MP3 audio format
The audio compression format developed largely at Fraunhofer IIS in Erlangen, fixed as an MPEG standard in 1992 and named .mp3 in 1995, which made digital music portable.
- Made by
- Fraunhofer-Gesellschaft
- From
- Munich, Bavaria
- Category
- Software & AI
- Introduced
- 1992
What it is
MP3, formally MPEG-1 Audio Layer III, is a method of compressing digital audio so that files take up a fraction of the space of uncompressed CD audio while sounding close to the original. Researchers at the Fraunhofer Institute for Integrated Circuits IIS in Erlangen played a central role in its development, working with partners including the University of Hanover and Thomson. The system was fixed at an MPEG meeting in London in November 1992. On 14 July 1995 the Fraunhofer team chose .mp3 as the file extension, replacing the in-house .bit.
How it works
MP3 relies on psychoacoustics, the science of how humans hear. Our ears cannot perceive some sounds, for instance quiet tones close in frequency to louder ones. The encoder analyses the audio, identifies information that listeners are unlikely to notice and discards it, then codes the remaining signal efficiently. The result can be a large reduction in file size with little audible loss.
Why it stands out
MP3 changed the economics of music. Small files could be stored on computers, sent over the early internet and carried in pocket-sized players. That enabled file sharing, legal download stores and, eventually, streaming. The project grew out of an idea by Dieter Seitzer of transmitting high-quality music over ISDN telephone lines and was carried forward by his doctoral student Karlheinz Brandenburg, Heinz Gerhäuser and others at Fraunhofer IIS. It shows how applied research in Germany can reshape a global industry.
Where you meet it
MP3 files are still found in music libraries, podcasts, audiobooks and in-car systems. Newer formats such as AAC, which Fraunhofer also helped develop, have taken over much streaming, but MP3 remains a universal format that almost every device can play.
Stories
Related products
Celonis process mining
Software from a Munich university spin-off that reconstructs how business processes really run from the event data in systems such as SAP, then helps fix them.
DeepL Translator
Neural machine translation from Cologne, released in August 2017 and known for translations that read naturally rather than word for word.
FLUX image models
Text-to-image AI models from Black Forest Labs of Freiburg, released from August 2024 by the researchers behind Stable Diffusion, in open and commercial versions.
Last reviewed 4 November 2026. Main source: heise.de/-10485563. Suggest a correction.