Archive
Multimodal & Vision

The Two-Letter Audio Model

Archive
Hard
200pts0 solves
Meta's neural audio codec that compresses speech and music to tens of tokens per second and decodes back to high-fidelity waveforms — used inside AudioGen, MusicGen, and SpeechTokenizer. Name the codec. Flag format: CONGRESS{name}. Example: CONGRESS{soundstream}.
Show hint
A prefix about encoding + a suffix for codec.

Archive — no submissions accepted

This challenge is preserved for reference. Play live challenges at /challenges.