Services
In-house AI tools for brands and agencies – from inaudible watermarks to automated radio spots. AI is a tool, not an end in itself: people make the decisions.
What it is about
Many tasks around a brand sound are routine: variants, languages, formats, searching, checking. That is exactly where AI saves time – if the tools fit the task.
That is why 42 Sounds builds its own tools: Python workflows that slot into existing productions – from voice synthesis to a Pro Tools integration.
Solutions
Know where a sound runs – and prove where it comes from.
Spots, music and sound logos get an inaudible identifier that a database recognises later – as proof of airing and rights, and as machine-readable labelling of AI content under the EU AI Act. Built into the in-house product AD-ID.
One brand sound becomes many versions – fast and consistent.
Text, AI voice and music bed automatically become a fully mixed spot – in variants for regions, offers or periods.
A brand voice that sounds the same in many languages. On request, the models run locally on in-house hardware – without the cloud.
Videos with music and voice in series: edit, music, speech and mix work together automatically.
The right sound in seconds instead of hours.
AI detects cuts and actions in the picture and suggests matching sounds and Foley – right inside Pro Tools. Topic of the talk at FMX 2026.
Samples and sound effects are recognised and tagged automatically. Search works in plain language, e.g. “door slams shut, reverberant”.
The tool finds tracks from a library that match the pictures, tempo and mood of a video.
Own product
Every ad. Identified. Hear it. Know it.
AD-ID recognises spots by their sound. Point the phone at the audio source – TV, radio, laptop or an in-store speaker – and AD-ID tells which spot is playing.
The same song runs for two brands. An audio fingerprint alone cannot tell them apart – the watermark can: AD-ID recognises which spot is playing for which brand and campaign. To test: play a track here, open ad‑id.42sounds.de on a phone, tap and point it at the speaker.
Verified original
Verified original
Under the hood
Three examples, step by step – from input to result.
Listen
For a fictional car dealership: text, AI voice, AI-generated music and mix – generated automatically.
Working together
The tools run at 42 Sounds, the brand receives finished results – spots, variants, analyses.
The tool runs at the brand or agency itself – with setup and onboarding.
Tailored to a specific task – from the first prototype to ongoing operation.
Process
Where does routine cost time today? That becomes a clear goal for the tool.
A first working draft with real material – early enough to decide together.
The tool runs alongside a real project and is sharpened there.
Integration into existing workflows and software – with maintenance and further development.
On stage
FMX 2026 · Stuttgart
Photo: Dominique Brewing
Custom AI tools read the edit, evaluate keyframes and suggest matching sounds and MIDI – as a Python workflow that fits into existing post-production. In the sound design track alongside Nina Hartstone and David Fluhr.
View in the FMX program
Workshop, 60 minutes, moderated by André Klar.
FAQ
No. AI speeds up variants and routine work. The idea, the selection and the fine-tuning stay handmade.
An inaudible identifier in the audio signal. It cannot be heard, but software can read it out and match it to a database – for example to prove airings or to label AI-generated content in a machine-readable way.
Only with licensed voices and the consent of the speakers. Usage rights and labelling under the EU AI Act are settled from the start.
Not necessarily. Many tools run locally on in-house hardware – for example for voices and audio that should not leave the company.
Yes. Besides use as a service, there are licences for in-house use and custom-built solutions.
A mix of established AI models and in-house Python tools built for each task – for example for shot detection, sound recognition, voice synthesis and the Pro Tools integration.
Contact
First idea or finished brief – a short message is all it takes.