● Covered by 1 source · 1 reportLow impact1 neutral
Maker compresses 2.9MB song to 21KB using AI codec, prints it on paper as QR codes
Key points
- A 2.9MB MP3 was compressed to 21KB using Meta's EnCodec, a 1000x reduction.
- The 21KB data was printed onto eight QR codes on a sheet of paper.
- The same 21KB file was also transferred wirelessly between two ESP32 boards via LoRa.
- Playback requires the specific 24 kHz EnCodec neural network decoder.
AI Compression for Physical Storage
Makestreme, a maker, successfully compressed a two-minute, 2.9MB MP3 song into a 21KB file using Meta's EnCodec neural audio codec. This represents a compression ratio of approximately 1,000 times. The compressed data was then printed onto a sheet of paper as eight QR codes, arranged to resemble a double-sided "paper cassette."
EnCodec Technology
Meta's open-source EnCodec, released in 2022, converts audio waveforms into discrete tokens. A matching decoder reconstructs the audio from these tokens. At a 3 kbps setting, the reconstructed song maintained quality, but at 1.5 kbps, the output was significantly degraded. This technology allows for substantial data reduction compared to conventional lossy compression like MP3.
QR Code Implementation
Each of the eight QR codes holds a 1-byte ID and a segment of the token stream, with four codes per side. The project utilized the lowest error correction level to maximize data density, meaning any damage to a single QR code would render the song unplayable. A single QR code can store approximately 3KB of binary data, with Makestreme reaching a practical limit of about 3.3KB.
Wireless Transfer Experiment
The 21KB file was also transmitted wirelessly between two ESP32 boards equipped with REYAX RYLR998 modules over LoRa radio at 866 MHz. This transfer, using 160-byte packets and an ACK-and-retry protocol, took approximately 90 minutes due to a conservative 40-second pause after each acknowledged packet, implemented for duty-cycle compliance.
Playback Dependency and Historical Context
Similar to historical methods like the 1985 Cauzin Softstrip, which printed software as 2D barcodes, this project's playback is dependent on specific technology. The QR codes store tokens, not raw audio, requiring EnCodec's exact 24 kHz decoder for playback. This means the song's longevity is tied to the continued availability and functionality of the neural network decoder.
✨ This summary was generated by AI from the outlets' reporting listed below. It is not independently verified and may contain errors — check the original sources. How BrevFeed works →
The daily brief
One email each morning: the day's tech stories, clustered across outlets and summarized. No account needed.
One email a day. Unsubscribe in one click, any time.
Today's brief
Spend a few minutes, get the whole day. Every topic's top stories in one hands-free rundown — listen, watch, or read the transcript.
▶ Play today's brief
New every morning, and the back catalogue is archived by date.
Reporting from
A maker named Makestreme used Meta's EnCodec neural audio codec to compress a 2.9MB MP3 song to 21KB, then printed it across eight QR codes on paper. This project demonstrates extreme audio compression using AI and explores novel data storage and transfer methods, though playback relies on the specific neural network decoder.