AudioCodex: fully causal streaming autoencoder compresses stereo audio to ~10Hz latents for live generation
serrjoa · x · 2026-11-12
ISMIR 2026 paper "AudioCodex: Streaming Autoencoders for Live Audio" (Marco Pasini et al.): a fully causal autoencoder compressing stereo audio into compact 10Hz continuous latents, strictly causal in both encoding and decoding, targeting real-time interactive/streaming generative audio. Combines a Mean Flow magnitude autoencoder with latent hypersphere projection, bounded-noise augmentation, representation alignment, and adversarial phase recovery; extensive ablations show it beats the state of the art on the generation-reconstruction trade-off.
More from Research
- Anthropic scaled 1,000 Claude agents to formalize Fermat's Last Theorem — SimplyAnnisa · 2026-11-12
- NeurIPS paper gives RNNs provable state-estimation guarantees via control-theory observers — chriswolfvision · 2026-11-12
- NeurIPS paper proves when RNN+MLP can provably recover true robot state — chriswolfvision · 2026-11-12
- Curated 9-paper reading list for LLM quantization beginners, from QLoRA to BitNet — East-Muffin-6472 · 2026-11-12
- A Quick Diagnostic Guide to 4 Common ML Training Errors and Their Causes — cneuralnetwork · 2026-11-12
- DLS runs 1.5x-2.4x faster than tangermeme on A100, verified within numerical precision — anshulkundaje · 2026-11-12