an AU component format for Mac users.


The confusion here may stem from the fact that OpenAI has a "ChatGPT plugin" directory. Plugins in this case refer to fine tuned models within ChatGPT, issued by third party developers. These are silly, low quality text generation experiences that use plain text to "generate" chord charts and lyrics.

ChatGPT MIDI chord progression generator

We've covered the most popular ChatGPT music plugins here, with screenshots and details to illuminate how they work, mostly so that you don't have to waste your time. You can see an example of a "ChatGPT melody generator" in the screenshot above, illustrating how ineffective they are.


DrumGPT is an actual digital audio plugin whereas ChatGPT "plugins" do not interface with a DAW in any way.


Are there any good alternatives DrumGPT?


Fadr is currently the only commercial text-to-instrument plugin company.


Native Instruments, known for their studio grade virtual instrument libraries, did publish a research paper in 2024 that demonstrated a working text-to-instrument model. However, there's no sign that Native Instruments will release that model commercially.



A senior manager from Native Instruments confirmed to me privately that they were concerned a model of this kind would cannibalize their other products and hurt their bottom line.


This means that, at least for the time being, Fadr is the only AI VST to generate a collection of percussive one shots. Fortunately, the product's audio quality and interface make it a fun, practical tool for whipping up new drum sounds on the fly.


Check out our article on AI drum VSTS for a broad overview of DAW plugins that apply machine learning to percussion samples. Those other tools do not offer text prompting interface but they may have something of interest to you.


You can visit Fadr's website to learn more about their product suite.

 
 
bottom of page