Khala: Scaling Acoustic Token Language Models Toward High-Fidelity Music Generation

• URL: http://arxiv.org/abs/2605.01790v1
• Date: Sun, 03 May 2026 09:13:20 GMT
• Title: Khala: Scaling Acoustic Token Language Models Toward High-Fidelity Music Generation
• Abstract summary: A common design pattern in high-quality music generation is to handle structure and fidelity in different representation spaces.<n>In this work, we explore an alternative view: both may be progressively modeled within a single deep acoustic-token hierarchy.<n>Our results suggest that high-quality music generation can be effectively pursued without separating structure and fidelity into heterogeneous representation spaces.
• Score: 31.997336001394284
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: A common design pattern in high-quality music generation is to handle structure and fidelity in different representation spaces: a generator first models high-level structure, followed by diffusion-based or neural decoding stages that reconstruct fine details. In this work, we explore an alternative view: both may be progressively modeled within a single deep acoustic-token hierarchy. To study this, we build a 64-layer residual vector quantization (RVQ) acoustic representation and propose a two-stage coarse-to-fine generation framework. A backbone model first generates coarse acoustic tokens for the full track, and a super-resolution model then completes finer tokens within the same acoustic token space. The super-resolution stage works at full-track scale and refines tokens layer by layer while running in parallel over time, leading to a fixed 62-step inference process. To jointly improve lyric alignment and fine-detail reconstruction, we further introduce hybrid-attention training: the alignment objective uses causal attention, while layer-wise refinement uses full attention. A key finding is that text--vocal alignment can emerge within pure acoustic-token language modeling, without requiring a separate semantic token stage. Moreover, initializing the super-resolution model from the trained backbone significantly improves convergence and final quality. Taken together, our results suggest that high-quality music generation can be effectively pursued without separating structure and fidelity into heterogeneous representation spaces. Instead, both can be progressively modeled within a unified acoustic-token hierarchy, pointing toward a simpler and more unified path to high-quality music generation.

