Building the infrastructure music needs today
Scroll to explore
Groundwork for What Comes Next
CORPUS develops the tools to check, describe, discover and license music for a future in which people and machines create together, software chooses much of what is heard, and music responds to the moment it is played in.
Checks, descriptions and search, built for our own library and now licensed to catalogues.
A legal basis that pays rights holders and gives companies models they can use with confidence.
Musicians and rights holders decide how their music is used, and every licence pays them.
The same principles apply wherever models are built from what people contribute, from language to health data.
How We Listen
More music is released every day than anyone could listen to, and a growing share of it is generated by machines. Our integrity checks and rich semantic descriptions, built with musicians, sound engineers and musicologists, give every track the same attention, however small its audience.
Verify
Screened for damage, duplicates, covers and generated material, and flagged for review before problems surface later.
Understand
Described in depth: its sound, lyrics, emotional character and the scenes it can carry.
Discover
Search by musical criteria, a reference track, a description or a creative brief, and refine it in conversation.
Modular APIs connect verification, description and search to the platforms and workflows you already have.
For Catalogues and Libraries
Your sync team pitches what it remembers. We help you get value from the rest of your repertoire too.
A full check of your catalogue for technical faults, duplicates, covers and generated material, before a buyer finds them.
Descriptions and search for your own team, on your website or through our API, so a brief reaches beyond the titles everyone remembers.
If you choose, your music on Soundbook and in chat assistants such as ChatGPT and Claude, where productions search and soon license.
You decide where your music appears and on what terms. We set it up with you.
For Film, TV and Game Makers
Stop falling for temp tracks you can’t keep.
Describe a scene or a feeling, name a genre, a tempo or an instrument, or play a reference track. Soundbook asks back and narrows it down with you until the track is there.
Play it against your cut and share a shortlist with your team, not another WeTransfer link full of MP3s.
Composition and recording together, with licences from festival run to worldwide release. Soon in minutes, not weeks of emails.
For Musicians
You make the music. Finding the people who need it shouldn’t be your second job.
The finished album, the forgotten cue, the track on your hard drive. Upload once, set your terms, keep your rights. Non-exclusive.
No genre boxes, no playlist to game. We describe what your music does, so the right scene finds it.
You switch each use on yourself: film and games, chat assistants, products that listen, training our own models. No fees for you.
For Products That Listen
Built by Sofilab, the Munich sound design lab that started CORPUS and has designed sound for Audi, Harman, Brainlab and Linde. Sofilab is developing sound that responds to the moment and is looking for pilot partners.
Scores that follow the player, made from human performances.
A night drive and a tense junction don’t have to sound the same. Cabin atmospheres that set the mood and guide attention.
A bedside device that sings along with songs from someone’s own life, to support carers.
Built from Practice
CORPUS is developed by Sofilab and co-financed by Creative Europe.
We have written music for feature films, documentaries and games. We have sat on both sides of the brief, and we know what it takes to deliver.
We run our own and open-weight models on our servers in Germany, under EU law, not on someone else’s AI service. Your scripts, briefs and music stay private in Soundbook.
The Corpus Is Open
CORPUS is in public beta. Here is where we stand:
- Integrity checks
- In-depth descriptions
- Search and exploration, try it on soundbook.fm →
- Rights chain: permissions and ownership for every track, in progress. Once it is in place, we leave the beta.
Follow along
Updates from the corpus, now and then.
The question we keep working on
What if musicians had a stake in the machines that learn from their music?
Cookie Settings
We use cookies to enhance your browsing experience, serve personalized content, and analyze our traffic. By clicking "Accept All", you consent to our use of cookies.