Scaling Open Discrete Audio Foundation Models with Interleaved Semantic, Acoustic, and Text Tokens

• URL: http://arxiv.org/abs/2602.16687v1
• Date: Wed, 18 Feb 2026 18:32:46 GMT
• Title: Scaling Open Discrete Audio Foundation Models with Interleaved Semantic, Acoustic, and Text Tokens
• Authors: Potsawee Manakul, Woody Haosheng Gan, Martijn Bartelds, Guangzhi Sun, William Held, Diyi Yang,
• Abstract summary: Current audio language models are predominantly text-first, either extending pre-trained text LLM backbones or relying on semantic-only audio tokens.<n>This paper presents a systematic empirical study of native audio foundation models that apply next-token prediction to audio at scale.
• Score: 62.56027815951259
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Current audio language models are predominantly text-first, either extending pre-trained text LLM backbones or relying on semantic-only audio tokens, limiting general audio modeling. This paper presents a systematic empirical study of native audio foundation models that apply next-token prediction to audio at scale, jointly modeling semantic content, acoustic details, and text to support both general audio generation and cross-modal capabilities. We provide comprehensive empirical insights for building such models: (1) We systematically investigate design choices -- data sources, text mixture ratios, and token composition -- establishing a validated training recipe. (2) We conduct the first scaling law study for discrete audio models via IsoFLOP analysis on 64 models spanning $3{\times}10^{18}$ to $3{\times}10^{20}$ FLOPs, finding that optimal data grows 1.6$\times$ faster than optimal model size. (3) We apply these lessons to train SODA (Scaling Open Discrete Audio), a suite of models from 135M to 4B parameters on 500B tokens, comparing against our scaling predictions and existing models. SODA serves as a flexible backbone for diverse audio/text tasks -- we demonstrate this by fine-tuning for voice-preserving speech-to-speech translation, using the same unified architecture.

Related papers

• UALM: Unified Audio Language Model for Understanding, Generation and Reasoning [124.19449187588832]
  Unified Audio Language Model (UALM) aims to unify audio understanding, text-to-audio generation, and multimodal reasoning in a single model.<n>We first present UALM-Gen, a text-to-audio language model that directly predicts audio tokens and is comparable to state-of-the-art diffusion-based models.<n>We present UALM-Reason, a multimodal reasoning model that utilizes both text and audio in the intermediate thinking steps to facilitate complex generation tasks.
  arXiv  Detail & Related papers  (2025-10-13T22:55:01Z)
• Generative Audio Language Modeling with Continuous-valued Tokens and Masked Next-Token Prediction [63.26850431270348]
  We research audio generation with a causal language model (LM) without discrete tokens.<n>We leverage token-wise diffusion to model the continuous distribution of the next continuous-valued token.<n>We propose a novel masked next-token prediction task that incorporates masked prediction into the causal LM framework.
  arXiv  Detail & Related papers  (2025-07-14T00:14:54Z)
• Baichuan-Audio: A Unified Framework for End-to-End Speech Interaction [9.101978573666546]
  Baichuan-Audio is an end-to-end audio large language model that seamlessly integrates audio understanding and generation.<n>It features a text-guided aligned speech generation mechanism, enabling real-time speech interaction with both comprehension and generation capabilities.
  arXiv  Detail & Related papers  (2025-02-24T15:16:34Z)
• C3LLM: Conditional Multimodal Content Generation Using Large Language Models [66.11184017840688]
  We introduce C3LLM, a novel framework combining three tasks of video-to-audio, audio-to-text, and text-to-audio together. C3LLM adapts the Large Language Model (LLM) structure as a bridge for aligning different modalities. Our method combines the previous tasks of audio understanding, video-to-audio generation, and text-to-audio generation together into one unified model.
  arXiv  Detail & Related papers  (2024-05-25T09:10:12Z)
• AudioFormer: Audio Transformer learns audio feature representations from discrete acoustic codes [6.375996974877916]
  We propose a method named AudioFormer, which learns audio feature representations through the acquisition of discrete acoustic codes. Our research outcomes demonstrate that AudioFormer attains significantly improved performance compared to prevailing monomodal audio classification models.
  arXiv  Detail & Related papers  (2023-08-14T15:47:25Z)
• AudioLDM 2: Learning Holistic Audio Generation with Self-supervised Pretraining [46.22290575167155]
  This paper proposes a framework that utilizes the same learning method for speech, music, and sound effect generation. Our framework introduces a general representation of audio, called "language of audio" (LOA)
  arXiv  Detail & Related papers  (2023-08-10T17:55:13Z)
• CLIPSonic: Text-to-Audio Synthesis with Unlabeled Videos and Pretrained Language-Vision Models [50.42886595228255]
  We propose to learn the desired text-audio correspondence by leveraging the visual modality as a bridge. We train a conditional diffusion model to generate the audio track of a video, given a video frame encoded by a pretrained contrastive language-image pretraining model.
  arXiv  Detail & Related papers  (2023-06-16T05:42:01Z)
• BEATs: Audio Pre-Training with Acoustic Tokenizers [77.8510930885778]
  Self-supervised learning (SSL) has been witnessed in language, vision, speech, and audio domains over the past few years. We propose BEATs, an iterative audio pre-training framework to learn Bidirectional representation from Audio Transformers. In the first iteration, we use random projection as the acoustic tokenizer to train an audio SSL model in a mask and label prediction manner. Then, we train an acoustic tokenizer for the next iteration by distilling the semantic knowledge from the pre-trained or fine-tuned audio SSL model.
  arXiv  Detail & Related papers  (2022-12-18T10:41:55Z)
• Matching Text and Audio Embeddings: Exploring Transfer-learning Strategies for Language-based Audio Retrieval [11.161404854726348]
  We present an analysis of large-scale pretrained deep learning models used for cross-modal (text-to-audio) retrieval. We use embeddings extracted by these models in a metric learning framework to connect matching pairs of audio and text.
  arXiv  Detail & Related papers  (2022-10-06T11:45:14Z)

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.