Research and infrastructure for generative audio.
ViMu is a research and infrastructure company with Vienna as home base. We design and train our own generative audio models and run them on our own servers at extreme speed. A handful of partners build their products on top of them.
Active programmes: · ·
01 — What we do
Research, inference and the products that happen in between.
Research
Over a decade of experience in machine learning research in music technology. Frontier generative audio and music models built on our own custom model IP, trained on dedicated hand-crafted sessions recorded with our partners.
Inference
We run our own GPU servers in Europe and the US. The serving stack is built for audio, where every bit of latency is something you feel.
Products
With our partners we turn the models into APIs, plugins and apps. Under their brand, ours, or both.
02 — Engagement
How we start working together.
- 01 Write to us. Tell us roughly what you want to build. A few lines is fine.
- 02 NDA. We sign one before either side shares any details.
- 03 Briefing. We show you what we have, what it can do and what your product would need.
- 04 Build and launch. We build the models and the integration, run the inference, and launch it with you.
03 — Principles
How we handle your work and your data.
Confidential
Partner projects stay under NDA until the partner decides to talk about them.
We don't keep your input
Input lives on the render server only while the render runs. We don't store it and we don't train on it.
Rendered in your region
Requests are processed and delivered in the user's region, EU or US.
Training data
We don't scrape. We train on sessions we record in collaboration with our partner companies. We train our own custom models from scratch. If the product needs base model adaptations, we customize the models to fit the product.
04 — Team
Four founders, a big data team.
Vienna
Although we're an international company, Vienna is our home base, and a city with a long history in music.
100 years
of music and ML R&D between our four founders. Each of us has spent our careers in leading roles, pushing the field forwards.
Data team
A large in-house team of data specialists, including engineers, composers, conductors and musicians — ensures extreme levels of detail and quality.
We already work with several established companies in professional audio.
Names under NDA
End of public file
The rest is better told in person.
Get an introduction, or just write to us.