The paper investigates how much original audio information remains in frozen representations from various pretrained audio encoders by attempting to reconstruct five‑second, 44.1‑kHz stereo music clips. Using a shared Stable Audio Open latent diffusion decoder, the authors reconstruct audio from embeddings produced by supervised classifiers (VGGish, ConvNeXt), an audio‑text contrastive model (CLAP), and a waveform‑reconstruction model (EnCodec). Results on the Million Song Dataset reveal that reconstructability varies across encoder families and that embeddings exposing finer temporal or spectral detail enable better recovery of source content, even for compressed, task‑oriented embeddings.
arXiv:2607. 03806v1 Announce Type: cross Abstract: Audio foundation models are widely adopted as general-purpose feature extractors, yet the internal structure of their learned representations remains insufficiently understood.
SAGE, a 105‑million‑parameter semantic audio generative encoder, compresses music waveforms into compact latent representations by distilling embeddings from a pretrained audio‑text model. It achieves a three‑way balance—high reconstruction quality, rich semantic latent structure, and fast inference—running three times faster than Stable Audio Open while matching or surpassing larger models in listening‑test quality and objective metrics. Across all nineteen latent‑semantics probing tasks, SAGE sets new state‑of‑the‑art performance both within and outside its training domain.
arXiv:2608. 14819v1 Announce Type: cross Abstract: Music foundation models are commonly used as frozen audio feature extractors, yet selecting which layer to extract from remains largely heuristic.
The paper introduces a supervised end-to-end benchmark for restoring orchestral historical music, addressing the lack of ground-truth pairs in early‑20th‑century recordings. It simulates the historical degradation process more accurately than prior work and trains a latent flow‑matching model that surpasses existing HMR baselines in various evaluations. Additionally, the authors release a 9.3‑hour license‑free test set, code, and audio demos for the community.
The paper introduces a lightweight technique to steer the pitch content of audio generated by the Stable Audio Open diffusion model. A small convolutional probe (~125k parameters) is trained to decode frame‑level pitch‑class activations from the model’s latent space using paired audio and MIDI data. During inference, the frozen probe acts as a differentiable loss, guiding generation toward a user‑specified pitch‑class sequence without retraining the base model, and improves melodic coherence by 2.4× over the unguided baseline.
arXiv:2607. 08545v1 Announce Type: cross Abstract: End-to-end neural audio models achieve high-fidelity compression and generation.
arXiv:2607. 00777v1 Announce Type: cross Abstract: Recognizing jazz standards from audio is a challenging form of tune-level music retrieval: different performances of the same standard may vary in tempo, key, arrangement, instrumentation, improvisational content, and even whether the head melody is present.
DuoTok is a source‑aware dual‑track music tokenizer designed for vocal‑accompaniment generation. It first learns a semantic audio representation via self‑supervised pretraining, then refines source‑aware structure with feature‑replacement noise and multi‑task supervision (spectral reconstruction, source separation regularization, and an ASR head for lyric alignment). The encoder is frozen and hard‑routed codebooks for vocals and accompaniment are learned, while a diffusion decoder restores fine acoustic detail from the discrete tokens, achieving a favorable predictability‑fidelity trade‑off at ultra‑low bitrate across public benchmarks.
arXiv:2511. 05350v3 Announce Type: replace-cross Abstract: We argue that training autoencoders to reconstruct inputs from noised versions of their encodings, when combined with perceptually motivated losses, yields encodings that are structured according to a perceptual hierarchy.
arXiv:2605.27840v2 Announce Type: replace-cross Abstract: Audio tokenizers are fundamental to unifying audio understanding and generation. Understanding requires high-level semantics, while generatio...
The paper introduces geometric iterative retrieval, a new approach for resynthesizing high‑quality audio from coarse Residual Vector Quantization (RVQ) codec tokens. Instead of choosing between discrete token prediction or continuous regression, the method performs contrastive retrieval within the continuous codebook space, leveraging the RVQ hierarchy as an iterative decomposition. Experiments on speech and music codec restoration tasks demonstrate that this technique outperforms both single‑pass token prediction and one‑step regression baselines.
arXiv:2608. 04142v1 Announce Type: cross Abstract: Existing reference-free methods for evaluating music perceptual quality alleviate the need for paired noisy-clean data, but they still rely on a background set, which is used to compute aggregated statistics of clean audio samples.