English

Beyond Reconstruction: Full-Context Generative DiT for Music Generation

Audio and Speech Processing 2026-08-09 v1

Abstract

Hybrid music generators combine the long-range planning of an autoregressive language model with the fidelity of a diffusion- or flow-based acoustic renderer. Yet renderers are trained with clean, target-derived codec tokens but deployed with imperfect language-model predictions, creating codecinterface exposure bias. Rather than treating rendering as a simple reconstruction task,we formulate it as full-context generation from an imperfect discrete plan.We introduce FullDiT, a conditional DiT that fuses eight frame-aligned RVQ streams with independently encoded captions and lyrics and uses non-causal self-attention over the complete acoustic latent sequence. During training, Error-Matched Distractor Conditioning (EMDC) matches per-codebook replacement rates to teacher-forced top-1 error rates and samples near-miss tokens from cosine-KNN neighborhoods without changing the acoustic target. At inference, four-way classifier-free guidance (4-CFG) independently scales codec, lyric, and caption guidance increments. Matched ablations show that EMDC improves ViSQOL by 0.77 under synthetic corruption and is clearly preferred in non-tied comparisons with fixed languagemodel tokens. Further ablations show gains from full-song context and renderer-side text conditioning. The complete system outperforms five commercial systems on 15 of 18 automatic metrics and ranks among the top three on the Artificial Analysis Music with Vocals Leaderboard. The demo page is available at https://selinacloudl.github.io/fulldit-demo/.

Keywords

Cite

@article{arxiv.2608.08787,
  title  = {Beyond Reconstruction: Full-Context Generative DiT for Music Generation},
  author = {Yunjia Li and Menglin Wu and Junyu Dai and Xinyue Fan and Xiangang Li and Haoxu Wang and Jianwei Yu and Huaicheng Zhang and Han Zhao and Weiqin Li and Yufei Shi and Cheng Wen and Sitong Zhao and Qixi Zheng and Haina Zhu and Wei Li},
  journal= {arXiv preprint arXiv:2608.08787},
  year   = {2026}
}