Related papers

• OmniVAE: An Audio-Video VAE with Cross-Modal Alignment for Joint Generation [73.57253966578186]
  We present OmniVAE, a jointly trained audio-video VAE that learns fine-grained semantic alignment between audio and video latent representations.<n>Experiments show that both objectives consistently improve the learnability of the latent spaces.
  arXiv  Detail & Related papers  (2026-07-26T21:51:28Z)
• HiTokSR: A Coarse-to-Fine Tokenizer with Hierarchical Codebooks for High-Fidelity Real-World Image Super-Resolution [2.4519602914420475]
  We present HiTokSR, a hierarchical token prediction framework.<n>We show that HiTokSR achieves state-of-the-art performance in both quality and reconstruction fidelity.
  arXiv  Detail & Related papers  (2026-05-31T11:02:58Z)
• HoliTok:A Coutinuous Holistic Tokenization with Robust Dual Capabilities of Speech Generation and Understanding [27.487275558312774]
  We propose a continuous Holistic speech Tokenization model designed for unified generation-understanding modeling.<n>HoliTok encodes 48kHz speech into a compact 25Hz sequence of 128-dimensional latents.<n> Experiments show that HoliTok achieves competitive reconstruction fidelity, improves generative learnability for high-quality and controllable synthesis.
  arXiv  Detail & Related papers  (2026-05-28T13:55:19Z)
• SymphonyGen: 3D Hierarchical Orchestral Generation with Controllable Harmony Skeleton [26.272034375469943]
  SymphonyGen is a 3D hierarchical framework for cinematic orchestration.<n>We introduce "short-score" conditioning via a beat-quantized multi-voice harmony skeleton.<n>We implement a dissonance-averse sampling algorithm to suppress unintended tonal clashes.
  arXiv  Detail & Related papers  (2026-04-28T11:01:21Z)
• Talker-T2AV: Joint Talking Audio-Video Generation with Autoregressive Diffusion Modeling [59.97930201616015]
  We propose Talker-T2AV, an autoregressive diffusion framework where high-level cross-modal modeling occurs in a shared backbone.<n>A shared autoregressive language model jointly reasons over audio and video in a unified patch-level token space.<n> Experiments on talking portrait benchmarks show Talker-T2AV outperforms dual-branch baselines in lip-sync accuracy, video quality, and audio quality.
  arXiv  Detail & Related papers  (2026-04-26T07:48:47Z)
• The World is Not Mono: Enabling Spatial Understanding in Large Audio-Language Models [17.675850481660863]
  We introduce a hierarchical framework for Auditory Scene Analysis (ASA)<n> guided by this framework, we introduce a system that enables models like Qwen2-Audio to understand and reason about the complex acoustic world.<n>Our work provides a clear pathway for leveraging the powerful reasoning abilities of large models towards holistic acoustic scene analysis.
  arXiv  Detail & Related papers  (2026-01-06T11:54:47Z)
• Scaling Self-Supervised Representation Learning for Symbolic Piano Performance [52.661197827466886]
  We study the capabilities of generative autoregressive transformer models trained on large amounts of symbolic solo-piano transcriptions.<n>We use a comparatively smaller, high-quality subset to finetune models to produce musical continuations, perform symbolic classification tasks, and produce general-purpose contrastive MIDI embeddings.
  arXiv  Detail & Related papers  (2025-06-30T14:00:14Z)
• Combining audio control and style transfer using latent diffusion [1.705371629600151]
  In this paper, we aim to unify explicit control and style transfer within a single model. Our model can generate audio matching a timbre target, while specifying structure either with explicit controls or through another audio example. We show that our method can generate cover versions of complete musical pieces by transferring rhythmic and melodic content to the style of a target audio in a different genre.
  arXiv  Detail & Related papers  (2024-07-31T23:27:27Z)
• Simple and Controllable Music Generation [94.61958781346176]
  MusicGen is a single Language Model (LM) that operates over several streams of compressed discrete music representation, i.e., tokens. Unlike prior work, MusicGen is comprised of a single-stage transformer LM together with efficient token interleaving patterns.
  arXiv  Detail & Related papers  (2023-06-08T15:31:05Z)
• The Power of Reuse: A Multi-Scale Transformer Model for Structural Dynamic Segmentation in Symbolic Music Generation [6.0949335132843965]
  Symbolic Music Generation relies on the contextual representation capabilities of the generative model. We propose a multi-scale Transformer, which uses coarse-decoder and fine-decoders to model the contexts at the global and section-level. Our model is evaluated on two open MIDI datasets, and experiments show that our model outperforms the best contemporary symbolic music generative models.
  arXiv  Detail & Related papers  (2022-05-17T18:48:14Z)
• DiffSinger: Diffusion Acoustic Model for Singing Voice Synthesis [53.19363127760314]
  DiffSinger is a parameterized Markov chain which iteratively converts the noise into mel-spectrogram conditioned on the music score. The evaluations conducted on the Chinese singing dataset demonstrate that DiffSinger outperforms state-of-the-art SVS work with a notable margin.
  arXiv  Detail & Related papers  (2021-05-06T05:21:42Z)
• Any-to-Many Voice Conversion with Location-Relative Sequence-to-Sequence Modeling [61.351967629600594]
  This paper proposes an any-to-many location-relative, sequence-to-sequence (seq2seq), non-parallel voice conversion approach. In this approach, we combine a bottle-neck feature extractor (BNE) with a seq2seq synthesis module. Objective and subjective evaluations show that the proposed any-to-many approach has superior voice conversion performance in terms of both naturalness and speaker similarity.
  arXiv  Detail & Related papers  (2020-09-06T13:01:06Z)
• Learning Interpretable Representation for Controllable Polyphonic Music Generation [5.01266258109807]
  We design a novel architecture, that effectively learns two interpretable latent factors of polyphonic music: chord and texture. We show that such chord-texture disentanglement provides a controllable generation pathway leading to a wide spectrum of applications.
  arXiv  Detail & Related papers  (2020-08-17T07:11:16Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.