Entirely Ethical AI Vocal Synthesis.
NeoDIVA is a group dedicated to creating unique AI Vocalists, utilising OpenVPI's DiffSinger engine.
NeoDIVA's "Lead" vocals are designed to provide the essential foundations for AI vocal synthesis, offering a diverse range of masculine, feminine, and androgynous tones optimized for multi-genre performance.
At NeoDIVA, we believe in the strength of community. We are committed to uplifting creators and expanding our lineup. If you'd like to collaborate, submit your vocal for consideration as a NeoDIVA feature.
Project Lead
Lead Developer & Project Lead. Alex is the mastermind behind the NeoDIVA project structure and training pipelines.
Audio & Art
Key Artist and Music Producer. They illustrate all of the key art for the NeoDIVA LEAD vocals.
Design
Character Designer. They create all of the new designs for the NeoDIVA LEAD vocals, and collaboration designs.
/// SELECT UNIT FOR ANALYSIS
License: CC BY-NC 4.0. Free for non-commercial projects. Credit "NeoDIVA". Resulting embeddings must not be public. Licensing fees apply for exceptions.
WARNING: Use of RVC/SVC for inference/training on this voice is PROHIBITED without explicit exception.
"NeoDIVA"s are specialized DiffSinger voicebanks trained within the larger NeoDIVA "Megamodel".
NeoDIVA "LEAD" vocals are our core collection. "FEATURE" vocals are community-submitted vocalists trained and hosted by us.
Yes. Our models are trained only on vocal data provided with explicit consent for this specific purpose. We do not scrape data. Training is performed locally, minimizing carbon footprint.
The NeoDIVA model supports 14 languages:
English, Japanese, Thai, Korean, Italian, Spanish, Chinese, Portuguese (BR), Georgian, Romanian, Polish, Russian, French and Greek
Currently, due to dataset licensing, commercial usage is NOT allowed. This may change in the future.
/// ETHICS & SUSTAINABILITY PROTOCOLS
NeoDIVA operates on a strict "Opt-In Only" policy. We do not scrape data from the internet. Every dataset used in our "Megamodel" is either provided by the vocalist with explicit, written consent for AI training, or is a made-for-purpose research dataset that was also created without scraping data.
We believe AI should amplify human creativity, not replace it. Our models are designed to be instruments—tools for composers to realize vocal performances that were previously impossible.
Training AI models consumes significant energy, even on local, consumer-grade, hardware. We believe in total transparency regarding our carbon footprint.
To mitigate this impact, we accurately track the power consumption of every GPU hour used and we donate monthly to Offset Britain to ensure our project remains carbon neutral (or negative).
By completing this form, you acknowledge that your dataset will be considered for training as a NeoDIVA Feature. Feature vocals are distinct from Lead vocals and serve to promote the community ecosystem.
Mandatory: Read Terms of Service before proceeding. Access Document Here
The DiffSingers showcased on the NeoDIVA site belong to and are managed by their individual creators. Our platform acts as a catalyst for character growth.
Submit now!