It’s becoming increasingly clear that not every company is being upfront about how their technology is built, which understandably makes producers skeptical.
Because of that, we thought this was the right time to give you a behind-the-scenes look at how Spawn’s AI is actually trained - and why you can trust that it’s ethical, artist-first, and consent-driven.
What is ethical AI?
Ethical AI means our AI is trained on licensed, consent-based data from real creators, not scraped or unverified sources.
Contributors explicitly agree to how their work is used, are compensated for their contributions, and retain ownership of their music. We believe producers should know how the tools they rely on are built, so we aim to make our approach as transparent and verifiable as possible.
Who built Spawn?
Spawn is a partnership between Sauceware Audio and Lemonaide Music.
The Lemonaide team is responsible for the MIDI generation inside Spawn. Lemonaide is made up of former Google and Amazon scientists, and their CEO, MJ, first connected with Sauceware in 2022 to talk through what was coming in the AI music world - including the exact ethical concerns producers are now raising.
To this day, Lemonaide remains one of the few companies in the space that can confidently say their models are trained ethically, with proper licensing, consent, and compensation. This is why they’ve worked with respected producers like Lex Luger, and why Forbes featured Lemonaide for prioritizing artists and consent in AI music.
If you want to go deeper, MJ, Ani, Saif, and Julian from Lemonaide put together a video breaking down their approach to AI: Watch it here
What part of Spawn is AI?
Spawn uses AI specifically for MIDI generation.
We’ve taken a deliberately narrow approach to AI so it acts as a creative assist, not a replacement for producers. Limiting AI’s role also allows its training and behavior to be clearly documented and verified.
All sound design, presets, workflow, and overall product direction are developed by Sauceware Audio.
How can I trust what you're saying?
Lemonaide’s models are Fairly Trained certified. This certification helps publicly verify that training data is licensed, consent-based, and does not rely on copyrighted material without permission.
Beyond certification, we’re making our approach as publicly verifiable as possible by documenting how the AI is trained and what standards we follow - including this page.
Below is a plain-language overview of the data and standards used to train the AI models inside Spawn. Where relevant, we link to supporting documentation so you can dig deeper if you want.
Training data overview
The AI models used in Spawn are trained on a curated dataset of MIDI created by real producers.
The dataset does not include:
Scraped public MIDI repositories
Unlicensed commercial music catalogs
MIDI generated by other AI systems
Unverified third-party sources
Dataset source categories
The training dataset is composed of the following categories:
Licensed contributor MIDI
MIDI licensed directly from individual creators who explicitly agreed to its use for AI training.
Commissioned / custom-created MIDI
Original MIDI created specifically for training purposes by commissioned contributors.
Contributor consent
All contributors whose material is used in training explicitly agree to:
Their MIDI being used to train AI models
The scope of that use (training only, not redistribution or resale)
The terms under which their work is licensed
Contributor agreements are consent-driven and documented.
Contributor compensation
All contributors are compensated for their participation, whether their work is licensed or custom-commissioned.
Compensation structures may include:
One-time license fees
Commission fees for custom-created material
Specific compensation amounts are private, but compensation is provided for every contribution used in training.
Legal review & rights
Contributor agreements are reviewed by legal counsel and grant Lemonaide the right to use contributed material for AI training, while allowing contributors to retain ownership of their underlying works.
Output originality & duplication safeguards
Spawn is designed to generate new, original MIDI, not to reproduce or retrieve training data.
Training data is not directly accessible to users or the model at runtime. The system is designed to reduce memorization and direct duplication, though all generative systems operate probabilistically.
Royalty-free use of your output
MIDI generated using Spawn is royalty-free and may be used commercially by the user, subject to the terms outlined in our Terms of Service.
Please note that some music libraries and platforms apply their own review or detection processes for submissions. We recommend reviewing a library’s specific terms and guidelines before submitting your work.